JPH10142490A - Environment recognition device and camera - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately recognize environment even when objects small in the difference of distance exist. SOLUTION: Even when plural objects 11 and 12 partially coming close to each other exist, the existence of the individual objects frequently appears as a separate peak in frequency distribution (peak-D and peak-E) in the case the average distances of the individual objects are different. By taking notice of this point, the frequency distribution of the information is made based on the information on the distance distribution of field space and area dividison is performed based on the information on the frequency distribution.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an environment recognizing apparatus and an environment recognizing apparatus for recognizing an environment by dividing a screen into individual objects based on distance distribution information of the outside world and evaluating each area. The present invention relates to improvement of a camera provided with a recognition device.

[0002]

2. Description of the Related Art Japanese Patent Publication No. 4-67607 by the present applicant discloses a technique for optically measuring the distance to an object existing in a plurality of directions. According to this technique, distance distribution information of an object existing in a scene can be obtained, and it is possible to recognize an object arrangement state of the scene from the distribution information.

[0003] A typical environment recognition method which has been conventionally performed will be described below.

A scene such as that shown in FIG.
Shoot with a stereo camera using Then, the two images having parallax obtained by the stereo camera are each divided into “m × n” blocks, and a signal in a certain block of an image captured by one camera and a signal captured by the other camera are used. When a known correlation operation is performed between the signals in the corresponding block, the distance to the object in the previous block and the defocus can be measured by the principle of triangulation. By performing this measurement for all blocks, distance distribution information composed of “m × n” blocks as shown in FIG. 21B is obtained.

[0005] Next, region division (grouping) is performed to separate each object constituting the scene on the screen. When this grouping is performed, the object space composed of the aforementioned “m × n” blocks is divided into regions for each object as shown in FIG. 21C (the hatched portions in FIG. 21 indicate insufficient contrast of the image signal). In this case, it is determined that the reliability of the correlation operation result is low.

As a method of area division (grouping),
There is a method of comparing the similarity between two parameters relating to a block constituting the object space and a block adjacent thereto, and determining that the similar object is the same if the similarity is high and another object if the similarity is low. The information used as the parameter is often a surface normal vector when dense distance distribution data is obtained, and is simply a distance value when the distance distribution data is relatively rough as in the conventional example. And a defocus value.

For example, the distance information of each block shown in FIG. 21B is compared with the distance information of two adjacent blocks on the right and left. It is determined that the constituent objects form the same object. "If the distance difference is greater than a predetermined value, it is determined that the objects forming the two blocks are different objects. By performing the above-described determination between all blocks and blocks adjacent to each other, the entire screen can be divided into regions for each object, and each divided region is treated as a group representing one object be able to.

As described above, the technique of recognizing the environment by dividing the screen into regions for each object includes, for example, an automatic traveling robot that determines its own traveling direction, and automatically avoids danger by judging obstacles before and after. Available to do cars.

Further, the main subject is automatically recognized by evaluating each area and detecting a main object in the screen, and a camera that focuses on the main subject, or a photometer of only the main subject, which is measured under a backlight condition. However, it can be used in a wide range of technical fields, such as a camera that can properly adjust the exposure to the main subject, and an air conditioner that determines the direction in which a person is present in a room and appropriately controls air blowing.

[0010]

Conventionally, when region division is performed based on relatively rough distance distribution information, it is determined whether a difference in distance (or defocus amount) between two adjacent regions is equal to or less than a predetermined value. As described above, it is determined whether the two blocks are the same object. However, there are cases where accurate region division cannot be performed by such a method.

[0011] For example, when two humans are arranged back and forth so that a part of the body overlaps, even if the average distance difference between the two persons is equal to or more than the predetermined value, only a part of the body is Just in contact with each other, the distance difference in this portion may be equal to or less than the above-described predetermined value, and the two objects may be combined as one region.

In addition, since the distance is obtained by the correlation operation, when two object images are reflected in the same operation block, the distance between the two objects may be derived. The distance depends on the pattern of the image signal). When such a state referred to as “perspective conflict” occurs, even if the distance difference between the two objects is equal to or greater than the predetermined value, an intermediate distance value is calculated in a part of the boundary, and as a result, the two In some cases, objects are grouped into one object and divided into regions.

As described above, in the conventional method, there is a case where the object arrangement in the object scene cannot be accurately recognized. Therefore, the recognition of the main object is erroneous. For example, in the case of a camera, the calculation of the focus adjustment amount and the exposure amount In some cases, calculations could not be performed accurately.

(Object of the Invention) A first object of the present invention is to use the frequency distribution information of distances for area division, thereby enabling accurate environment recognition even in a situation where an object having a small distance difference exists. It is an object of the present invention to provide an environment recognition device that can perform the operation.

A second object of the present invention is to provide an environment recognizing apparatus capable of detecting a maximum value of distance frequency distribution information and setting a boundary value before and after the maximum value, thereby enabling accurate environment recognition. It is in.

A third object of the present invention is to evaluate the sharpness of the extreme value of the frequency distribution information of the distance, and to use only the information of the extreme value having a predetermined sharpness for the area division, thereby obtaining a more accurate value. An object of the present invention is to provide an environment recognition device capable of performing environment recognition.

A fourth object of the present invention is to provide an environment recognizing apparatus capable of performing more accurate environment recognition by using the minimum value of distance frequency distribution information for area division.

A fifth object of the present invention is to provide an environment recognizing apparatus capable of performing accurate environment recognition by using frequency distribution information of defocus for area division.

A sixth object of the present invention is to provide an environment recognizing apparatus capable of detecting a local maximum value of the defocus frequency distribution information and setting a boundary value before and after the local maximum value, thereby enabling more accurate environmental recognition. Is to do.

A seventh object of the present invention is to evaluate the sharpness of the extreme value of the frequency distribution information of defocus, and to use only the information of the extreme value having a predetermined sharpness for region division.
It is another object of the present invention to provide an environment recognition device capable of performing accurate environment recognition.

An eighth object of the present invention is to utilize the minimum value of the defocus frequency distribution information for area division,
It is another object of the present invention to provide an environment recognition device capable of performing accurate environment recognition.

A ninth object of the present invention is to provide a camera capable of detecting a main subject region from an environment recognition result obtained by an environment recognition device and setting appropriate photographing parameters based on the main subject region detection result. Is to do.

A tenth object of the present invention is to adjust the focus with respect to the detected main subject so that the photographer can accurately determine the main subject without being conscious of the focus adjustment position and automatically determine the main subject. It is to provide a camera that can be focused.

An eleventh object of the present invention is to adjust the exposure based on the detected light amount of the main subject area,
An object of the present invention is to provide a camera capable of accurately adjusting the exposure of a main subject even under adverse conditions such as backlight.

A twelfth object of the present invention is to adjust the focal length based on the detected information of the main subject region, so that the photographer can adjust the focal length without adjusting the focal length. An object of the present invention is to provide a camera capable of automatically setting a corner.

[0026]

In order to achieve the first object, the present invention according to the first or ninth aspect of the present invention provides an image processing system in which a screen is divided into individual objects based on distance distribution information of the outside world. In an environment recognition apparatus comprising an area dividing means for dividing and performing environment recognition based on an output of the area dividing means, the area dividing means creates distance frequency distribution information based on the distance distribution information, The distribution information is used as one of the determination factors for region division.

