JP2018040692A - Self-vehicle position estimation device and self-vehicle position estimation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a self-vehicle position estimation device and a self-vehicle position estimation method capable of reducing a possibility to erroneously recognize a start position of branching or confluence.SOLUTION: An ECU 20 includes: a sign recognition section 21 for recognizing a road surface sign indicating a boundary of a lane in front of a course of a self- vehicle on the basis of an imaging result by an imaging section 32; an extraction section 22 for extracting an angle change part to be changed at a predetermined angle with respect to a direction along a lane within a recognized road surface sign as a candidate of a shape change part indicating a road shape change; an information acquisition section 23 for acquiring position information indicating a position where a road shape change occurs in front of a course; a narrowing section 24 for narrowing extracted candidates to a candidate present within a predetermined distance in a direction along a lane from a position indicated by the position information; a sign identification section 25 for identifying a narrowed candidate as a shape change part in front of a course; and a position estimation section 26 for identifying a self-vehicle position on a map on the basis of an identified branch point or confluence point of a lane.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an in-house position estimation device that recognizes the shape of a road and a vehicle position estimation method.

  2. Description of the Related Art Conventionally, a method is known in which a highly accurate digital map is provided, and the latitude / longitude (position) and direction of the host vehicle are estimated on the map to know the road shape ahead of the course and to steer the host vehicle.

  Patent Document 1 discloses a method of adjusting the position of the host vehicle at the position of the branch road based on a captured image obtained by capturing the front of the path of the host vehicle. In the apparatus disclosed in Patent Document 1, when a feature point of a thick frame line indicating a boundary between a main line and a branch road on a road or a zebra zone is extracted from a captured image, it is determined that a branch road exists ahead of the course. .

Japanese Patent Laid-Open No. 2007-3286

  As described in Patent Document 1, the method of aligning with a broken line indicating a branch has the following problems. In a real environment, there are branch paths in which the broken line is not visible, and branch paths in which the broken line is not drawn. On the road under such conditions, the method disclosed in Patent Document 1 cannot be used.

  On the other hand, when recognizing the start position of a lane branch or the like on the map based on the shape of a lane line or the like extracted from the image, a road surface dirt or a scratch on the road surface is erroneously recognized as the start position of this branch or the like. There is a case. For example, if the dirt on the road surface extends in a direction resembling a branch, such as a road surface repair mark such as coal tar or a white lined area between the black lines of the brake marks of a double tire on a truck May be misrecognized as a lane marking indicating the branch start position. In the vehicle position estimation, if the front and rear positions are not accurately matched before the branch position, for example, the timing of the lane change to the branch road may be shifted, leading to a great sense of discomfort for the driver. In place of recognition by lane markings, there is a method of aligning the front and rear positions with a sign or the like. However, there are cases where there is no sign or the sign cannot be recognized by shielding a preceding vehicle or the like. For this reason, appropriately recognizing the lane branching and merging position is an important technique for estimating the position of the vehicle on the map.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a host vehicle position estimation device and a host vehicle position estimation method that can reduce the possibility of misrecognizing the start position of branching or merging. .

  In order to solve the above problem, in the present invention, a map recognition unit that recognizes a road marking indicating a boundary of a lane in front of the course of the host vehicle based on a result of imaging by the imaging unit, and a map in the recognized road marking, An angle change part whose shape changes at an angle corresponding to the angle of the branch lane relative to the main line at the lane junction above or the angle of the merging lane relative to the main line at the lane junction is extracted as a candidate. And an information acquisition unit for acquiring position information indicating a position of a branching point or a merging point of the lane in front of the course of the host vehicle, and the extracted candidate from the position indicated by the position information. A narrowing-down part that narrows down to the candidates existing within a predetermined distance in a direction along the lane, and an indication that identifies the narrowed-down candidates as a position indicating a branch point or a merging point of the lane in front of the course Comprising a tough, based on the identified said candidate, a position estimation unit for estimating a vehicle position on the map, the.

