CN106407887A - Method and apparatus for acquiring step size in search of candidate frame - Google Patents
Method and apparatus for acquiring step size in search of candidate frame Download PDFInfo
- Publication number
- CN106407887A CN106407887A CN201610716184.0A CN201610716184A CN106407887A CN 106407887 A CN106407887 A CN 106407887A CN 201610716184 A CN201610716184 A CN 201610716184A CN 106407887 A CN106407887 A CN 106407887A
- Authority
- CN
- China
- Prior art keywords
- search
- raw partition
- target
- candidate frame
- divisional type
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/94—Hardware or software architectures specially adapted for image or video understanding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and apparatus for acquiring step size in search of a candidate frame, relates to the field of computer vision and object detection in pattern recognition, and can determine the step size in search of an object candidate frame in a to-be-detected video by establishing a Poisson distribution model. The method comprises the following steps: acquiring a to-be-searched image; acquiring image information of every original section of the to-be-searched image and the Poisson distribution function corresponding to every original section; based on the image information and the Poisson distribution function of every original section, determining the type of the section of every original section; and based on the type of the section, determining the step size in search of the candidate frame in every original section. According to the embodiments of the invention, the technical solution can be applied to object detection, such as pedestrian and vehicle detection in scenes like static monitoring videos and onboard monitoring videos.
Description
【Technical field】
The present invention relates to the object detection field in computer vision and pattern-recognition, more particularly, to a kind of candidate frame search
The acquisition methods of step-length and device.
【Background technology】
Developing rapidly and extensively apply with computer image processing technology, for target detection technique demand also by
Gradually rise.Target detection has become as computer vision and the basic problem of area of pattern recognition, and detects the candidate frame of target
The determination of step-size in search is an important previous work of target identification classification.The existing at present side generating target candidate frame
Method be usually sliding window way of search, sliding window way of search when carrying out target search, candidate frame whole scanning window with
Fixing length stepping.
In realizing process of the present invention, inventor finds that in prior art, at least there are the following problems：
According to existing Target Searching Method, during target is scanned for, candidate frame is in whole scanning window
With fixing length stepping, all with fixing length step-searching it is possible that leaking in the different region of search target number
Inspection, Search Results Bu Shi global optimum.
【Content of the invention】
In view of this, a kind of acquisition methods of candidate frame step-size in search and device, Ke Yigen are embodiments provided
According to search target, the frequency occurring in region and density information determine candidate frame step-size in search.
On the one hand, a kind of acquisition methods of candidate frame step-size in search are embodiments provided, methods described includes：
Obtain image to be searched；
Obtain the image information of each raw partition and each raw partition in described image to be searched each self-corresponding
Poisson distribution function；
Image information according to each raw partition and Poisson distribution function, determine the divisional type of each raw partition；
According to described divisional type, determine the candidate frame step-size in search in each raw partition.
On the other hand, embodiments provide a kind of acquisition device of candidate frame step-size in search, described device includes：
First acquisition unit, for obtaining image to be searched；
Second acquisition unit, for obtaining in described image to be searched the image information of each raw partition and each is former
The each self-corresponding Poisson distribution function of beginning subregion；
First determining unit, for the image information according to each raw partition and Poisson distribution function, determines that each is former
The divisional type of beginning subregion；
Second determining unit, for according to described divisional type, determining the candidate frame search step in each raw partition
Long.
A kind of acquisition methods of candidate frame step-size in search provided in an embodiment of the present invention and device, set up Poisson by subregion
Model constantly study update the positional information of the specific search target obtaining, and can adjust the time in every piece of region in every two field picture
Select frame step-size in search, the quantity of constrained objective candidate frame, thus detecting to search target for different regions.This side
Method and device improve Detection results, it is possible to obtain higher recall rate.
【Brief description】
In order to be illustrated more clearly that the technical scheme of the embodiment of the present invention, below will be attached to use required in embodiment
Figure be briefly described it should be apparent that, drawings in the following description are only some embodiments of the present invention, for this area
For those of ordinary skill, without having to pay creative labor, can also be other attached according to the acquisition of these accompanying drawings
Figure.