More specifically, a frequency distribution of the information is created from the distance distribution information of the object space. In this frequency distribution, for example, even when there are a plurality of objects that are only partially close to each other, as long as the average distance of each object is different, the existence of each object is Focusing on the fact that they often appear as separate peaks, the region is divided based on information on frequency distribution.

In order to achieve the second object, the present invention according to claim 2 or 9, detects a maximum value of the frequency distribution information, and sets a boundary value at a distance separated by a predetermined amount before and after the maximum value. Is set, and the area is divided based on the information of the boundary value.

In order to achieve the third object, the present invention according to claim 3 or 9 evaluates the sharpness of all the detected extreme values, and evaluates the extreme value having a sharpness not less than a predetermined value. Only the value information is used as one of the determination elements for the area division.

In order to achieve the fourth object, the present invention according to claim 4 or 9, detects a minimum value of the frequency distribution information, sets a boundary value with a distance of the minimum value, and sets a boundary value of this boundary value. The region is divided based on the information.

In order to achieve the fifth object, the present invention according to claim 5 or 10 divides a screen into regions based on distribution information of defocus of an object field as if each screen were divided into individual objects. In an environment recognition apparatus comprising the area dividing means and performing environment recognition based on an output of the area dividing means, the area dividing means creates frequency distribution information of defocus based on the defocus distribution information, The frequency distribution information is used as one of the determination elements for the area division.

More specifically, a frequency distribution of the information is created from the defocus distribution information in the object space. In this frequency distribution, for example, even when there are a plurality of objects that are only partially adjacent to each other, as long as the average defocus of each object is different, the existence of each object is Focusing on the fact that they often appear as separate peaks in the distribution, the region is divided based on the information on the frequency distribution.

In order to achieve the sixth object, according to the present invention, a maximum value of the frequency distribution information is detected and a defocus value separated by a predetermined amount before and after the maximum value is detected. A boundary value is set, and area division is performed based on the information of the boundary value.

In order to achieve the seventh object, the present invention according to claim 7 or 10 evaluates the sharpness of all the detected extreme values, and evaluates the extreme having a sharpness not less than a predetermined value. Only the value information is used as one of the determination elements for the area division.

In order to achieve the eighth object, the present invention according to claim 8 or 10, detects the maximum value and the minimum value of the frequency distribution information, and defocuses the minimum value existing before and after the maximum value. The value is set as a boundary value, and the area is divided based on the information of the boundary value.

According to a ninth aspect of the present invention, there is provided an environment recognition apparatus according to the ninth or tenth aspect, and an area of a main subject is obtained from an environment recognition result of the environment recognition apparatus. And a main subject detecting means for detecting the

In order to achieve the tenth object, the present invention according to the twelfth aspect has an automatic focus adjusting means for focusing on a main subject based on an output result of the main subject detecting means. .

In order to achieve the eleventh object, the present invention according to a thirteenth aspect is directed to a measuring means for measuring a light amount of a main subject area based on an output result of a main subject detecting means, An automatic exposure amount adjusting means for adjusting the exposure amount is provided.

In order to achieve the twelfth object, according to the present invention, the focal length of the lens is adjusted so that the main subject has an appropriate size based on the output result of the main subject detecting means. The automatic focus adjustment means is provided.

[0040]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail based on illustrated embodiments.

A first embodiment of the present invention for recognizing an object arrangement state of an object scene will be described with an example of an automatic ranging point selection function of a camera.

FIG. 1 is an arrangement diagram of optical components of a camera for detecting a defocus or distance distribution in a photographing screen. In the drawing, reference numeral 1 denotes a photographing lens, 2 denotes a field lens, and 3 denotes a secondary connection. The image lens 4 is an area sensor.
Light beams from different pupil positions of the photographing lens 1 are guided onto the two imaging screens 4a and 4b of the area sensor 4, respectively, and re-formed at an imaging magnification determined by the field lens 2 and the secondary imaging lens 3. Imaged. The area sensor 4 is located at a position optically equivalent to the photographic film surface with respect to the photographic lens 1, and the imaging screens 4a and 4b each have a part of the photographic screen or a field of view equal to the photographic screen.

FIG. 2 shows a layout when the detection optical system shown in FIG. 1 is applied to a camera.
5 is a quick return mirror, 6 is a pentaprism, 7
Is a split prism, 8 is a reflection mirror, and the other is the same as FIG.

FIG. 3 is a view of the layout of FIG. 2 as viewed from above the camera.

With the above configuration, the imaging screens 4a and 4b having a predetermined parallax can be obtained.

The camera having the above configuration is disclosed in detail in Japanese Patent Application No. 5-278433.

FIG. 4 is a circuit diagram showing an example of a specific configuration of a camera provided with each of the above devices. First, the configuration of each unit will be described.

In FIG. 4, a PRS controls various controls of the camera, for example, a CPU (central processing unit) and an R
This is a one-chip microcomputer (hereinafter, referred to as a microcomputer) having an OM, a RAM, and an A / D conversion function. The microcomputer PRS performs a series of camera operations such as an automatic exposure control function, an automatic focus adjustment function, and film winding / rewinding in accordance with a camera sequence program stored in the ROM. For that, the microcomputer PRS
Communication signals SO, SI, SCLK, communication selection signal CLC
By using M, CDDR, and CICC, communication is performed with peripheral circuits in the camera body and a control unit in the lens described later to control the operation of each circuit and lens.

SO is a data signal output from the microcomputer PRS, SI is a data signal input to the microcomputer PRS, and SCLK is a synchronous clock of the data signals SO and SI.

LCM is a lens communication buffer circuit,
When the camera is in operation, power is supplied to the lens power supply terminal VL and a selection signal C from the microcomputer PRS is supplied.
When the LCM is at a high potential level (hereinafter, abbreviated as “H” and the low potential level is abbreviated as “L”), it serves as a communication buffer between the camera and the lens.

When the microcomputer PRS sets the selection signal CLCM to "H" and sends out predetermined data as SO in synchronization with the clock SCLK, the lens communication buffer circuit L
The CM is connected to the SCL via the camera-lens communication contact.
The buffer signals LCK and DCL of K and SO are output to the lens. At the same time, the buffer signal of the signal DLC from the lens is output as SI, and the microcomputer PRS outputs SC.
The lens data is input as SI in synchronization with LK.

DDR is a circuit for detecting and displaying various switches SWS, is selected when the signal CDDR is "H", and is controlled by the microcomputer PRS using SO, SI and SCLK. That is, based on data sent from the microcomputer PRS, the display of the display member DSP of the camera is switched, and the on / off state of various operation members of the camera is notified to the microcomputer PRS by communication. OLC is an external liquid crystal display device located above the camera, and ILC is a liquid crystal display device in the viewfinder.

SW1 and SW2 are switches linked to a release button (not shown). The switch SW1 is turned on by pressing the release button in the first stage, and the second switch is subsequently turned on.
The switch SW2 is turned on by pressing down in stages. Microcomputer PR
In step S, photometry and automatic focus adjustment are performed when the switch SW1 is turned on, and exposure control and subsequent film winding are performed with the switch SW2 turned on as a trigger.