  The inventor of the present application has a map in the first place when estimating the vehicle position, and tire marks that tend to cause misrecognition enter the branch road while braking the truck before the main line. Because it is made at the time, the angle formed with the lane marking of the main line is difficult to coincide with the angle of the branch road, and further, attention is paid to the point that road surface repair marks and cracks are also uneven and difficult to coincide with the angle of the branch road did. Therefore, by using the information on the angles of branching and merging that are already known on the map, the part having the angle corresponding to the branching and merging is extracted as a candidate, and the extracted candidate is narrowed down based on the position information, so that the vehicle We decided to specify the position of branching and merging used to specify the position. For this reason, from the vehicle position that has been estimated to some extent in advance, the presence and angle of the front branch or merging and its angle are obtained from the map information to narrow down other than the angle. Then, the vehicle position is estimated using the narrowed candidates. As a result, it is possible to reduce the possibility of misrecognizing the start position of branching or merging, and to appropriately estimate the vehicle position on the map.

  In addition, it is possible to estimate the vehicle position in the front-rear direction in the same way not only at the folding point of the lane marking but also at the bending point of the guardrail at the road edge. This makes it possible to match even when the lane marking is faint or when it cannot be seen by backlight. Also, by providing a means for outputting road edge information on the map, it is possible to estimate the position of the vehicle in the same way, and the fluctuation of the shape due to trees at the road edge that becomes a disturbance at that time, at an already known angle By narrowing down, errors can be suppressed.

The block diagram of a vehicle control apparatus. The figure explaining the map currently recorded on the external memory. The figure explaining specification of the own vehicle position. The figure explaining the extraction method of a road marking provided with an angle change part. The flowchart explaining the specification of the own vehicle position which ECU implements. The flowchart which shows the process of FIG.5 S13 in detail. The figure explaining the angle of an angle change part. The figure explaining extraction of a candidate. The figure explaining narrowing down of a candidate. The figure explaining exclusion of the candidate using a luminance pattern. The figure explaining the exclusion method of the candidate by determination of a tire mark. The flowchart explaining the process implemented by FIG.4 S14. The figure explaining the process which searches the position where the shape change of a road produces in 3rd Embodiment.

  Embodiments of a host vehicle position estimation apparatus and a host vehicle position estimation method according to the present invention will be described with reference to the drawings. In the following embodiments, parts that are the same or equivalent to each other are denoted by the same reference numerals in the drawings, and the description of the same reference numerals is used.

(First embodiment)
FIG. 1 shows a vehicle control device 100 to which the vehicle position estimation device and the vehicle position estimation method are applied. The vehicle control apparatus 100 is an example of a vehicle system mounted on a vehicle, for example, estimates the position of the host vehicle on a map, and supports driving of the host vehicle based on the estimation result. Further, in this embodiment, the road marking is a display body that forms a road boundary line formed on the road surface, and means road paint, road fences, curbs, and the like.

  As shown in FIG. 1, the vehicle control device 100 includes various sensors 30, an ECU 20 that functions as a host vehicle position estimation device, and a driving support device 40.

  The various sensors 30 include a GPS receiver 31, a camera 32, a vehicle speed sensor 35, and a yaw rate sensor 36.

  The GPS receiver 31 functions as part of a well-known satellite positioning system (GNSS), thereby receiving radio waves transmitted from the satellite as GPS information. The GPS information includes DOP (error information) that is information indicating an error in the position of the satellite, the time when the radio wave is transmitted, and the position of the satellite. The GPS receiver 31 calculates the distance from the satellite to the host vehicle based on the difference between the reception time when the GPS information is received and the transmission time included in the GPS information. Then, the calculated distance and the position of the satellite are output to the ECU 20.

  The camera 32 functions as an imaging unit, and is installed in the vehicle interior with the imaging axis facing the front of the host vehicle so that the front of the host vehicle can be captured. In this embodiment, the camera 32 is a monocular camera, and is composed of, for example, a CCD image sensor or a CMOS image sensor. Moreover, the image imaged with the camera 32 is output to ECU20 with a predetermined period. The camera 32 may be a stereo camera having a right camera and a left camera arranged in parallel in the vehicle width direction.

  The vehicle speed sensor 35 is provided on a rotating shaft that transmits power to the wheels of the host vehicle, and detects the speed of the host vehicle based on the number of rotations of the rotating shaft. The yaw rate sensor 36 detects a change in the direction of the host vehicle per unit time based on the yaw rate actually generated in the host vehicle, that is, the angular velocity around the center of gravity of the vehicle.