Fig. 1 is a kind of acquisition methods flow chart of candidate frame step-size in search provided in an embodiment of the present invention；
Fig. 2 is the acquisition methods flow chart of another kind candidate frame step-size in search provided in an embodiment of the present invention；
Fig. 3 is the acquisition methods flow chart of another kind candidate frame step-size in search provided in an embodiment of the present invention；
Fig. 4 is the acquisition methods flow chart of another kind candidate frame step-size in search provided in an embodiment of the present invention；
Fig. 5 is a kind of composition frame chart of the acquisition device of candidate frame step-size in search provided in an embodiment of the present invention；
Fig. 6 is the composition frame chart of the acquisition device of another kind candidate frame step-size in search provided in an embodiment of the present invention；
Fig. 7 is the composition frame chart of the acquisition device of another kind candidate frame step-size in search provided in an embodiment of the present invention；
Fig. 8 is the composition frame chart of the acquisition device of another kind candidate frame step-size in search provided in an embodiment of the present invention.
【Specific embodiment】
In order to be better understood from technical scheme, below in conjunction with the accompanying drawings the embodiment of the present invention is retouched in detail
State.
It will be appreciated that described embodiment is only a part of embodiment of the present invention, rather than whole embodiment.Base
Embodiment in the present invention, those of ordinary skill in the art obtained under the premise of not making creative work all its
Its embodiment, broadly falls into the scope of protection of the invention.
The term using in embodiments of the present invention is the purpose only merely for description specific embodiment, and is not intended to be limiting
The present invention." a kind of ", " described " and " being somebody's turn to do " of singulative used in the embodiment of the present invention and appended claims
It is also intended to including most forms, unless context clearly shows that other implications.
It should be appreciated that term "and/or" used herein is only a kind of incidence relation of description affiliated partner, represent
There may be three kinds of relations, for example, A and/or B, can represent：, there is A and B in individualism A simultaneously, individualism B these three
Situation.In addition, character "/" herein, typically represent forward-backward correlation to as if a kind of relation of "or".
Depending on linguistic context, word as used in this " if " can be construed to " ... when " or " when ...
When " or " in response to determining " or " in response to detection ".Similarly, depending on linguistic context, phrase " if it is determined that " or " if detection
(condition of statement or event) " can be construed to " when determining " or " in response to determining " or " when the detection (condition of statement
Or event) when " or " in response to detecting (condition of statement or event) ".
Embodiments provide a kind of acquisition methods of candidate frame step-size in search, can be applied to including static monitoring
During the target detection such as pedestrian detection, vehicle detection in the scenes such as video, vehicle-mounted monitoring video, as shown in figure 1, methods described
Including：
101st, obtain image to be searched.
Wherein, described image to be searched refers to the altimetric image all to be checked during target detection.
102nd, the image information of each raw partition and each raw partition in described image to be searched are obtained each right
The Poisson distribution function answered.
Wherein it is desired to explanation, the embodiment of the present invention is for the scene such as static surveillance video and vehicle-mounted monitoring video
Monitoring, in image, search target occurs in the frequency in certain block region and distribution density obeys Poisson distribution.
Wherein, each raw partition described refers to detection zone is carried out the regional after piecemeal.
Wherein, the image information of each raw partition described include searching for the frequency information that target occurs in every piece of region with
And the density information of search target distribution.
Wherein, described search target refers to the object to be detected during target detection, such as people, vehicle and object
Deng.
Wherein, described Poisson distribution function is set up based on mathematics Poisson model function is it is adaptable in the description unit interval
The number of times that chance event occurs.
103rd, the image information according to each raw partition and Poisson distribution function, determines the subregion class of each raw partition
Type.
Wherein, described divisional type is by how many determinations searching for target number in raw partition.
Wherein, described search target number refers to the quantity of search target distribution in the unit interval, goes out with search target
Existing frequency information is related with the density information of distribution.
104th, according to described divisional type, determine the candidate frame step-size in search in each raw partition.
A kind of acquisition methods of candidate frame step-size in search provided in an embodiment of the present invention, set up Poisson model simultaneously by subregion
Constantly study updates the positional information of the specific objective obtaining, and can adjust the candidate frame search step in every piece of region in every two field picture
Long, the quantity of constrained objective candidate frame, thus detect to target for different regions.This method improves detection effect
Really, it is possible to obtain higher recall rate.