The switch SW2 is connected to the "interrupt input terminal" of the microcomputer PRS. Even when a program is executed when the switch SW1 is turned on, an interrupt is generated by turning on the switch SW2, and control is immediately transferred to a predetermined interrupt program. Can be.

MTR1 is for film feeding, MTR2 is for mirror up / down and shutter spring charging,
Each is a motor, and each motor drive circuit MDR
1, MDR2 controls forward rotation and reverse rotation. The signals M1F, M1R, M2F, M2R input from the microcomputer PRS to the motor drive circuits MDR1, MDR2 are
This is a signal for motor control.

MG1 and MG2 denote magnets for starting the first and second curtains of the shutter, respectively.
Is supplied through the amplification transistors TR1 and TR2, and the microcomputer PRS controls the shutter.

The motor drive circuits MDR1, MDR
2 and shutter control are not directly related to the present invention, and therefore, detailed description is omitted.

A signal DCL input to a lens control circuit (hereinafter referred to as a lens microcomputer) LPRS in synchronization with LCK is command data from the camera to the lens LNS. It is predetermined. The microcomputer LPRS in the lens analyzes the instruction in accordance with a predetermined procedure, and performs the operation of focus adjustment and aperture control, the operation status of each part of the lens by the DLC (driving status of the focus adjustment optical system, driving status of the aperture, etc.) and various parameters. (An open F number, a focal length, a defocus amount versus a coefficient of a moving amount of the focus adjusting optical system, various focus correction amounts, and the like) are output.

This embodiment shows an example of a zoom lens. When a focus adjustment command is sent from a camera, a focus adjustment motor LTMR is sent to a signal LMF in accordance with the simultaneously transmitted drive amount and direction. , LMR to adjust the focus by moving the optical system in the direction of the optical axis.
The amount of movement of the optical system is monitored by a pulse signal SENCF of an encoder circuit ENCF that detects a pattern of a pulse plate that rotates in conjunction with the optical system with a photocoupler and outputs a number of pulses corresponding to the amount of movement. The microcomputer LPRS in the lens counts with a counter in the microcomputer LPRS, and when the predetermined movement is completed, the microcomputer LPRS in the lens itself outputs signals LMF and LMR.
Is set to “L” to brake the focus adjustment motor LMTR.

For this reason, once the focus adjustment command is sent from the camera, the microcomputer PRS on the camera side does not need to be involved in driving the lens at all until the driving of the lens is completed. Also, if there is a request from the camera,
The content of the counter can be transmitted to the camera.

When an aperture control command is sent from the camera, a known stepping motor DMTR for driving the aperture is driven in accordance with the number of aperture stages sent at the same time.

Since the stepping motor can perform open control, it does not require an encoder for monitoring the operation.

ENCZ is an encoder circuit associated with the zoom optical system, and the microcomputer LPRS in the lens receives the signal SENCZ from the encoder circuit ENCZ to detect the zoom position. The lens microcomputer LPRS stores lens parameters at each zoom position, and when requested by the microcomputer PRS on the camera side, sends parameters corresponding to the current zoom position to the camera.

The ICC is a sensor / drive circuit including a focus detection and exposure control (photometry) area sensor composed of a CCD or the like and a drive unit therefor. The ICC is selected when the signal CICC is "H". It is controlled by the microcomputer PRS using SO, SI, and SCLK.

.Phi.V, .phi.H, and .phi.R are signals for reading the outputs of the two area sensors and reset signals, and a sensor control signal is generated by the drive unit in the sensor / drive circuit ICC based on the signal from the microcomputer PRS. . The sensor output is amplified after reading from the sensor unit, and the output signal IM
The signal is input to the analog input terminal of the microcomputer PRS as AGE, and the microcomputer PRS converts the signal into a digital signal.
The digital values are sequentially stored at predetermined addresses on the RAM. Then, using these digitally converted signals, environment recognition and focus adjustment or photometry of the object scene are performed.

In FIG. 4, the camera and the lens are shown as being separate (lens can be replaced).
There is no problem even if the camera and lens are integrated, and the invention is not limited to these.

FIG. 5 shows an operation of a camera having a function of recognizing an object arrangement state using a frequency distribution of distances in the object space, detecting a main subject, and then focusing a lens on the main subject. In particular, it is a flowchart showing an operation of selecting a distance measuring point for recognizing an environment and determining a main subject, and will be described below.

When the photographer presses the release button, the switch SW1 is turned on, and the CPU in the microcomputer PRS switches to RO.
According to the program recorded in M, control according to the flowchart of FIG. 5 is started.

In step (101), a sensor image is captured. The capture of the sensor image is executed as follows.

First, the sensor is reset. More specifically, the control signals φV, φH, and φR are simultaneously set to “H” for a certain period of time by the microcomputer PRS, so that the sensor / drive circuit I
A reset operation is performed inside the CC. Next, the microcomputer P
An accumulation start command is sent from the RS to start accumulation, and the end of accumulation is detected later. Then, the control signals φV and φH are driven to sequentially read the sensor output IMAGE, and the microcomputer PRS
A / D converted and stored in RAM. Thus, the capture of the sensor image in this step (101) is completed. The output signal data of the two sensors is stored in predetermined areas IMG1 and IMG2 on the RAM.

In the next step (102), “m
A distance distribution information (distance map) composed of “× n” blocks (m and n are integers equal to or greater than 1) is created. The creation of the distance map executed here will be described with reference to the flowchart of FIG.

In step (201) of FIG. 6, variables x and y indicating the coordinates of the block are initialized. Then, in the next step (202), block (x,
The signal necessary for the distance calculation of y) is stored in the image data I on the RAM.
It is extracted from MG1 and copied to a predetermined address A on the RAM. In the following step (203), another signal necessary for the distance calculation of the block (x, y)
G2 is extracted and copied to a predetermined address B on the RAM. Next, in step (204), a well-known correlation operation COR (A, B) is performed on the luminance distribution signals recorded at the address A and the address B, and a shift amount δ between the two image signals is calculated.

In the next step (205), the defocus or the calculation of the distance value is executed from the above-mentioned image shift amount δ by a known function f (δ), and a predetermined value secured for the distance distribution recording on the RAM is recorded. The distance value or the defocus is stored at the address D (x, y) of the data. Here, a case where the distance value is calculated will be described.

In the next step (206), the value of x is increased by one, and the processing target is moved to an adjacent block. Then, in step (207), x is compared with the resolution m of the distance map in the x direction, and when it is determined that “x <m” is true (meaning that this relationship is satisfied), the previous step (207) is performed. Returning to 202), the calculation and storage of the distance value are repeated for the adjacent block in the x direction in the same manner as described above.

On the other hand, in the above step (207),
If it is determined that “x <m” is false (meaning that this relationship is not satisfied), the process proceeds to step (208), x is initialized, and y is incremented by one. Then, the next step (209)
In step (2), the value of y is evaluated, and when it is determined that "y <n" is true, the process returns to step (202) again to start the operation for the next block sequence. When it is determined that “y <n” is false, the distance calculation for all the blocks is completed, and the creation of the distance map ends.