  ECU20 is comprised as a computer provided with CPU, ROM, and RAM. The CPU can function as each unit shown in FIG. 1 by executing a program stored in the internal memory. The ECU 20 is connected to an external memory 45 in which a map is recorded, and can read each piece of information recorded on the map from the external memory 45.

  In the map shown in FIG. 2, a link L indicating a lane on the road and a node N indicating a connection point of the lane are recorded. The coordinates on the map are recorded in the node N, and by referring to the position of the node N, it is possible to detect the position where the shape change of the road on the map occurs.

  As a change in the shape of the road on the map, FIG. 2A shows a lane branch, and FIG. 2C shows a lane merge. As shown in FIG. 2A, a link L3 indicating a branch path is connected to a node N1 indicating a branch point in addition to links L1 and L2 indicating the main line. In addition, as shown in FIG. 2C, a node L11 indicating a junction is recorded with a link L13 indicating a merge channel in addition to the links L11 and L12 indicating the main line. In addition to the node N indicating the lane connection, the map may include auxiliary points indicating the position on the road for each predetermined distance.

  Further, the map records, as road information, shape information indicating the shape and position of the feature on the road, and attribute information indicating the name of the feature and related information. The shape information and attribute information are stored in a database so that the values can be referred to in association with each position or feature on the map. For example, as shape information, in FIG. 2B, the shape (approximate curve) of the road surface paint indicating the road boundary line and the branch angle θr of the branch road branched from the main line are recorded in association with the position on the map. ing. For example, the shape of the road surface paint is formed by well-known vector data including coordinates of representative points and approximate curves of lines connecting the representative points.

  Moreover, ECU20 recognizes the shape change part which shows the position where the shape change of a road arises in a road marking. The shape changing portion has a portion inclined at a predetermined angle with respect to the direction along the lane in the road marking indicating the boundary line of the road. In this embodiment, the shape changing portion indicates a lane branch point, a merging point, or the like. It is a part. ECU20 recognizes the relative position on the basis of the own vehicle in the shape change part ahead of a course by the edge point obtained from the imaging result by the camera 32. FIG. In addition, the recognized shape changing portion is used for estimating the vehicle position on the map.

  A method for recognizing the vehicle position using the shape change unit recognized by the ECU 20 will be described with reference to FIG. First, the ECU 20 calculates the detected position DP on the absolute coordinates by adding the position EP of the host vehicle acquired on the map based on the GPS information or the like to the relative position of the recognized shape changing unit. Then, the amount of deviation between the two positions DP and MP is calculated by matching the calculated detection position DP with the position MP of the shape changing portion on the map corresponding to the detection position DP. For example, the relationship between the detected position DP and the position MP on the map is matched using a determinant having the shift amount as an element, and the shift amount can be calculated by solving each element of the determinant. Then, the vehicle position CP is estimated by correcting the position EP based on the calculated deviation amount. Here, a part of the shape changing portion is inclined at a predetermined angle from the direction along the lane, and both positions DP and MP can be matched in the traveling direction using this portion. Therefore, the amount of deviation in the traveling direction can be calculated with high accuracy by using the shape changing portion for estimating the vehicle position.

  The driving support device 40 controls the traveling of the host vehicle based on the host vehicle position CP estimated by the ECU 20. For example, the driving support device 40 predicts the future position of the host vehicle based on the estimated host vehicle position CP, the vehicle speed, and the yaw rate, and uses the predicted future position and the road recognition result to determine whether the host vehicle is a lane marking. It is determined whether or not there is a risk of deviating from the own lane divided by the road surface paint. For example, if the driving support device 40 has an alarm function, if it is determined that the host vehicle may deviate from the host lane, a warning is displayed on a display included in the host vehicle, or a speaker included in the host vehicle is used. A warning sound is generated. Further, if the driving support device 40 has a driving assist function, when it is determined that the host vehicle may deviate from the own lane, a steering force is applied to the steering device.