Furthermore, it is understood that combining preceding method flow process, in the alternatively possible implementation of the embodiment of the present invention, pin
To step 103 according to the image information of each raw partition and Poisson distribution function, determine the divisional type of each raw partition
Realization provide flow process in detail below, as shown in Fig. 2 include：
201st, the image information according to each raw partition and Poisson distribution function, determines search mesh in each raw partition
Target number.
Wherein, described Poisson distribution function has two parameters in continuous distributed, and one is that in time domain, chance event occurs
Frequency, another is the distribution density of chance event in spatial domain, in embodiments of the present invention, chance event in described time domain
The frequency occurring refers to search for the frequency that target occurs in each raw partition, the distribution of chance event in described spatial domain
Density refers to search for the density of target distribution.
Wherein, described Poisson distribution function is trained during search target is detected, in learning process
Middle dynamic access searches for target in every piece of region frequency information occurring and the distribution density information searching for target.
Wherein, when detecting to search target, to each frame, image to be searched detects, Poisson distribution function becomes
For discrete distribution it is determined that the time, the number of search target is the density of search target distribution in the unit interval.
202nd, when the number of described search target is in the range of first threshold, then by the number of search target in the first threshold
The divisional type of the raw partition in the range of value is defined as target sparse distributed areas.
Wherein, a fewer number range of described first threshold scope refers to search target number, be designated as [L,
Lmin).
Wherein, described L is the integer more than 0.
Wherein, described Lmin is the integer more than L.
203rd, when the number of described search target is in the range of Second Threshold, then by the number of search target in the first threshold
The divisional type of the raw partition in the range of value is defined as the medium distributed areas of target.
Wherein, described Second Threshold scope refers to search for the number model that target number is more than first threshold scope
Enclose, be designated as [Lmin, Lmax].
Wherein, described Lmax is the integer more than Lmin.
204th, when the number of described search target is in the 3rd threshold range, then by the number of search target in the first threshold
The divisional type of the raw partition in the range of value is defined as heavy dense targets distributed areas.
Wherein, described 3rd threshold range refers to search for the number model that target number is more than Second Threshold scope
Enclose, be designated as [Lmax ,+∝).
Wherein, when the search target number of described raw partition is more than Lmin, it is more that raw partition is considered as target appearance
Region.
Wherein, when the search target number of described raw partition is less than L, raw partition is considered as target and less area
Domain.
Wherein, for search target, more region occurs and determine divisional type, for search target, less area occurs
Domain directly determines the step-size in search of candidate frame according to the number of search target.
Furthermore, it is understood that combining preceding method flow process, provide in the alternatively possible implementation of the embodiment of the present invention
How according to described divisional type, determine the concrete steps of the candidate frame step-size in search in each raw partition, for step
Rapid 104 realization provides flow process in detail below, as shown in figure 3, including：
301st, the candidate frame step-size in search setting in the raw partition being target sparse distributed areas by described divisional type
For the first step-length.
Wherein, described first step-length can be according to the search target distribution situation self-defining of described target sparse distributed areas
Length.
302nd, the candidate frame step-size in search setting in the raw partition being the medium distributed areas of target by described divisional type
For the second step-length.
Wherein, described second step-length can be according to the search target distribution situation self-defining of the medium distributed areas of described target
Length.
Wherein, described second step-length is less than described first step-length.
303rd, the candidate frame step-size in search setting in the raw partition being heavy dense targets distributed areas by described divisional type
For the 3rd step-length.
Wherein, described 3rd step-length can be according to the search target distribution situation self-defining of described heavy dense targets distributed areas
Length.
Wherein, described 3rd step-length is less than described second step-length.
It should be noted that when raw partition belongs to the less region of search target appearance, being occurred according to search target
Number directly determine step-length, described step-length be not less than described first step-length.
Furthermore, it is understood that combining preceding method flow process, embodiments provide alternatively possible implementation, such as
Shown in Fig. 4, before described acquisition image to be searched, also include：
401st, obtain the raw partition in original image.
Wherein, described original image refers to the n two field picture in detection zone.
Wherein, described n is greater than 0 integer.
Wherein, described raw partition carries out piecemeal according to detection zone size and feature.
Wherein, when described raw partition number is more, function more can accurately react intrapartition destination mark existing frequency and
Distribution density information, the statistical distribution within each raw partition is believed that basically identical.
Wherein, when the number of described raw partition is more, calculative data is more, from the complexity of computational methods
From the aspect of order of accuarcy two, the number of described raw partition determines according to the size of region of search and feature.