Returning to the description of FIG. 5, the next step (10
In 3), a frequency distribution is created. Hereinafter, creation of the frequency distribution will be described with reference to the flowchart of FIG.

In step (301), the variable H secured in the RAM for creating frequency distribution data
The contents of (0) to H (t _num -1) are cleared to 0. Here, H (t) (where 0 <t <t _num ) represents the number of data existing within a predetermined distance range represented by t.
t _num represents the number of frequency distribution data.

An example of a method for setting the distance range will be described.

The distance detection range to be calculated is, for example, 10
In the case of cm to 100 m, a method of dividing this distance range into equal parts, for example, every 1 m, and dividing the area into 100 parts can be considered (in this case, t _num = 100).

Since the depth of focus of the lens increases in inverse proportion to the distance, the distance range is converted to a logarithmic space to reflect this characteristic.
_{2.0~2.0 (= log 10 0.01~log 10 100} ) as a method of 20 divided areas separated this range every 0.1 are considered (in this case, t _num = 20).

In addition, it is of course possible to perform some kind of non-linear division according to the characteristics and use of the device.

In this embodiment, a case will be described in which regions are divided so as to be equally spaced in logarithmic space as described above.

In the next step (302), variables x and y indicating the coordinates of the block are initialized. In the following step (303), the distance value D (x, y) of the block (x, y) is included. A distance range is searched for, and a number t representing the range is determined. When the distance range has complicated nonlinear characteristics, the distance range is searched and examined one by one, but in the case of the present embodiment, the distance ranges are equally spaced in logarithmic space as described above. Is divided into
t can be directly obtained by t = INT (log ₁₀ (D (x, y) / d _min ) / s). Here, INT is a function of rounding down to the nearest whole number, and s is a range width in logarithmic space (s = 0.1 in the above example). Also, d _min is the minimum value of the distance detection range (d _min = 10 cm in the case described above), and is used for adjusting the closest range number to 0.

In the next step (304), the value of the frequency distribution corresponding to t is increased by one. Then, in step (305), the value of x is incremented by one and the process is shifted to an adjacent block. In the following step (306), x is compared with the resolution m in the distance map x direction,
If it is determined that “x <m” is true, the process returns to step (303), and the generation of the frequency distribution is continued for the next block as described above.

On the other hand, if it is determined in step (306) that "x <m" is false, the process proceeds to step (307), where x is initialized and y is incremented by one. Then, in step (308), the value of y is evaluated, and “y <
When it is determined that "n" is true, the process returns to step (303) again, and the calculation for the next block example is started. Also,
When it is determined that “y <n” is false, the work of reflecting the data of all the blocks in the frequency distribution data is completed, and the creation of the frequency distribution ends.

Although not explicitly shown in the flowchart of FIG. 7, in the case of a block for which the reliability of the correlation calculation is determined to be low due to a low contrast or the like, the reliability of the distance value of the calculation result naturally also increases. Because of its low, the data of these blocks need not be added to the frequency distribution at all.

In such processing, when creating a distance map, the reliability of the operation result of each block is recorded in an array variable, and when creating a frequency distribution, the reliability of the block is referred to by referring to the value of the array variable. In the case of a low block, data registration to the frequency distribution (steps (303) and (304) in FIG. 7) may be passed, and can be easily realized. Is omitted.

Returning to the description of FIG. 5, step (10)
In 4), an extremum in the frequency distribution, particularly a maximal value in the case of the present embodiment, is detected. For example, when a histogram as shown in FIG. 8 is obtained, the local maximum value p
The distance values of peak-A, peak-B, and peak-C are detected.

The method of this detection will be described with reference to the flowchart of FIG.

In step (401), the numerical value u representing the maximum value number is initialized. Next step (402)
In, the numerical value t representing the data number of the frequency distribution is initialized to 1. In the following step (403), the value H (t) of the frequency distribution and the value H (t-
1) and H (t + 1) are checked to see if H (t) forms a vertex. As a result, H (t)> H (t
-1) and if H (t)> H (t + 1), then H
(T) can be determined to form a local maximum,
Move to step (404). Then, this step (4)
At 04), a representative value of the distance represented by t is stored in the extreme value storage variable P (u) secured on the RAM. As the representative value, an appropriate value at the center of the distance range represented by t is appropriate, for example, given as “10 ^{s (t + 0.5)} · d _min ”.

In the next step (405), the numerical value u representing the number of the vertex is incremented by one, and in the following step (405), the numerical value t representing the data number of the frequency distribution to be detected is incremented by one. Change the frequency distribution of the target. Then, in step (406), whether or not t is within the range of the frequency distribution data, that is, “t <
examine whether it is a t _num "," t <t _num "it is if it is true, because it is the way the search of the extreme step (403)
If “t <t _num ” is false, it is determined that the search has been completed, and the process proceeds to the next step (407).

In step (407), the number of vertices actually registered for later processing is stored in a variable u _num , and the vertex detection processing ends.

Returning to the description of FIG. 5, when the detection of the maximum point is completed as described above, the process proceeds to the step (105) for setting the boundary value. This boundary value is determined in the next step (106).
Is a value that is referred to when performing region division by.

The concept of setting a boundary value in the present embodiment will be described.

In the histogram, there is a high possibility that an object exists around the distance of the peak portion. Therefore, it is appropriate to recognize the predetermined distance range before and after the peak as one group. That is, two predetermined values R
f and Rb (both positive real numbers) are set in advance, and “d−R−R” is set as a boundary value on the short distance side with respect to the distance d of the peak position.
If "f" and "d + Rb" are set as the boundary values on the long distance side, the object can be appropriately divided into regions. For example, when the detection target of the main object is limited to a human, particularly in the case of a camera or the like, since a human often faces the camera, the depth is considered to be at most about 50 cm, and Rf and Rb are set to about 30 cm, respectively. It is possible.

Considering from the viewpoint of dividing a region for each object, it is possible to provide a non-linear property such that an object at a short distance has a fine division with respect to a distance, and an object at a long distance has a large division width. It is effective, and in order to realize this effect, it is appropriate to determine a boundary value by adding or subtracting a predetermined value in a logarithmic space. For example, "log ₁₀ (d) -R
A method of setting a boundary value in logarithmic space as “f” and “log ₁₀ (d) + Rb” can be considered. This is equivalent to setting the boundary values to “d × 10 ^−Rf ” and “d × 10 ^Rb ” at the actual distance, and the above-described effect is obtained. Of course, it is obvious that “d × 10 ^−Rf ” and “d × 10 ^Rb ” can be directly obtained and set as the boundary values at the actual distance.

As described above, by setting the boundary values before and after the maximum point, it is possible to divide the area as a group of the objects reflected on the maximum point.