  When the ECU 20 recognizes the shape change portion in the road marking based on the imaging result of the camera 32, dirt or scratches on the road surface may be erroneously recognized as the shape change portion. For example, in the example of FIG. 3, the dirt on the road surface extends at a predetermined angle in the direction along the road, and the camera 32 misrecognizes the dirt position Di as the shape change portion. If the vehicle position CP is specified based on the position Di in which dirt is erroneously recognized, an error may occur in the vehicle position CP on the map. Further, in the vehicle position estimation, if the front and rear positions are not accurately matched before the branch position, for example, the timing of the lane change to the branch road is shifted, and the driver feels uncomfortable. Therefore, appropriately recognizing a part on the road used for estimating the position of the vehicle is an important technique for realizing accurate matching. Therefore, the ECU 20 is provided with each function shown in FIG. 1 to reduce the possibility of erroneously recognizing a position where a road shape change occurs.

  Next, each function of the ECU 20 shown in FIG. 1 will be described. The sign recognition unit 21 recognizes a road marking indicating the boundary of the lane ahead of the course of the host vehicle based on the imaging result of the camera 32. For example, the sign recognition unit 21 performs bird's-eye processing on a captured image captured by the camera 32 to convert the position of an object or the like in the captured image into a positional relationship when viewed from above the host vehicle. Then, feature points such as edge points are extracted from the captured image after the bird's-eye view processing. Then, the extracted feature points are grouped according to their shapes, luminance patterns, etc., so that the grouped feature points are recognized as road surface display.

  In the road marking recognized by the sign recognition unit 21, the extraction unit 22 determines the angle of the branch lane with respect to the main line at the branch point of the lane on the map, or the merging lane with reference to the main line at the merging point of the lane. An angle changing portion whose shape changes at an angle corresponding to the angle is extracted as a candidate. For example, in FIG. 4, the extraction unit 22 changes the angle having a shape change point HP whose shape changes at a predetermined inclination angle θb with respect to the direction along the lane among the road markings recognized in the captured image. The part ACP is extracted as a candidate for the shape change part. At this time, the inclination angle θb is determined based on an angle indicating a change in the shape of the road existing ahead of the host vehicle. In FIG. 4, an angle change portion ACP including an inclination portion Sb inclined at an inclination angle θb from the shape change point HP with respect to the straight portion Ss parallel to the center line M of the road is extracted as a shape change portion candidate. In addition, as a calculation method of each part Ss and Sb for calculating inclination-angle (theta) b, it can calculate by well-known Hough (Hough) transformation of the feature point of a road marking, for example.

  The information acquisition unit 23 acquires position information indicating the position of a lane branch point or a lane junction point where a change in the shape of the road ahead of the course occurs based on the position of the host vehicle. In the first embodiment, the information acquisition unit 23 acquires, as position information, the position of a node N indicating a lane junction or merging point recorded on the map, based on the position of the host vehicle estimated on the map. To do.

  The narrowing-down unit 24 narrows down the shape change unit candidates extracted by the extraction unit 22 to candidates existing within a predetermined distance in the direction along the lane from the position indicated by the position information. Since the position information acquired by the information acquisition unit 23 includes an error, the narrowing unit 24 sets a position where a candidate to be narrowed exists in consideration of this error. For example, the narrowing-down unit 24 sets an area for narrowing down each candidate based on the position indicated by the position information, and uses candidates belonging to this area as a narrowing target.

  The sign specifying unit 25 specifies the narrowed candidates as a shape changing unit in front of the course. That is, the position of the angle change portion of the narrowed road marking is specified as a position where a shape change such as branching or merging actually occurs on the road.

  The position estimating unit 26 estimates the vehicle position CP on the map using the identified candidate shape changing unit. In this embodiment, the position estimation unit 26 is acquired based on the position on the map acquired based on the output of the GPS receiver 31 functioning as a positioning sensor, or based on the outputs of the vehicle speed sensor 35 and the yaw rate sensor 36. The vehicle position CP is specified by correcting the position on the map.

  Next, the identification of the own vehicle position CP performed by the ECU 20 will be described with reference to FIG. The process shown in FIG. 5 is a process performed by the ECU 20 at a predetermined cycle. Further, in the process shown in FIG. 5, a case where a lane branch point is used as a road shape changing portion and a road surface paint indicating the branch point is specified will be described as an example.

  In step S10, the position of the host vehicle on the map is acquired based on the GPS information. For example, the ECU 20 acquires an estimated position estimated on a map using a well-known map matching process. Step S10 functions as a positioning position acquisition unit.