402nd, the frequency of occurrences of search target and distribution density in described raw partition are gathered.
Wherein, the described frequency of occurrences gathering described raw partition interior search target and distribution density refer to described in collection
Each raw partition of n two field picture searches for the frequency of occurrences and the distribution density of target.
403rd, the frequency of occurrences according to described search target and density, determine the Poisson distribution function of described raw partition.
Wherein, described Poisson distribution function passes through to gather the frequency of occurrences of search target and density letter in described n two field picture
Breath carries out parameter Estimation and obtains initial value.
A kind of acquisition methods of candidate frame step-size in search provided in an embodiment of the present invention, set up Poisson model simultaneously by subregion
Constantly study updates the positional information of the specific search target obtaining, and the candidate frame that can adjust every piece of region in every two field picture is searched
Suo Buchang, the quantity of constrained objective candidate frame, thus detect to search target for different regions.This method lifting
Detection results, it is possible to obtain higher recall rate.
In order to be better understood from the technical program, embodiments provide more specific embodiment, need
The bright embodiment of the present invention is suitable for but is not limited to implementation below.
Step 1, detection zone is divided into 12 (3x4) block zonules, pedestrian is set up to every piece of region of front 2400 two field pictures
Poisson position go forward side by side line parameter estimate obtain model initial value, such that it is able to obtain the Poisson model unit interval at every piece of region
The average originating rate initial value that one skilled in the art occurs.
Step 2, training Poisson model, the frequency information that dynamic access pedestrian occurs in every piece of region in learning process with
And the density information of pedestrian's distribution.
Step 21, according to set up Poisson model initial value using gradient descent method study renewal obtain pedestrian in every frame
The density information of each raw partition distribution in image.
Step 22 is when each band of position pedestrian obtaining for step 21 is distributed, for embodiment Scene, right
In pedestrian, more segmented areas occur：
When within the pedestrian's number occurring being 10, this region is regarded as pedestrian's sparse distribution region；
When the pedestrian's number occurring is 10 to 30, this region is regarded as the medium distributed areas of pedestrian；
When number of times pedestrian is more than 30, this region is regarded as pedestrian's dense distribution region.
Step 23, pedestrian is there are to less segmented areas, the frequency threshold value that pedestrian in setting is 8.
Step 3, the density information being distributed in every two field picture according to the pedestrian that the study of Poisson position model obtains, in search mesh
The region of mark different distributions density collection arranges the different step-size in search of candidate frame, controls the candidate frame number of whole two field picture.
Step 31, for search target distribution more region setting step-length as follows：Pedestrian's sparse distribution region step=
16, pedestrian medium distributed areas step=8, pedestrian's dense distribution region step=4.
Step 32, for search target distribution less region setting step-length as follows：8 are more than for number of times pedestrian
Situation, step-length step=16, conversely, setting step-length step=32.
In embodiments of the invention, the ratio of width to height of the candidate frame of each region setting is locked as 0.5, in addition, for each region
Place occurs without the situation of pedestrian, then do not generate candidate frame.
A kind of acquisition methods of candidate frame step-size in search provided in an embodiment of the present invention, set up Poisson model simultaneously by subregion
Constantly study updates the positional information of the specific search target obtaining, and the candidate frame that can adjust every piece of region in every two field picture is searched
Suo Buchang, the quantity of constrained objective candidate frame, thus detect to search target for different regions.This method lifting
Detection results, it is possible to obtain higher recall rate.
Embodiments provide a kind of acquisition device of candidate frame step-size in search, can be used for realizing aforementioned approaches method stream
Journey, its composition is as shown in figure 5, described device includes：
First acquisition unit 51, for obtaining image to be searched.
Second acquisition unit 52, for obtain in described image to be searched the image information of each raw partition and each
The each self-corresponding Poisson distribution function of raw partition.
First determining unit 53, for the image information according to each raw partition and Poisson distribution function, determines each
The divisional type of raw partition.
Second determining unit 54, for according to described divisional type, determining the candidate frame search in each raw partition
Step-length.
Optionally, as shown in fig. 6, described first determining unit 53 includes：
First determining module 531, for the image information according to each raw partition and Poisson distribution function, determines each
The number of target is searched in raw partition.