An example will be described. When the processing described so far is performed on the scene shown in FIG. 10A, three scenes are obtained as shown in FIG.
Peak-D, peak-E, and peak-F
Is generated. peak-D and p
The peak-E is formed by the subjects 11 and 12, respectively, and the peak-F is formed by the background. peak-
When boundary values are set for D, peak-E, and peak-F, border-Df, border-D in FIG.
r, border-Ef, border-Er, bor
The distance represented by der-Ff and border-Fr is the boundary value. In the subsequent region division, if these boundary values exist between two distances between adjacent blocks, it is determined that the boundary between the adjacent blocks is the boundary of the object. By performing this processing between blocks constituting the adjacent relationship of all the screens, in the case described above, the three subjects 11, 1
2 and the background can be clearly separated.

Furthermore, when a part of two objects is close to each other, an erroneous determination occurs because the conventional method divides an area based on a mere distance difference between adjacent blocks. The block with a small distance difference is shown in FIG.
Since the object exists in the area-A distribution of (b), the two objects can be clearly separated. In other words, even when two objects are close to each other, the presence of each object often appears as a separate peak in the frequency distribution, as long as the average distance of each object is different. And focus on the frequency distribution information to determine the boundary value of the area division. Therefore, even if there is an object with a small distance difference, or even if there is a perspective conflict, accurate Can be separated into two objects.

The setting of the boundary value performed in step (105) will be described with reference to the flowchart of FIG.

In step (501), a numerical value u representing the number of the maximum value is initialized, and in the next step (502), “P (u) ·
10 ^−Rf ”to store the boundary value in the memory S (2
Record in u). In the following step (503),
“P (u) · 10 ^Rb ” as the boundary value on the long distance side of the maximum value
Is stored in the memory S (2u + 1). Then, in step (504), the value of the numerical value u is increased by 1 in order to proceed to the processing of the next maximum value.

In the next step (505), the value of u is compared with the maximum number u _num of the maximum values obtained in the preprocessing to check whether or not the processing for all the maximum values has been completed. As a result, if “u <u _num ”, the process is not completed, and the process returns to step (502). If “u <u _num ” is false, the processing for all extreme values is completed.

By the above processing, all the boundary values become R
The memories S (0) to S (2u _num −
It is stored in 1).

Returning to the description of FIG. 5 again, when the setting of the boundary value is completed as described above, next, in step (106), the area is actually divided.

For example, as shown in FIG. 12, when performing division processing while performing raster scanning from the upper left block of the screen as indicated by the arrow in the figure, the block G (x, y) above the block of interest G (x, y) , Y−1) and the left block G (x−1,
y), if it is determined whether the group is the same,
As a result, it can be determined whether or not all adjacent blocks are the same block. At this time, the blocks on the upper side (y = 0) and the left side (x = 0) of the screen do not have the upper block and the left block, respectively, so that no processing is performed on them.

The result of the judgment is stored in the memory G on the RAM.
Record in (0,0) to G (m-1, n-1). First,
The block of (x, y) = (0, 0) has a group number g =
Registered as 1, and if a group having a different area is detected, the number of g is increased by one and set as the group number of the block.

By this processing, for example, a photographing scene as shown in FIG. 10A is given a number for each group as shown in FIG.

Since the numbering process itself is a known technique called “labeling method”, a flowchart of the entire area division is omitted, but it is one of the key points of the present embodiment. The “algorithm for determining whether adjacent blocks are in the same group” performed in the description will be described below.

FIG. 14 shows that two blocks are the same object or different objects between a block of interest G (x, y) being scanned and a block G (x−1, y) on the left. It is a flowchart at the time of determining whether it is.

[0110] In step (601), the block of interest G (x, y) distance value D (x, y) of the variable d ₁ for work, in the next step (602), comparison block G (x The distance value D (x−1, y) of (−1, y) is copied to the variable for work d ₂ .

In the next step (603), a numerical value u representing a boundary value number and a flag Flag representing whether or not two blocks are the same are initialized. "Flag = 0"
Indicates that the two objects are the same, and “Flag =
"1" indicates that the two objects are distinct.

In the following step (604), it is checked whether or not a boundary value S (u) exists between the two distances to be compared. “D ₁ <S (u) and S (u) <d ₂ ” is a condition for “d ₁ <d ₂ ”, and d ₂ <S (u) an
d S (u) <d ₁ "is a condition in the case of" d ₂ <d ₁ ".

If the result of the logical sum is true, it can be determined that the two blocks are separate objects. Therefore, in the next step (605), a flag is set and the processing is terminated.

On the other hand, if the logical sum is false, the value of u is increased by one in step (606) in order to compare with the next boundary value. Then, in the next step (607), u
Is checked to see if it exceeds the number of boundary values, and if not, the flow returns to step (604) to check for the next boundary value. If it is over, the check ends with Flag = 0, and the two blocks are ended as the same object.

The above determination is made for all the adjacencies of all the blocks, and the area division (grouping) is completed.

Returning to the description of FIG. 5, the next step (1)
In step 07), the characteristics of each region (each group) constituting the shooting space are evaluated, and a group representing the main subject is determined from all the groups.

For example, in the case of FIG.
The average distance for each of the ~ 17 groups,
Characteristics such as the width, height, position on the screen, and the like of the region are calculated, and are comprehensively evaluated to determine a region considered as a main subject.

For example, a main subject degree evaluation function represented by the following equation (1) can be considered.

(Main subject degree) = W ₁ × (width) × (height) + W ₂ / (distance from screen center) + W ₃ / (average distance) (1) In the above equation (1) _{Te, W 1, W 2, W} 3 is a constant weighting, the distance from the center is the distance between the center of gravity position of the screen center and the regions, the average distance represents the average distance of all blocks in the area . The main subject degree is calculated for all areas, and the subject having the highest main subject degree is determined as the main subject.

The area representing the main subject is, for example, as shown in FIG.
After determining 11 in (a), in step (108), the “focusing distance” is determined based on the distance information within this area,
The lens is driven so as to focus on this distance. As a method of determining the focusing distance, the average of the distances of all the blocks included in the area, the distance of the nearest block in the area, and the like can be considered.

As described above, the calculations from step (102) to step (108) are performed in the microcomputer PRS.

In the following step (109), a command for focus adjustment is sent from the microcomputer PRS to the lens LNS so that the focus is adjusted to the distance determined in step (108). The main object is focused by controlling the motor LMTR, and the focus adjustment on the main object is completed.

As described above, the screen is divided into regions using the frequency distribution information of distances, divided into groups for each object, and a group considered as the main subject is selected from all the groups, and the object is selected. By driving the lens so as to focus on, the photographer can obtain a photograph in which the main subject is focused only by focusing on the composition without being conscious of the position of the main subject.

Further, in this embodiment, by setting the boundary value of the area division before and after the maximum value of the frequency distribution of the distance, it becomes possible to separate a plurality of objects that could not be separated by the conventional method. , Can accurately recognize the environment. As a result, in the case of the camera as in the embodiment,
An accurate determination of the focusing distance can be performed.

Further, in this embodiment, the result of environment recognition is used for focus detection. However, by determining the exposure based on the luminance information of the area considered as the main subject, it can be used under adverse conditions such as a backlight condition. Even so, a photograph in which the main subject is exposed can be obtained.

When the main subject area is known, the focal length is adjusted by driving the zoom motors ZMF and ZMR so that the main subject occupies 50% of the area of the screen, and the main subject is automatically set. The automatic zoom function that adjusts the angle of view to the angle can be easily realized.

In the above description, the case where distance distribution information is used has been described. However, similar processing can be performed using the defocus amount instead of the distance as an element of the distribution information.

Further, in the present embodiment, the distance map is created for the blocks on the “m × n” cells. However, the distribution of the distance measurement blocks does not necessarily have to be a cell shape. However, the same processing can be performed without any problem.

(Second Embodiment) A second embodiment of the present invention will be described.
In this embodiment, the detection accuracy of the local maximum value is further improved with respect to the first embodiment described above, and the entire processing flow is in accordance with FIG.

In the second embodiment, the method of detecting the maximum value in step (104) in FIG. 5 is different.

The point to be detected as the maximum value is indicated by p in FIG.
A peak having a certain degree of sharpness, such as peak-G, peak-H, and peak-J, is desirable. The maximum value caused by a slight fluctuation, such as peak-I, depends on the presence of a specific object. Since it is often not a generated peak, it is desirable not to detect it as a local maximum.

Therefore, in a second embodiment of the present invention,
The sharpness of the maximum value is evaluated, and only the extreme value having a sharpness equal to or more than a predetermined value is used for determining a region division point later.

This maximum value evaluation will be described with reference to the flowchart of FIG.

In step (701), the numerical value u representing the number of the maximum value is initialized, and in the next step (702), the numerical value t representing the data number of the frequency distribution is initialized to 1. Then, in step (703), the value H (t) of the frequency distribution, the values H (t−1) of the frequency distribution before and after,
Compare with H (t + 1) to see if H (t) forms a vertex. As a result, H (t)> H (t-1)
If H (t)> H (t + 1), it can be determined that H (t) forms a local maximum, and the flow proceeds to step (704). If the maximum value has not been formed, the process proceeds to step (709).

Step (704) and Step (705)
, Variables a and b in which the difference between the detected local maximum value and the value of the frequency distribution separated by a predetermined value e before and after is stored in the RAM.
Copy to Then, in step (706),
These values are compared with a threshold value T, and if any of the values exceeds the threshold value, it is determined that the peak has a predetermined sharpness, and the local maximum value registration after step (707) is performed. Move to In step (706), if none of the values a and b exceeds the predetermined value, it is determined that the sharpness of the local maximum value is low, and the process proceeds to step (709).

In step (707), the representative value of the distance represented by t is stored in the extreme value storage variable P (u) secured on the RAM. The central value of the distance range represented by t is appropriate as the representative value of the distance represented by t.
It is given by ^{s (t + 0.5)} · d _min . The next step (70
In 8), the numerical value u representing the number of the vertex is increased by 1 and in the following step (709), the numerical value t representing the data number of the frequency distribution of the detection target is increased by 1 to change the frequency distribution of the detection target. .

In the next step (710), it is checked whether or not the numerical value t representing the data number of the frequency distribution is within the range of the frequency distribution data, ie, whether or not “t <t _num ”. If “t _num ” is true, the process returns to step (703) because the search for the extremum is in progress. Also, "t
If <t _num ”is false, it is determined that the search has been completed, and the process proceeds to step (711).

In step (711), the number of vertices actually registered for later processing is stored in a variable u _num and the vertex detection processing ends.

According to the above flow, the local maximum value used as the boundary value for the subsequent area division is limited to the local maximum value having a sharpness equal to or more than the predetermined value, and therefore, the local maximum value is affected by the fine local maximum value. Therefore, it is possible to recognize the environment more accurately than in the first embodiment.

Also, in the present embodiment, the use data is not limited to the distance, and a similar determination can be made based on the defocus information and the image shift amount.

(Third Embodiment) A third embodiment of the present invention.
In the embodiment, a boundary value used for area division is set by using a local minimum value of the frequency distribution.

Although the flow of the entire process conforms to the first embodiment, the contents of the extreme value detection in step (104) in FIG. 5 and the setting of the boundary value in step (105) are different.

In the present embodiment, step (10) in FIG.
In the extremum detection of 4), the positions of the maximal value and the minimal value are stored. This processing will be described with reference to the flowchart in FIG.

In step (801), a variable u representing the number of the maximum value is initialized, and in the next step (802), the numerical value t representing the data number of the frequency distribution is initialized to 1. In the next step (803), the value of the flag First indicating the start of the process is set to 1. Then, in step (804), the value H (t) of the frequency distribution
A comparison is made between the values H (t-1) and H (t + 1) of the frequency distribution before and after to determine whether or not H (t) forms a local maximum value. As a result, H (t)> H (t−1) and H
If (t)> H (t + 1), it can be determined that H (t) forms a local maximum, and step (805)
Move on to

In step (805), it is checked whether or not the extremum is the first one detected. If the extremum is the first extremum, the process proceeds to step (806), and it is determined that the first extremum is the maximum value. Set a flag to represent. Then, in the next step (807), since the first extreme value registration has been completed, the value of the flag First indicating the start of the process is set to 0.

If it is determined in step (804) that the value is not the local maximum, the process proceeds to step (808) to check whether or not H (t) forms a local minimum.
As a result, H (t) <H (t-1) and H (t) <
If it is H (t + 1), it can be determined that H (t) forms a minimum value, and the routine goes to step (809).

In step (809), it is checked whether or not the extreme value is the first one detected. If the extreme value is the first extreme value, the process proceeds to step (810), and it is determined that the first extreme value is the minimum value. Set a flag to represent. Then, in the next step (811), since the first extreme value registration has been completed, the value of the flag First indicating the start of the process is set to 0.

If the maximum value or the minimum value is detected, in step (812), the representative value of the distance represented by t is stored in the variable P (u) for storing the extreme value secured on the RAM.
The central value of the distance range represented by t is appropriate as the representative value of the distance represented by t, and is given by, for example, 10 ^{s (t + 0.5)} · d _min .

Next, in step (813), the numerical value u representing the extreme value number is increased by 1, and in the following step (814), the numerical value t representing the data number of the frequency distribution to be detected is increased by 1. Change the detection target. Then, in step (815), it is checked whether or not t is within the range of the frequency distribution data, that is, whether or not “t <t _num ”. If “t <t _num ” is true, the extreme value Since the search is in progress, the process returns to step (804).

When this loop is repeated, the maximum value and the minimum value are alternately recorded in P (u). Therefore, if the flag FirstMAX is 1, P is set for odd and even u.
(U) represents the maximum value and the minimum value, respectively, and if the flag FirstMAX is 0, the opposite is true.

In the above step (815), "t <
If “t _num ” is false, it is determined that the search has been completed, and the process _proceeds to step (816), where the number of vertices actually registered is stored in a variable u _unm for later processing, and the vertex detection process ends. I do.

Next, the setting of the boundary value in step (105) will be described.

In the present embodiment, the boundary value of the area division is set to the minimum value existing before and after the maximum value. When the maximum value and the minimum value are more than a predetermined distance, the boundary value is set not at the minimum value but at a distance away from the maximum value by a predetermined distance before. The flow of setting the boundary value will be described with reference to FIG.

In step (901), a variable v representing the number of the boundary value is initialized. In the next step (902), the flag Fir set in step (104) is set.
Checks whether stMAX is 1. If "FirstMAX =
If "1", the first extremum is the maximum value, and if 0, the first extremum is the minimum value. So, if the maximum value,
In step (904), the numerical value u representing the number of the local maximum value is set to the first local maximum value number 0. If the local minimum value is first, the numerical value u is set to the first local maximum number number 1 in step (905).
Set to.

Next, in step (905), it is checked whether the numerical value u representing the number of the maximum value is 1 or more, and if it is 1 or more, the step (905) is performed in order to set the minimum value before the maximum value as the boundary value. In step 906), it is checked whether the distance difference between the maximum value and the minimum value is larger than a predetermined distance T. As a result, if it is smaller than the predetermined distance, the next step (907)
, A minimum value is set to a variable S (v) representing a boundary value.
Move to 8), and set a boundary value at a position a predetermined distance before the local maximum value. Then, in the next step (909), the value of the variable v indicating the number of the boundary value is increased by one.

When the value u representing the number of the local maximum value is smaller than 1 (ie, when u = 0) in step (905), there is no local minimum value before the first local maximum value. Then, the process directly proceeds to step (910).

In step (910), it is checked whether or not the numerical value u representing the number of the maximum value is smaller than "u _max -1".
If it is smaller, the process moves to step (911) to set the local minimum value after the local maximum value as the boundary value. Then, in this step (911), it is checked whether or not the distance difference between the local maximum value and the local minimum value is larger than the predetermined distance T. As a result, if the distance is smaller than the predetermined distance, in step (912), the value of the minimum value is set to the variable S (v) representing the boundary value.
Move to 3), and set a boundary value at a position distant by a predetermined distance from the local maximum value. Then, in the next step (914), the value of the variable v indicating the number of the boundary value is increased by one.

If the numerical value u representing the maximum number in step (910) is equal to or greater than u _max , it means that there is no minimum value after the maximum value, and the process directly proceeds to step (915). .

In the step (915), the value of the numerical value u representing the number of the maximum value is increased by 2 in order to shift the processing to the next maximum value. Then, in the next step (916), it is checked whether or not the value of the numerical value u is smaller than u _num . If it is smaller, there is still data, so the flow returns to step (905) to set the next boundary value. If the value of the numerical value u is equal to or greater than u _num , the boundary value setting has been completed for all extreme values, and the process ends.

By the above processing, the boundary value is changed to the point 1 in FIG.
A boundary value is set at a local minimum value before and after the local maximum value, as in 4,15. If the distance between the local maximum value and the local minimum value is too long, the boundary value is set at a position away from the local maximum value by a predetermined amount, such as the point 13, so that an appropriate boundary value can be set.

Although the example in which the threshold value T of the distance difference between the local maximum value and the local minimum value is simply set by the actual distance has been described, a method of obtaining the difference in a state where the distance is converted into a logarithmic space is also possible.

According to the above flow, the boundary value used as the boundary value for the subsequent area division is set between the objects, and the environment can be appropriately recognized.

Also, in the present embodiment, the use data is not limited to the distance, and a similar determination can be made based on defocus information.

(Fourth Embodiment) A fourth embodiment of the present invention will be described.
Is a combination of a conventional method and a method based on frequency distribution data, and is extremely accurate as a region dividing method. The flow of the entire process conforms to FIG.

In the fourth embodiment, the method of determining the area division in step (106) in FIG. 5 is different, and the other processes are the same as those in the first embodiment.

The method of determining the area division will be described with reference to the flowchart of FIG.

In step (1001), the distance value D (x, y) of the block of interest G (x, y) is set to the variable d ₁ for the work. In the next step (1002), the comparison block G (x−1) is set. , Y) are copied to the variable for work d ₂ .

In the next step (1003), a numerical value u representing a boundary value number and a flag Flag representing whether or not two blocks are the same are initialized. "Flag =
"0" indicates that the two objects are the same, and "Flag
= 1 "indicates that the two objects are distinct.

In the next step (1004), it is checked whether the difference between the distances of the two blocks is smaller than a predetermined value Ts. If it is smaller, it is determined that the two objects are likely to be the same object, and the process proceeds to step (1005). If the distance difference is equal to or greater than Ts, the two objects are determined as different objects, and the process proceeds to step (1006).

In step (1005), it is checked whether or not a boundary value S (u) exists between the two distances to be compared. “D ₁ <S (u) and S (u) <d ₂ ” is a condition for “d ₁ <d ₂ ”, and “d ₂ <S (u) a
“nd S (u) <d ₁ ” is a condition in the case of “d ₂ <d ₁ ”.

If the result of the logical sum is true, it is possible to determine that the two blocks are separate objects, so that a flag is set in step (1006) and the processing ends.

On the other hand, if the logical sum is false, the value of the numerical value u representing the maximum number is increased by one in step (1007) in order to compare with the next boundary value. Then, in the next step (1008), it is checked whether or not the numerical value u representing the number of the maximum value exceeds the number of boundary values, and if not, the process returns to step (1005), and the next boundary value is used. Perform a check. If it is over, the check ends with "Flag = 0", and the two blocks are ended as the same object.

The above determination is made for all adjacent relations of all blocks, and the area division (grouping) is completed.

By performing the area division by the above method,
Even in the case of a plurality of objects in which only a part which is conventionally divided into a single region and which is only partly close to each other, the region can be correctly divided by further checking with a boundary value set by the frequency distribution.

(Modification) Although the present invention has been described as applied to a single-lens reflex camera, it is also applicable to video devices such as video cameras and electronic still cameras. Further, the present invention is applicable to the following devices.

For example, the present invention can be used for an automatic traveling robot that determines its own traveling direction, and a vehicle that determines obstacles in front and behind to automatically avoid danger. Furthermore, the present invention can be used in a wide range of technical fields, such as an air conditioner that determines the direction in which a person is present in a room and controls air blowing appropriately.

Further, the present invention may have a configuration in which the above embodiments or their techniques are appropriately combined.

[0178]

As described above, according to the first and ninth aspects of the present invention, by using the extreme value information of the frequency distribution information of the distance for the area division, the distance which has conventionally been frequently misjudged is determined. Even in a situation where there is an object having a small difference, it is possible to correctly perform region division, and it is possible to realize accurate environment recognition.

According to the second and ninth aspects of the present invention, a local maximum value of distance frequency distribution information is detected, and a boundary value used as a determination element for area division is set before and after the local maximum value. Thus, the region can be correctly divided, and accurate environment recognition can be realized.

According to the third and ninth aspects of the present invention, the sharpness of the extreme value of the distance frequency distribution information is evaluated, and only the extreme value information having a predetermined sharpness is used for area division. As a result, the region can be more correctly divided, and accurate environment recognition can be realized.

Further, according to the present invention as set forth in claims 4 and 9, the minimum value of the frequency distribution information of the distance is used for the area division, so that the area can be correctly divided, and an accurate environment can be obtained. Recognition can be realized.

According to the present invention as set forth in claims 5 and 10, by using extreme value information of the defocus frequency distribution information for area division, the distance difference in which erroneous determination has conventionally been large is small. Even in a situation where an object exists, it is possible to correctly perform region division, and to realize accurate environment recognition.