  In step S11, it is determined whether or not a lane branch point exists within a predetermined distance ahead of the course from the position acquired in step S10. Here, the predetermined distance may be a distance of 10 meters or more and less than 100 meters from the estimated position, for example, a position within 50 meters. If there is a node N indicating a branch point within a predetermined distance from the estimated position on the map, the ECU 20 determines that a lane branch point exists. On the other hand, when a branch point does not exist within a predetermined distance from the estimated position (step S11: NO), the process of FIG. 5 is once ended.

  When a branch point of the lane exists within a predetermined distance ahead of the course from the estimated position (step S11: YES), in step S12, based on the captured image that is the imaging result of the camera 32, the vehicle ahead of the course Recognize the position of road paint. Step S12 functions as a sign recognition process.

  In step S13, in the recognized road surface paint, an angle changing portion that changes at a predetermined angle with respect to the direction along the lane is extracted as a candidate for a shape changing portion that indicates lane branching or merging. FIG. 6 is a flowchart showing in detail the process of step S13. Step S13 functions as an extraction process.

  In step S21, the captured image is subjected to bird's-eye processing. In step S22, an angle change part included in the image after bird's-eye view processing is extracted. For example, the ECU 20 extracts a part having a shape change point as an angle change part in the road surface paint.

  In step S23, it is determined whether or not the branching angle θr of the branching lane based on the main line at the branching point of the lane is recorded on the map. As shown in FIG. 7A, if the branch angle θr of the branch lane is recorded as shape information on the map (step S23: YES), in step S24, this branch angle θr is recorded on the map. Obtain from information.

  When the lane junction is used as a position where the road shape changes, in step S23, the ECU 20 obtains the angle of the merging lane with respect to the main line at the lane junction.

  When the branching angle θr of the branch lane is not recorded as shape information on the map (step S23: NO), in step S25, the angle formed by the link L indicating the branch road and the link L indicating the main line is set. Based on this, the branch angle θr of the branch lane is calculated. For example, as shown in FIG. 7B, the ECU 20 is connected to a node N21 indicating a branch point, calculates an angle formed by a link L21 indicating a main line and a link L22 indicating a branch path, and calculates the angle. The branch angle of the branch lane is θr.

  In step S26, the branching angle θr acquired in step S24 or step S25 is compared with the inclination angle θb of the angle changing unit extracted in step S22, and a candidate for the shape changing unit is extracted based on the comparison result. For example, the ECU 20 extracts an angle change unit in which the difference between the inclination angle θb of the angle change unit and the branch angle θr of the branch lane is equal to or less than a threshold Th as a shape change unit candidate. The threshold value Th can be set to a value between 5 degrees and 10 degrees, but the value varies depending on the recognition accuracy of the camera 32 and the accuracy of the branching angle θr recorded on the map. In the example of FIG. 8, the shape change portion candidates C <b> 1 to C <b> 3 are extracted from the captured image by the determination in step S <b> 26.

  Returning to FIG. 5, in step S <b> 14, position information indicating the position where the shape change of the road in front of the course occurs is acquired with reference to the position of the host vehicle. The recognition accuracy of the imaging result obtained by the camera 32 decreases as the distance from the host vehicle increases. Therefore, as shown in FIG. 9A, in the first embodiment, the ECU 20 on the map, the position of the lane branch point (node N) within the predetermined distance D1 from the position EP acquired in step S10. And the searched position is acquired as position information. Here, the predetermined distance D1 is a distance set according to the imaging accuracy of the camera 32. Step S14 functions as an information acquisition process.

  In step S15, each candidate in the captured image is narrowed down to candidates existing within a predetermined distance in the direction along the road from the position indicated by the position information. Since there is an error in the position information, as shown in FIG. 9B, the ECU 20 narrows down the candidates from the position information Pi acquired in step S14 in the direction along the road by a distance D2 back and forth. Is set as the narrowed-down area NA. Then, the candidate C within the set narrowing area NA is set as a narrowing target. The distance D2 is set according to the accuracy of the position information, and can be set to a value of 5 meters or more and 20 meters or less, for example.

  If there is no candidate as a result of the narrowing down in step S15 (step S16: NO), the processing in FIG. On the other hand, if there is a candidate as a result of the narrowing down in step S15 (step S16: YES), the process proceeds to step S17. In the example of FIG. 9B, since the candidate C1 is located in the narrowed down area NA, each candidate is narrowed down to the candidate C1.