Second determining module 532, for when the number of described search target is in the range of first threshold, searching for target
The divisional type of raw partition in the range of first threshold for the number be defined as target sparse distributed areas.
3rd determining module 533, for when the number of described search target is in the range of Second Threshold, searching for target
The divisional type of raw partition in the range of first threshold for the number be defined as the medium distributed areas of target.
4th determining module 534, for when the number of described search target is in the 3rd threshold range, searching for target
The divisional type of raw partition in the range of first threshold for the number be defined as heavy dense targets distributed areas.
Optionally, as shown in fig. 7, described second determining unit 54 includes：
First setup module 541, for by described divisional type be target sparse distributed areas raw partition in time
Frame step-size in search is selected to be set to the first step-length.
Second setup module 542, for by described divisional type be the medium distributed areas of target raw partition in time
Frame step-size in search is selected to be set to the second step-length.
3rd setup module 543, for by described divisional type be heavy dense targets distributed areas raw partition in time
Frame step-size in search is selected to be set to the 3rd step-length.
Optionally, as shown in figure 8, described device also includes：
3rd acquiring unit 55, for obtaining the raw partition in original image.
Collecting unit 56, for gathering the frequency of occurrences and the distribution density searching for target in described raw partition.
3rd determining unit 57, for the frequency of occurrences according to described search target and density, determines described raw partition
Poisson distribution function.
A kind of acquisition device of candidate frame step-size in search provided in an embodiment of the present invention, sets up Poisson model simultaneously by subregion
Constantly study updates the positional information of the specific search target obtaining, and the candidate frame that can adjust every piece of region in every two field picture is searched
Suo Buchang, the quantity of constrained objective candidate frame, thus detect to search target for different regions.This device lifting
Detection results, it is possible to obtain higher recall rate.
Those skilled in the art can be understood that, for convenience and simplicity of description, the system of foregoing description,
Device and the specific work process of unit, may be referred to the corresponding process in preceding method embodiment, will not be described here.
It should be understood that disclosed system in several embodiments provided by the present invention, apparatus and method are permissible
Realize by another way.For example, device embodiment described above is only schematically, for example, described unit
Divide, only a kind of division of logic function, actual can have other dividing mode when realizing, for example, multiple units or group
Part can in conjunction with or be desirably integrated into another system, or some features can be ignored, or does not execute.Another, shown
Or the coupling each other that discusses or direct-coupling or communication connection can be by some interfaces, device or unit indirect
Coupling or communication connection, can be electrical, mechanical or other forms.
The described unit illustrating as separating component can be or may not be physically separate, show as unit
The part showing can be or may not be physical location, you can with positioned at a place, or can also be distributed to multiple
On NE.The mesh to realize this embodiment scheme for some or all of unit therein can be selected according to the actual needs
's.
In addition, can be integrated in a processing unit in each functional unit in each embodiment of the present invention it is also possible to
It is that unit is individually physically present it is also possible to two or more units are integrated in a unit.Above-mentioned integrated list
Unit both can be to be realized in the form of hardware, it would however also be possible to employ the form that hardware adds SFU software functional unit is realized.
The above-mentioned integrated unit realized in the form of SFU software functional unit, can be stored in an embodied on computer readable and deposit
In storage media.Above-mentioned SFU software functional unit is stored in a storage medium, including some instructions with so that a computer
Device (can be personal computer, server, or network equipment etc.) or processor (Processor) execution the present invention each
The part steps of embodiment methods described.And aforesaid storage medium includes：USB flash disk, portable hard drive, read-only storage (Read-
Only Memory, ROM), random access memory (Random Access Memory, RAM), magnetic disc or CD etc. various
Can be with the medium of store program codes.
The foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all essences in the present invention
Within god and principle, any modification, equivalent substitution and improvement done etc., should be included within the scope of protection of the invention.
Claims (8)
1. a kind of acquisition methods of candidate frame step-size in search are it is characterised in that include：
Obtain image to be searched；
Obtain the image information of each raw partition and each self-corresponding Poisson of each raw partition in described image to be searched
Distribution function；
Image information according to each raw partition and Poisson distribution function, determine the divisional type of each raw partition；
According to described divisional type, determine the candidate frame step-size in search in each raw partition.