According to the sixth and tenth aspects of the present invention, the local maximum value of the defocus frequency distribution information is detected.
By setting a boundary value to be used as a determination element for area division before and after the maximum value, area division can be performed correctly, and accurate environment recognition can be realized.

According to the present invention, the sharpness of the extreme value of the frequency distribution information of defocus is evaluated, and only the information of the extreme value having a predetermined sharpness is used for area division. By doing so, it becomes possible to perform the area division more correctly, and it is possible to realize accurate environment recognition.

According to the present invention as set forth in claims 8 and 10, by using the minimum value of the frequency distribution information of defocus for area division, it becomes possible to perform area division correctly, and Environmental recognition can be realized.

Further, according to the present invention described in claim 11,
A camera that includes an environment recognition device that uses distance or defocus information and can automatically set appropriate shooting parameters for a main subject can be realized.

Further, according to the present invention described in claim 12,
A camera that includes an environment recognition device that uses distance or defocus information and that automatically adjusts the focus of a main subject can be realized.

Further, according to the present invention described in claim 13,
It is possible to realize a camera that includes an environment recognition device that uses distance or defocus information and can accurately adjust the exposure of the main subject even under adverse conditions such as backlight.

According to the fourteenth aspect of the present invention,
It is possible to realize a camera that has an environment recognition device that uses distance or defocus information and allows the photographer to adjust the angle of view appropriately for the main subject without adjusting the focal length of the lens. it can.

[Brief description of the drawings]

FIG. 1 is an arrangement diagram of an optical system of a camera according to a first embodiment of the present invention.

FIG. 2 is a perspective view showing a more detailed structure of the optical system of FIG. 1;

FIG. 3 is a plan view of the optical system of FIG. 2;

FIG. 4 is a block diagram showing an electrical configuration of the camera according to the first embodiment of the present invention.

FIG. 5 is a flowchart showing an operation of a main part of the camera of FIG. 4;

FIG. 6 is a flowchart showing an operation at the time of creating a distance map in step (102) of FIG. 5;

FIG. 7 is a flowchart showing an operation at the time of creating a frequency distribution in step (103) of FIG. 5;

8 is an example of a histogram created by the operation of FIG. 7;

FIG. 9 is a flowchart showing an operation at the time of detecting an extreme value in step (104) of FIG. 5;

FIG. 10 is a diagram for helping to explain the operation of setting the boundary value in FIG. 11;

FIG. 11 is a flowchart showing an operation when setting a boundary value in step (105) of FIG. 5;

FIG. 12 is a diagram for describing an area dividing method according to the first embodiment of the present invention.

FIG. 13 is a diagram illustrating an example of a labeling result according to the first embodiment of the present invention.

FIG. 14 is a flowchart showing an operation at the time of boundary determination according to the first embodiment of the present invention.

FIG. 15 is a diagram for explaining peaks to be detected by the camera according to the second embodiment of the present invention.

FIG. 16 is a flowchart showing an operation of the camera according to the second embodiment of the present invention when detecting a boundary maximum value.

FIG. 17 is a flowchart showing an operation of the camera according to the second embodiment of the present invention when detecting a boundary value.

FIG. 18 is a flowchart showing an operation when setting a boundary value of the camera according to the third embodiment of the present invention.

FIG. 19 is a diagram showing an example of setting a boundary value in a camera according to a third embodiment of the present invention.

FIG. 20 is a flowchart showing an operation of the camera at the time of boundary determination according to the fourth embodiment of the present invention.

FIG. 21 is a diagram illustrating an example of a shooting scene with a camera.

[Explanation of symbols]

 Reference Signs List 1 shooting lens 4 area sensor 4a, 4b imaging screen 11-17 subject example extracted from distance map PRS microcomputer ICC area sensor LNS lens LPRS microcomputer in lens

Claims (14)

[Claims] 1. An environment recognition apparatus comprising: a region dividing unit that divides a screen into individual objects based on distance distribution information of an object scene, and performs environment recognition based on an output of the region dividing unit. In the environment recognition apparatus, the region dividing means generates distance frequency distribution information based on the distance distribution information, and uses the frequency distribution information as one of the determination factors for region division. 2. The area dividing means detects a local maximum value of the frequency distribution information, sets a boundary value at a distance separated by a predetermined amount before and after the local maximum value, and sets an area based on the boundary value information. The environment recognition apparatus according to claim 1, wherein the environment recognition apparatus performs division. 3. The region dividing means evaluates the sharpness of all the detected extreme values, and uses only the information of the extreme value having a sharpness equal to or more than a predetermined value as one of the determination elements of the region division. The environment recognition device according to claim 1, wherein the environment recognition device is used. 4. The method according to claim 1, wherein the region dividing means detects a minimum value of the frequency distribution information, sets a boundary value with a distance of the minimum value, and performs region division based on the boundary value information. The environment recognition device according to claim 1. 5. The image processing apparatus according to claim 1, further comprising a region dividing unit that divides a screen into individual regions based on distribution information of defocus of the object scene, and performs environment recognition based on an output of the region dividing unit. In the environment recognition device, the area dividing unit creates defocus frequency distribution information based on the defocus distribution information, and uses the frequency distribution information as one of the determination elements for the area division. Environment recognition device. 6. The region dividing means detects a local maximum value of the frequency distribution information, sets a boundary value at a defocus value separated by a predetermined amount before and after the local maximum value, and based on the boundary value information. 6. The environment recognition apparatus according to claim 5, wherein the area is divided by performing the area division. 7. The region dividing means evaluates the sharpness of all the detected extreme values, and uses only the information of the extreme value having a sharpness equal to or more than a predetermined value as one of the determination elements of the region division. The environment recognition device according to claim 5, wherein the environment recognition device is used. 8. The region dividing means detects a local maximum value and a local minimum value of the frequency distribution information, sets a defocus value of a local minimum value before and after the local maximum value as a boundary value, and sets the boundary value information. 6. The environment recognition apparatus according to claim 5, wherein the area is divided based on the following. 9. The apparatus according to claim 1, further comprising a distance measuring means for measuring distances in a plurality of directions, wherein said area dividing means divides the area based on distance distribution information obtained by said distance measuring means. The environment recognition device according to 1, 2, 3, or 4. 10. Defocus measuring means for measuring defocus in a plurality of directions with respect to a focal position of a photographing optical system, wherein said area dividing means is based on defocus distribution information obtained by said defocus measuring means. 9. The environment recognition apparatus according to claim 5, wherein the area is divided. 11. A camera comprising: the environment recognition device according to claim 9; and a main subject detection unit configured to detect a main subject region from an environment recognition result of the environment recognition device. 12. The camera according to claim 11, further comprising an automatic focus adjustment unit that focuses on a main subject based on an output result of the main subject detection unit. 13. An image forming apparatus comprising: a measuring means for measuring a light amount of a main subject area based on an output result of the main subject detecting means; and an automatic exposure adjusting means for adjusting an exposure amount based on the light quantity. The camera according to claim 11, wherein 14. An apparatus according to claim 11, further comprising an automatic focus adjusting means for adjusting a focal length of the lens so that the main subject has an appropriate size based on an output result of said main subject detecting means. Camera.