  In step S17, the candidates narrowed down in step S16 are specified as the shape changing part on the road. In the example of FIG. 9B, the candidate C1 is specified as the shape changing unit corresponding to the branch point. Step S18 functions as a sign specifying step.

  In step S18, the vehicle position CP is estimated using the shape change part (candidate) specified in step S17. Specifically, the ECU 20 calculates a deviation amount based on a result of matching the detected position DP of the shape change portion with a position MP on the map of the shape change portion, and corrects the position EP by using the deviation amount. The vehicle position is estimated from the corrected position. Step S18 functions as a position estimation step.

  As described above, in the first embodiment, the ECU 20 recognizes the road marking indicating the boundary of the lane in front of the course of the host vehicle based on the imaging result of the camera 32, and in the recognized road marking, An angle changing unit having an angle corresponding to the angle of the branch lane with respect to the main line at the branch point of the above lane or the angle of the merged lane with reference to the main line at the junction point of the lane is extracted as a candidate. In addition, position information indicating the position of a branch point or merging point of the lane in front of the course of the own vehicle is acquired, and the extracted candidates exist within a predetermined distance in the direction along the own lane from the position indicated by the position information. Narrow down to candidates. Then, the narrowed candidates are identified as lane junctions or merging points in front of the course, and the vehicle position on the map is estimated based on the identified candidates. In this case, by narrowing down candidates that satisfy both the angle of the angle changing unit and the range based on the position information, it is possible to reduce the possibility of misrecognizing road markings indicating the start of lane branching or merging. As a result, it is possible to specify the position of the vehicle on the map based on the road marking that has been properly recognized.

  The ECU 20 acquires the vehicle position on the map based on the output from the GPS receiver 31, and searches for a branch point or junction of lanes located within a predetermined distance from the vehicle position on the map. And the position of the branch point or merging point of the searched lane is acquired as position information. The recognition accuracy of the imaging result obtained by the camera 32 decreases as the distance from the host vehicle increases. In this regard, in the above configuration, the position of the shape changing portion existing within a predetermined distance from the vehicle position on the map is acquired as position information. In this case, since the range in which the candidates are narrowed down is a range in which the recognition accuracy of the camera 32 is ensured, it is possible to suppress the possibility of erroneously recognizing the shape change portion due to the decrease in the recognition accuracy of the camera 32. .

(Second Embodiment)
In the second embodiment, candidates corresponding to dirt or scratches on the road among the candidates extracted based on the imaging result of the camera 32 are excluded from the candidates.

  For example, in the candidate narrowing in step S15 in FIG. 5, each candidate in the image is narrowed down based on the luminance pattern in addition to the distance from the position indicated by the position information. When comparing road paint with dirt and scratches on the road, road paint shows a pattern in which the road color appears on both sides of the white line area, whereas dirt and scratches on the road can be such a pattern. The nature is low. Therefore, ECU20 determines whether each extracted candidate is road surface paint based on the surrounding luminance pattern containing the candidate in the road width direction of a road, and excludes a candidate according to a determination result. Therefore, step S15 in FIG. 5 functions as a luminance pattern exclusion unit.

  FIG. 10 is a diagram for explaining candidate exclusion using a luminance pattern. As an example of the determination method of the luminance pattern, the ECU 20 grayscales surrounding pixels including the candidate periphery, and calculates a histogram of the grayscaled pixels. In FIG. 10A, the periphery of the candidate C1 that is a road surface paint is composed of a white line (high luminance) region that is a paint region, and a road surface (low luminance) region that appears on both sides of the white line. For this reason, in the histogram around the candidate C1 shown in FIG. 10B, the number of high- and low-luminance pixels PN is large and the number of medium-luminance pixels PN is small. On the other hand, in FIG. 10A, the periphery of the candidate C3, which is a scratch on the road surface, is mainly composed of a road surface (low luminance) region. For this reason, in the histogram around the candidate C3 shown in FIG. 10C, the number of high luminance pixels PN is small, and the number of medium luminance to low luminance pixels PN is large. Therefore, ECU20 can exclude the candidate which is not road surface paint by determining the luminance pattern obtained by the histogram around each candidate.

  As described above, in the second embodiment, the ECU 20 determines whether or not the candidate is road paint based on the surrounding luminance pattern including the candidate in the road width direction, and the candidate that has not been determined as road paint. Are excluded from narrowing down. When comparing road paint with dirt and scratches on the road, road paint shows a brightness pattern in which the road color appears on both sides of the white line area, whereas dirt and scratches on the road have such a brightness pattern. The possibility of not appearing is increased. In this regard, in the above configuration, whether or not the candidate is road surface paint is determined based on the luminance pattern around the candidate in the road width direction of the road, and if it is not road surface paint, it is excluded from the target of narrowing down. In this case, dirt and scratches on the road whose shape approximates that of the shape change portion in the road surface paint can be excluded, and the accuracy of narrowing down candidates can be improved.

(Third embodiment)
In the third embodiment, the candidates corresponding to the tire marks are excluded from the candidates extracted based on the imaging result of the camera 32.

  FIG. 11 is a diagram for explaining a candidate exclusion method in the third embodiment. If there is a tire mark TC extending at a predetermined angle in the direction along the road on the road surface, the tire mark may be erroneously recognized as a part indicating a lane branch or merging point. Accordingly, the ECU 20 narrows down candidates in step S15 in FIG. 5 based on the determination result of whether each candidate in the image is a tire mark in addition to the distance from the position indicated by the position information. Narrow down. Therefore, in this 3rd Embodiment, Step S15 of Drawing 5 functions as a tire mark exclusion part.

  For example, the ECU 20 extracts a feature amount such as a texture of each candidate, and excludes a candidate having a feature amount similar to a tire mark. As an example of the feature amount calculation method, the ECU 20 calculates a histogram indicating the luminance distribution of the pixels constituting each candidate, and determines whether the texture corresponds to the tire mark based on the histogram. And the candidate (TC in FIG. 11) determined to be a tire mark is excluded from the candidates for the shape change portion.

  As a method for calculating the feature amount of each candidate, a known pattern matching may be used instead of the luminance histogram. Moreover, you may combine the determination whether it is a tire trace shown in this 3rd Embodiment, and the determination of the luminance pattern shown in 2nd Embodiment.

  As described above, in the third embodiment, the ECU 20 determines whether or not the candidate is a tire mark, and excludes the candidate determined as the tire mark from the target of narrowing down. If tire marks extending at a predetermined angle in the direction along the road exist on the road surface, the tire marks may be erroneously recognized as a shape change portion. In this regard, in the above configuration, it is determined whether or not the extracted candidate is a tire mark, and the candidate determined as the tire mark is excluded from the narrowing targets. In this case, tire marks that are easily misrecognized as shape change portions can be excluded from the candidates, and the accuracy of narrowing down the candidates can be improved.

(Fourth embodiment)
In the fourth embodiment, position information for setting a range for narrowing candidates is acquired based on the imaging result of the camera 32.

  FIG. 12 is a flowchart illustrating the process performed in step S14 of FIG.

  In step S31, the position of the road surface in the captured image is estimated. For example, the ECU 20 estimates a region indicating a road using road surface paint indicating left and right road boundaries recognized in step S11 of FIG.

  In step S32, an increase in road width is detected in the direction along the road. ECU20 extracts the road surface paint which shows the right and left road boundary in the road surface estimated by step S31, and detects the increase in road width by the change of the space | interval in the vehicle width direction (road width direction) of this road surface paint. Step S32 functions as a road width determination unit.

  If an increase in road width is not detected (step S33: NO), it is determined that there is no position where the road shape changes in front of the course, and the processing in FIGS. 5 and 12 is temporarily terminated. On the other hand, when an increase in road width is detected (step S33: YES), the process proceeds to step S34.

  In step S34, a position where the road width is increasing is acquired as position information. For example, in FIG. 13A, the ECU 20 obtains, as the position information Pi, the position closest to the host vehicle in front of the course among the positions where the road width detected in step S32 increases.

  Returning to FIG. 5, in step S15, as shown in FIG. 13B, a range of a predetermined distance in the direction along the road from the position indicated by the position information calculated in step S34 is set as a narrowed area NA. To do. And ECU20 makes the candidate which belongs to narrowing-down area | region NA the object of narrowing down.

  As described above, in the fourth embodiment, the ECU 20 detects the position where the road width increases in front of the course based on the imaging result of the camera 32, and detects this position when the area where the road width increases is detected. Is acquired as position information. When a position where a road shape change appears in front of the course, the appearance of this position can be determined by an increase in road width. In this regard, in the above configuration, a position where the road width increases in front of the course is detected based on the imaging result, and the detected position is acquired as position information. In this case, position information for narrowing down candidates can be acquired without using a map, so a range for narrowing down candidates can be set even if the position at which the road shape change occurs is not recorded on the map. It becomes possible to do.

(Other embodiments)
The shape changing portion on the road may be an entrance or exit of a retreat path provided on the road side in addition to a lane junction or a lane junction. In this case, the ECU 20 recognizes the shape change portion indicating the entrance or exit of the retreat path in the road marking in step S12 of FIG.

  ECU20 replaces with specifying the own vehicle position on a map based on the shape change part in the specified road marking, lane keep control which maintains the own vehicle on the own lane based on this shape change part, shape change Vehicle-to-vehicle communication for notifying the vehicle to the surrounding vehicle may be performed.

  DESCRIPTION OF SYMBOLS 21 ... Marking recognition part, 22 ... Extraction part, 23 ... Information acquisition part, 24 ... Narrowing part, 25 ... Marking specification part, 26 ... Position estimation part.

Claims (6)

A sign recognition unit for recognizing a road marking indicating a boundary of a lane in front of the course of the host vehicle based on an image pickup result by the image pickup unit (32);
In the recognized road marking, the shape changes at an angle corresponding to the angle of the branch lane relative to the main line at the lane branch point on the map or the angle of the merging lane relative to the main line at the lane junction. An angle changing unit to extract as candidates,
An information acquisition unit for acquiring position information indicating a position of a branch point or a merging point of the lane in front of the course of the host vehicle;
A narrowing unit that narrows down the extracted candidates to the candidates existing within a predetermined distance in a direction along the own lane from the position indicated by the position information;
A sign identifying unit that identifies the narrowed candidates as a branch point or a merge point of the lane in front of the course;
A vehicle position estimation apparatus comprising: a position estimation unit that estimates a vehicle position on a map based on the identified candidate. Based on the output from the positioning sensor (31, 35, 36), a positioning position acquisition unit that acquires the vehicle position on the map is provided.
The information acquisition unit searches for a branch point or a merge point of the lane located within a predetermined distance from the vehicle position on the map acquired by the positioning position acquisition unit, and searches for the branch point of the lane Or the own vehicle position estimation apparatus of Claim 1 which acquires the position of a confluence | merging point as said position information. A road width determination unit that determines whether or not the road width has increased in front of the course based on the imaging result,
The host vehicle position estimation apparatus according to claim 1, wherein the information acquisition unit acquires the position information when a position where the road width increases is detected by the road width determination unit. The road marking is road paint,
Based on the surrounding luminance pattern including the candidate in the width direction of the road, it is determined whether or not the candidate is the road surface paint, and the candidate that is not determined to be the road surface paint is selected from the targets to be narrowed by the narrowing unit. The own vehicle position estimation apparatus according to any one of claims 1 to 3, further comprising a luminance pattern exclusion unit to be excluded.   5. The tire mark removing unit according to claim 1, further comprising: a tire mark removing unit that determines whether the candidate is a tire mark and excludes the candidate determined as the tire mark from a target of narrowing by the narrowing unit. The vehicle position estimation device according to one item. A sign recognition step for recognizing a road marking indicating a boundary of a lane in front of the course of the host vehicle based on an imaging result by the imaging unit;
In the recognized road marking, the shape changes at an angle corresponding to the angle of the branch lane relative to the main line at the lane branch point on the map or the angle of the merging lane relative to the main line at the lane junction. An extraction step of extracting the angle changing part as a candidate;
An information acquisition step of acquiring position information indicating a position of a branch point or a merging point of the lane in front of the course of the host vehicle;
A narrowing step of narrowing down the extracted candidates to the candidates existing within a predetermined distance in a direction along the own lane from the position indicated by the position information;
An indication specifying step for specifying the narrowed candidates as a branch point or a junction of the lane in front of the course;
A vehicle position estimation method comprising: a position estimation step of identifying a vehicle position on a map based on the identified candidate.

Images