2. method according to claim 1 is it is characterised in that the described image information according to each raw partition and Poisson
Distribution function, determines that the divisional type of each raw partition includes：
Image information according to each raw partition and Poisson distribution function, determine and search for the individual of target in each raw partition
Number；
When the number of described search target is in the range of first threshold, then by the number of search target in the range of first threshold
The divisional type of raw partition be defined as target sparse distributed areas；
When the number of described search target is in the range of Second Threshold, then by the number of search target in the range of first threshold
The divisional type of raw partition be defined as the medium distributed areas of target；
When the number of described search target is in the 3rd threshold range, then by the number of search target in the range of first threshold
The divisional type of raw partition be defined as heavy dense targets distributed areas.
3. method according to claim 2 it is characterised in that described according to described divisional type, determine original at each
Candidate frame step-size in search in subregion includes：
The candidate frame step-size in search in raw partition that described divisional type is target sparse distributed areas is set to the first step
Long；
The candidate frame step-size in search in raw partition that described divisional type is the medium distributed areas of target is set to second step
Long；
The candidate frame step-size in search in raw partition that described divisional type is heavy dense targets distributed areas is set to the 3rd step
Long.
4. method according to claim 1 is it is characterised in that before described acquisition image to be searched, also include：
Obtain the raw partition in original image；
Gather the frequency of occurrences of search target and distribution density in described raw partition；
The frequency of occurrences according to described search target and density, determine the Poisson distribution function of described raw partition.
5. a kind of acquisition device of candidate frame step-size in search is it is characterised in that described device includes：
First acquisition unit, for obtaining image to be searched；
Second acquisition unit, for obtaining in described image to be searched the image information of each raw partition and each original point
The each self-corresponding Poisson distribution function in area；
First determining unit, for the image information according to each raw partition and Poisson distribution function, determines each original point
The divisional type in area；
Second determining unit, for according to described divisional type, determining the candidate frame step-size in search in each raw partition.
6. device according to claim 5 is it is characterised in that described first determining unit includes：
First determining module, for the image information according to each raw partition and Poisson distribution function, determines each original point
The number of target is searched in area；
Second determining module, for when the number of described search target is in the range of first threshold, by the number of search target
The divisional type of the raw partition in the range of first threshold is defined as target sparse distributed areas；
3rd determining module, for when the number of described search target is in the range of Second Threshold, by the number of search target
The divisional type of the raw partition in the range of first threshold is defined as the medium distributed areas of target；
4th determining module, for when the number of described search target is in the 3rd threshold range, by the number of search target
The divisional type of the raw partition in the range of first threshold is defined as heavy dense targets distributed areas.
7. device according to claim 6 is it is characterised in that described second determining unit includes：
First setup module, for by described divisional type be target sparse distributed areas raw partition in candidate frame search
Step-length is set to the first step-length；
Second setup module, for by described divisional type be the medium distributed areas of target raw partition in candidate frame search
Step-length is set to the second step-length；
3rd setup module, for by described divisional type be heavy dense targets distributed areas raw partition in candidate frame search
Step-length is set to the 3rd step-length.
8. device according to claim 5 is it is characterised in that described device also includes：
3rd acquiring unit, for obtaining the raw partition in original image；
Collecting unit, for gathering the frequency of occurrences and the distribution density searching for target in described raw partition；
3rd determining unit, for the frequency of occurrences according to described search target and density, determines the Poisson of described raw partition
Distribution function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610716184.0A CN106407887B (en) | 2016-08-24 | 2016-08-24 | Method and device for acquiring search step length of candidate frame |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610716184.0A CN106407887B (en) | 2016-08-24 | 2016-08-24 | Method and device for acquiring search step length of candidate frame |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106407887A true CN106407887A (en) | 2017-02-15 |
| CN106407887B CN106407887B (en) | 2020-07-31 |
Family
ID=58004405
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610716184.0A Active CN106407887B (en) | 2016-08-24 | 2016-08-24 | Method and device for acquiring search step length of candidate frame |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106407887B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101144784A (en) * | 2007-09-04 | 2008-03-19 | 杭州电子科技大学 | Automatic tracking method for video frequency microscopic image cell |
| CN103150566A (en) * | 2011-12-06 | 2013-06-12 | 中国科学院电子学研究所 | Automatic detecting method of remote sensing ground object target based on random geometric model |
| CN104715451A (en) * | 2015-03-11 | 2015-06-17 | 西安交通大学 | Seamless image fusion method based on consistent optimization of color and transparency |
| CN104778358A (en) * | 2015-04-09 | 2015-07-15 | 西安工程大学 | Method for tracking extended target by multiple sensors with partially overlapped monitoring areas |
| CN105205450A (en) * | 2015-08-24 | 2015-12-30 | 辽宁工程技术大学 | SAR image target extraction method based on irregular marked point process |
-
2016
- 2016-08-24 CN CN201610716184.0A patent/CN106407887B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101144784A (en) * | 2007-09-04 | 2008-03-19 | 杭州电子科技大学 | Automatic tracking method for video frequency microscopic image cell |
| CN103150566A (en) * | 2011-12-06 | 2013-06-12 | 中国科学院电子学研究所 | Automatic detecting method of remote sensing ground object target based on random geometric model |
| CN104715451A (en) * | 2015-03-11 | 2015-06-17 | 西安交通大学 | Seamless image fusion method based on consistent optimization of color and transparency |
| CN104778358A (en) * | 2015-04-09 | 2015-07-15 | 西安工程大学 | Method for tracking extended target by multiple sensors with partially overlapped monitoring areas |
| CN105205450A (en) * | 2015-08-24 | 2015-12-30 | 辽宁工程技术大学 | SAR image target extraction method based on irregular marked point process |
Non-Patent Citations (4)
| Title |
|---|
| K.GILHOL等: ""Spatial distribution model for tracking extended objects"", 《IEEE PROCEEDINGS-RADAR,SONAR AND NAVIGATION》 * |
| KEVIN GILHOLM等: ""Poisson models for extended target and group tracking"", 《SIGNAL & DATA PROCESSING OF SMALL TARGETS》 * |
| 张慧: ""天基红外传感器对中段目标群跟踪技术研究"", 《中国博士学位论文全文数据库 工程科技Ⅱ辑》 * |
| 覃剑等: ""采用在线高斯模型的行人检测候选框快速生成方法"", 《光学学报》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106407887B (en) | 2020-07-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111696128B (en) | High-speed multi-target detection tracking and target image optimization method and storage medium | |
| CN109977782B (en) | Cross-store operation behavior detection method based on target position information reasoning | |
| CN111709416B (en) | License plate positioning method, device, system and storage medium | |
| CN100545867C (en) | Aerial shooting traffic video frequency vehicle rapid checking method | |
| CN102509075B (en) | Remnant object detection method and device | |
| CN102496001B (en) | Method of video monitor object automatic detection and system thereof | |
| CN109800682B (en) | Driver attribute identification method and related product | |
| JP2002373332A (en) | Device and method for recognizing image | |
| CN106611512A (en) | Preceding start processing method, device and system | |
| CN105184291B (en) | A kind of polymorphic type detection method of license plate and system | |
| CN111898581A (en) | Animal detection method, device, electronic equipment and readable storage medium | |
| CN107529659B (en) | Seatbelt wearing detection method, device and electronic equipment | |
| CN109840904B (en) | Detection method for large-scale difference parts of high-speed rail contact network | |
| JP7374001B2 (en) | Learning model generation method, program, and information processing device | |
| CN108460336A (en) | A kind of pedestrian detection method based on deep learning | |
| US11948373B2 (en) | Automatic license plate recognition | |
| JP2002190013A (en) | Traffic jam detection system and traffic jam detection method by image recognition | |
| CN109800678A (en) | The attribute determining method and device of object in a kind of video | |
| CN106407887A (en) | Method and apparatus for acquiring step size in search of candidate frame | |
| CN112836697A (en) | Yoov 5-based vehicle type detection and quality estimation method | |
| CN111639597A (en) | Detection method of flag-raising touring event | |
| CN108960199A (en) | Target pedestrian detection method, device and electronic equipment | |
| CN116051927A (en) | Track recognition model training method, track recognition device and track recognition system | |
| JP7383954B2 (en) | Learning data collection device | |
| JPH09288795A (en) | Car number reader |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20200325 Address after: 400044 Shapingba District Sha Street, No. 174, Chongqing Applicant after: Chongqing University Address before: 400044 Shapingba District Sha Street, No. 174, Chongqing Applicant before: Chongqing University Applicant before: SHENZHEN TINNO WIRELESS TECHNOLOGY Co.,Ltd. |
|
| TA01 | Transfer of patent application right | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |