CN108647624A - A kind of road vehicle detection method, device, equipment and storage medium - Google Patents
A kind of road vehicle detection method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN108647624A CN108647624A CN201810419768.0A CN201810419768A CN108647624A CN 108647624 A CN108647624 A CN 108647624A CN 201810419768 A CN201810419768 A CN 201810419768A CN 108647624 A CN108647624 A CN 108647624A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- component
- road
- viewpoint
- sample
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- 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
- G06V2201/08—Detecting or categorising vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
This application discloses a kind of road vehicle detection method, device, equipment and storage mediums, including：Component display model is established for the different components of vehicle sample under multiple viewpoints, and exports the overall appearance information of vehicle sample；Road structure probability viewpoint figure is generated according to road information structure and viewpoint distribution, the view prediction when vehicle sample is in specific position is provided；Target vehicle component detection is carried out by the matching of current road image and component display model；In conjunction with overall appearance information, view prediction result and target vehicle component testing result, the probability that target vehicle component constitutes vehicle is inferred.The application passes through in different points of view off-line learning component display model and road structure probability viewpoint figure, it is provided for vehicle reconstruct and compares help, and carry out the detection of online processing vehicle part and vehicle probability inference, the robustness and precision of vehicle detection are improved under complicated more traffic environments in this way, multiple views and occlusion issue are solved, while reducing calculating cost.
Description
Technical field
The present invention relates to road vehicle detection field, more particularly to a kind of road vehicle detection method, device, equipment and
Storage medium.
Background technology
Currently, road vehicle detection method is widely studied, it is efficient driving assistance system (Advanced
Driver assistance systems, ADAS) and automated driving system critical issue.
Existing method is using the various feature such as symmetry of automobile appearance, edge, and characteristics of image such as shade etc. are wide
It is general to be used for automotive check and classification, such as Haar features and orientation characteristic histogram.But the appearance of vehicle may be regarded in different
Change on point very greatly, or sometimes due to blocking for basic common structure can only observe part vapour with limited camera coverage
The problem of vehicle parts etc. are brought.
Therefore, the multiple views and occlusion issue under multilane mode of transportation scene how are solved, are those skilled in the art
Technical problem urgently to be resolved hurrily.
Invention content
In view of this, the purpose of the present invention is to provide a kind of road vehicle detection method, device, equipment and storages to be situated between
Matter can improve the precision and robustness of vehicle detection, solve the multiple views under multilane mode of transportation scene and block to ask
Topic.Its concrete scheme is as follows：
A kind of road vehicle detection method, including：
Component display model is established for the different components of vehicle sample under multiple viewpoints, and exports the entirety of vehicle sample
Appearance information；
Road structure probability viewpoint figure is generated according to road information structure and viewpoint distribution, is provided when vehicle sample is specific
View prediction when position；
Target vehicle component detection is carried out by the matching of current road image and the component display model；
In conjunction with the overall appearance information, the view prediction result and the target vehicle component testing result, reasoning
Go out the probability that target vehicle component constitutes vehicle.
Preferably, it is vehicle under multiple viewpoints in above-mentioned road vehicle detection method provided in an embodiment of the present invention
The different components of sample establish component display model, specifically include：
It will be mounted on the component filter of the different components of vehicle sample and the entire vehicle sample of covering under multiple viewpoints
The data that obtain of root filter and the component filter carry out HOG relative to the spatial positional information of described filter
Feature detects, and generates HOG characteristic look models；
According to the HOG characteristic looks model, position of the different components of vehicle sample relative to entire vehicle sample is exported
Set offset and view information；
According to the position offset and the view information, the different components of vehicle are trained by the component filter, it is right
All parts carry out space constraint, obtain the component display model in each component of at least four viewpoints.
Preferably, in above-mentioned road vehicle detection method provided in an embodiment of the present invention, according to road information structure and
Viewpoint distribution generates road structure probability viewpoint figure, provides the view prediction when vehicle sample is in specific position, specifically includes：
Pass course figure and viewpoint channel detector obtain traffic direction and lane structure；
According to the traffic direction and lane structure of acquisition, the grid chart of self vehicle frame is generated；
The probability distribution that vehicle sample is observed in grid chart under different points of view is determined, to generate road structure probability viewpoint
Figure；
According to the road structure probability viewpoint figure, the view prediction when vehicle sample is in specific position is provided.
Preferably, in above-mentioned road vehicle detection method provided in an embodiment of the present invention, by current road image and
The matching of the component display model carries out target vehicle component detection, specifically includes：
The study of root filter characteristic is carried out to vehicle sample；
Component filter characteristic is obtained from described filter characteristic；
By the component filter characteristic, the target vehicle component position extracted in current road image is detected
With Viewing-angle information and be added to the component display model；
By the component display model, space constraint is carried out to the target vehicle component.
Preferably, in above-mentioned road vehicle detection method provided in an embodiment of the present invention, current road image is detected
The target vehicle component position of middle extraction and Viewing-angle information, specifically include：
When the target vehicle component extracted in current road image is not blocked, the inspection of the target vehicle component is calculated
Survey score；
When the target vehicle component extracted in current road image is blocked, a steady state value is assigned to substitute the mesh blocked
Mark vehicle part.
The embodiment of the present invention additionally provides a kind of road vehicle detecting device, including：
Model building module, the different components under multiple viewpoints being vehicle sample establish component display model, and
Export the overall appearance information of vehicle sample；
Viewpoint figure generation module, for generating road structure probability viewpoint figure according to road information structure and viewpoint distribution,
View prediction when vehicle sample is in specific position is provided；
Component detection module, for carrying out target vehicle by the matching of current road image and the component display model
Component detects；
Probability inference module, in conjunction with the overall appearance information, the view prediction result and the target vehicle
Component testing result infers the probability that target vehicle component constitutes vehicle.
The embodiment of the present invention additionally provides a kind of road vehicle detection device, including processor and memory, wherein described
Processor realizes such as above-mentioned road vehicle provided in an embodiment of the present invention when executing the computer program preserved in the memory
Detection method.
The embodiment of the present invention additionally provides a kind of computer readable storage medium, for storing computer program, wherein institute
It states and realizes such as above-mentioned road vehicle detection method provided in an embodiment of the present invention when computer program is executed by processor.
A kind of road vehicle detection method, device, equipment and storage medium provided by the present invention, including：It is regarded multiple
Point is lower to establish component display model for the different components of vehicle sample, and exports the overall appearance information of vehicle sample；According to road
Road message structure and viewpoint distribution generate road structure probability viewpoint figure, and the viewpoint provided when vehicle sample is in specific position is pre-
It surveys；Target vehicle component detection is carried out by the matching of current road image and component display model；In conjunction with overall appearance information,
View prediction result and target vehicle component testing result infer the probability that target vehicle component constitutes vehicle.
The present invention is divided into off-line learning and online processing two parts, and component display model is established in different points of view off-line learning
With generation road structure probability viewpoint figure, is provided for vehicle reconstruct and compare help, and carry out online processing vehicle part detection
With vehicle probability inference, the robustness of vehicle detection is improved under complicated more traffic environments in this way, in a probability inference frame
The precision of middle enhancing identification, solves the multiple views and occlusion issue under multilane mode of transportation scene, while reducing and being calculated as
This, has stronger practicability.
Description of the drawings
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The embodiment of invention for those of ordinary skill in the art without creative efforts, can also basis
The attached drawing of offer obtains other attached drawings.
Fig. 1 is the flow chart of road vehicle detection method provided in an embodiment of the present invention；
Fig. 2 is the structural schematic diagram of road vehicle detecting device provided in an embodiment of the present invention.
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation describes, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
The present invention provides a kind of road vehicle detection method, as shown in Figure 1, including the following steps：
S101, for the different components of vehicle sample component display model is established under multiple viewpoints, and exports vehicle sample
Overall appearance information；
S102, road structure probability viewpoint figure is generated according to road information structure and viewpoint distribution, provided when vehicle sample
View prediction in specific position；
S103, target vehicle component detection is carried out by the matching of current road image and component display model；
S104, in conjunction with overall appearance information, view prediction result and target vehicle component testing result, infer target carriage
Component constitutes the probability of vehicle.
In above-mentioned road vehicle detection method provided in an embodiment of the present invention, it can specifically be divided into off-line learning and online
Two parts are handled, establish component display model in different points of view off-line learning first and generate road structure probability viewpoint figure, are
Vehicle reconstruct, which provides, compares help, the detection of online processing vehicle part and vehicle probability inference is then carried out again, in this way in complexity
The robustness that vehicle detection is improved under more traffic environments, enhances the precision of identification in a probability inference frame, solves more
Multiple views and occlusion issue under lane traffic mode scene, while calculating cost is reduced, there is stronger practicability.
Further, in the specific implementation, in above-mentioned road vehicle detection method provided in an embodiment of the present invention, step
S101 establishes component display model under multiple viewpoints for the different components of vehicle sample, can specifically include following steps：
The first step mounted on the component filter of the different components of vehicle sample and will cover entire vehicle under multiple viewpoints
The data and component filter that the root filter of sample obtains carry out HOG relative to the spatial positional information of root filter
(Histogram of Oriented Gradient, histograms of oriented gradients) feature detects, and generates HOG characteristic look models；
It should be noted that HOG characteristic look models are that a vehicle sample is amplified resolution ratio (general 2 times),
HOG feature detections are carried out for all parts.Specifically, the root filter of entire vehicle sample will be covered and with high-resolution
The two is combined matching to get up by the component filter of rate using spatial position structure, is given birth at highest scoring point under different viewpoints
At HOG characteristic look models, Q can be denoted asn=(F0,Pi,...,Pn, b), F0It is root filter, PiIt is the mould of i-th of component
Type, b are the real number values for indicating deviation；Wherein Pn=(Fi,vi,di), FiIt is the filter of i-th of component, viA two dimension to
Amount, diCoordinate of the normal place relative to whole vehicle model when not deformed upon for i-th of filter；
Second step, according to HOG characteristic look models, export the different components of vehicle sample relative to entire vehicle sample
Position offset and view information；
Specifically, it is assumed that use four viewpoints, export the different components of all vehicle samples detected relative to entire
The position offset and view information of vehicle sample, i.e.,Wherein k=1 ..., 4.
Third walks, according to position offset and view information, the different components of vehicle is trained by component filter, to each portion
Part carries out space constraint, obtains the component display model in each component of at least four viewpoints；
Specifically, four viewpoints may include face, it is left front regard, right forward sight, side view；Due to all parts HOG features
There are one position offsets for meeting for sample after detection, and component display model is mainly to HOG characteristic look all parts
Space constraint so that generally position offset is not too large relative to sample for all parts.
Further, in the specific implementation, in above-mentioned road vehicle detection method provided in an embodiment of the present invention, step
S102 generates road structure probability viewpoint figure according to road information structure and viewpoint distribution, provides when vehicle sample is in specific position
When view prediction, can specifically include following steps：
The first step, pass course figure and viewpoint channel detector obtain traffic direction and lane structure；
Second step, traffic direction and lane structure according to acquisition, generate the grid chart of self vehicle frame；
Third step determines the probability distribution that vehicle sample is observed in grid chart under different points of view, to generate road structure
Probability viewpoint figure；
4th step, according to road structure probability viewpoint figure, the view prediction when vehicle sample is in specific position is provided.
Specifically, in the generation of road structure probability viewpoint figure, it is assumed that generate four visual point images in automotive vehicle frame, often
A Pixel Dimensions are 2 meters, and each net of viewpoint figure is obtained using the detect and track result (i.e. track) based on laser radar
Case sets the statistics of the viewpoint at place.For each grid position, by the track of vehicle of grid cell for estimating ballot viewpoint.
Then, the ballot of each grid position is in k=1 ..., is normalized on 4 viewpoint figures.It is seen at grid position i at viewpoint k
Examine the probabilistic forecasting value of vehicle
Further, in the specific implementation, in above-mentioned road vehicle detection method provided in an embodiment of the present invention, step
S103 carries out target vehicle component detection by the matching of current road image and component display model, can specifically include following
Step：
The first step, the study that root filter characteristic is carried out to vehicle sample；
Second step obtains component filter characteristic from root filter characteristic；
Third walks, by component filter characteristic, detects the target vehicle component place extracted in current road image
Position and Viewing-angle information are simultaneously added to component display model；
4th step, by component display model, space constraint is carried out to target vehicle component.
Specifically, learn root filter characteristic from root filter according in the vehicle sample of different points of view firstAgain
FromMiddle acquisition component filter characteristic Fi k(i=1 ... n), n are the number of component filter, generally preset constant
Value.After the component filter characteristic for obtaining different visual angles, index for detecting specific recording-member and viewpoint and its
The corresponding examples of components in position on image detects one group of target vehicle component in each viewpoint kI.e. regarding
Target vehicle component on point k is detected on the corresponding position of this component filter.Wherein,It is a target carriage
Examples of components,Be detection score, the detection score include with HOG to component display model detection it is matched loss and
Offset loss between different components.In addition, in the specific implementation, detecting the target vehicle portion extracted in current road image
Part position and Viewing-angle information, when can specifically include the target vehicle component extracted in current road image and not being blocked,
Calculate the detection score of target vehicle component；When the target vehicle component extracted in current road image is blocked, one is assigned
A steady state value substitutes the target vehicle component blocked.For detecting the probability of component set, vehicle partIf not yet
There are the probability for detecting remainder set, vehicle partDue to blocking or other factors are not observed completely
Automobile is whole, so it is difficult to define one it is rightExplicit estimation, therefore it is regarded as penalizing number, that is, assigns
Give a steady state value φocc=η φmin, mean detector threshold φminWith the product for penalizing (η=0.7) several η, to simulate occlusion part
Part simultaneously filters the examples of components that low quality detects.Following component display model carries out space constraint to target vehicle component,
So that all parts are either either with or without blocking, both relative to sample, generally position offset is not too large.
In view of the set of the detection vehicle part of all main viewpoints, the problem of vehicle reasoning, can be indicated with probabilistic manner
It is as follows：Wherein k=1 ..., 4, rkFor the position vehicle of the vehicle center estimation on viewpoint k
It constitutes.
Target is to find rk, maximizeIt is as follows Bayes rule can be further expanded：Wherein,It can be further simplified as P (rk), it is contemplated that road
Line structure information this be probability of the vehicle at the r of position.Mean the mould of given modular construction and space layout
TypeAnd vehicle, in the probability of position r, observes parts example at viewpoint kSet.AboutIt is's
The a subset of complete structure, because being only able to detect visible parts in the case where parts block,It is
Remaining set of complete structure, whereinIndicate blocking parts,Estimation be converted to as
Under:
Summarize above-mentioned derivation：
Thus, it is possible to infer the probability that target vehicle component constitutes vehicle.
Based on same inventive concept, the embodiment of the present invention additionally provides a kind of road vehicle detecting device, due to the road
The principle that vehicle detection apparatus solves the problems, such as is similar to a kind of aforementioned road vehicle detection method, therefore road vehicle detection dress
The implementation set may refer to the implementation of road vehicle detection method, and overlaps will not be repeated.
In the specific implementation, road vehicle detecting device provided in an embodiment of the present invention, as shown in Fig. 2, specifically including：
Model building module 11, the different components under multiple viewpoints being vehicle sample establish component display model,
And export the overall appearance information of vehicle sample；
Viewpoint figure generation module 12, for generating road structure probability viewpoint according to road information structure and viewpoint distribution
Figure, provides the view prediction when vehicle sample is in specific position；
Component detection module 13, for carrying out target vehicle portion by the matching of current road image and component display model
Part detects；
Probability inference module 14, in conjunction with overall appearance information, view prediction result and target vehicle component detection knot
Fruit infers the probability that target vehicle component constitutes vehicle.
In above-mentioned road vehicle detecting device provided in an embodiment of the present invention, the mutual of aforementioned four module can be passed through
Effect, establishes component display model in different points of view and generates road structure probability viewpoint figure, and comparison side is provided for vehicle reconstruct
It helps, and carries out vehicle part detection and vehicle probability inference, improve the Shandong of vehicle detection under complicated more traffic environments in this way
Stick, enhances the precision of identification in a probability inference frame, solve multiple views under multilane mode of transportation scene and
Occlusion issue, while calculating cost is reduced, there is stronger practicability.
Corresponding contents disclosed in previous embodiment can be referred to about the more specifical course of work of above-mentioned modules,
This is no longer repeated.
Correspondingly, the embodiment of the invention also discloses a kind of road vehicle detection device, including processor and memory；Its
In, processor realizes road vehicle detection method disclosed in previous embodiment when executing the computer program preserved in memory.
Corresponding contents disclosed in previous embodiment can be referred to about the more specifical process of the above method, herein no longer
It is repeated.
Further, the invention also discloses a kind of computer readable storage mediums, for storing computer program；It calculates
Machine program realizes aforementioned disclosed road vehicle detection method when being executed by processor.
Corresponding contents disclosed in previous embodiment can be referred to about the more specifical process of the above method, herein no longer
It is repeated.
Each embodiment is described by the way of progressive in this specification, the highlights of each of the examples are with it is other
The difference of embodiment, just to refer each other for same or similar part between each embodiment.For being filled disclosed in embodiment
It sets, for equipment, storage medium, since it is corresponded to the methods disclosed in the examples, so description is fairly simple, correlation
Place is referring to method part illustration.
Professional further appreciates that, unit described in conjunction with the examples disclosed in the embodiments of the present disclosure
And algorithm steps, can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate hardware and
The interchangeability of software generally describes each exemplary composition and step according to function in the above description.These
Function is implemented in hardware or software actually, depends on the specific application and design constraint of technical solution.Profession
Technical staff can use different methods to achieve the described function each specific application, but this realization is not answered
Think to exceed scope of the present application.
The step of method described in conjunction with the examples disclosed in this document or algorithm, can directly be held with hardware, processor
The combination of capable software module or the two is implemented.Software module can be placed in random access memory (RAM), memory, read-only deposit
Reservoir (ROM), electrically programmable ROM, electrically erasable ROM, register, hard disk, moveable magnetic disc, CD-ROM or technology
In any other form of storage medium well known in field.
A kind of road vehicle detection method, device, equipment and storage medium provided in an embodiment of the present invention, including：More
Component display model is established for the different components of vehicle sample under a viewpoint, and exports the overall appearance information of vehicle sample；Root
Road structure probability viewpoint figure is generated according to road information structure and viewpoint distribution, regarding when vehicle sample is in specific position is provided
Point prediction；Target vehicle component detection is carried out by the matching of current road image and component display model；In conjunction with overall appearance
Information, view prediction result and target vehicle component testing result infer the probability that target vehicle component constitutes vehicle.This hair
It is bright to be divided into off-line learning and online processing two parts, it establishes component display model in different points of view off-line learning and generates road knot
Structure probability viewpoint figure provides for vehicle reconstruct and compares help, and carries out the detection of online processing vehicle part and pushed away with vehicle probability
Reason improves the robustness of vehicle detection under complicated more traffic environments in this way, the enhancing identification in a probability inference frame
Precision solves multiple views and occlusion issue under multilane mode of transportation scene, while reducing calculating cost, has stronger
Practicability.
Finally, it is to be noted that, herein, relational terms such as first and second and the like be used merely to by
One entity or operation are distinguished with another entity or operation, without necessarily requiring or implying these entities or operation
Between there are any actual relationship or orders.Moreover, the terms "include", "comprise" or its any other variant meaning
Covering non-exclusive inclusion, so that the process, method, article or equipment including a series of elements includes not only that
A little elements, but also include other elements that are not explicitly listed, or further include for this process, method, article or
The intrinsic element of equipment.In the absence of more restrictions, the element limited by sentence "including a ...", is not arranged
Except there is also other identical elements in the process, method, article or apparatus that includes the element.
Detailed Jie has been carried out to road vehicle detection method provided by the present invention, device, equipment and storage medium above
It continues, principle and implementation of the present invention are described for specific case used herein, and the explanation of above example is only
It is the method and its core concept for being used to help understand the present invention；Meanwhile for those of ordinary skill in the art, according to this hair
Bright thought, there will be changes in the specific implementation manner and application range, in conclusion the content of the present specification should not manage
Solution is limitation of the present invention.
Claims (8)
1. a kind of road vehicle detection method, which is characterized in that including：
Component display model is established for the different components of vehicle sample under multiple viewpoints, and exports the overall appearance of vehicle sample
Information；
Road structure probability viewpoint figure is generated according to road information structure and viewpoint distribution, is provided when vehicle sample is in specific position
When view prediction；
Target vehicle component detection is carried out by the matching of current road image and the component display model；
In conjunction with the overall appearance information, the view prediction result and the target vehicle component testing result, mesh is inferred
Mark the probability that vehicle part constitutes vehicle.
2. road vehicle detection method according to claim 1, which is characterized in that for vehicle sample under multiple viewpoints
Different components establish component display model, specifically include：
It will be mounted on the root of the component filter of the different components of vehicle sample and the entire vehicle sample of covering under multiple viewpoints
The data and the component filter that filter obtains carry out HOG features relative to the spatial positional information of described filter
Detection generates HOG characteristic look models；
According to the HOG characteristic looks model, the different components for exporting vehicle sample are inclined relative to the position of entire vehicle sample
Shifting and view information；
According to the position offset and the view information, the different components of vehicle are trained by the component filter, to each
Component carries out space constraint, obtains the component display model in each component of at least four viewpoints.
3. road vehicle detection method according to claim 1, which is characterized in that according to road information structure and viewpoint point
Cloth generates road structure probability viewpoint figure, provides the view prediction when vehicle sample is in specific position, specifically includes：
Pass course figure and viewpoint channel detector obtain traffic direction and lane structure；
According to the traffic direction and lane structure of acquisition, the grid chart of self vehicle frame is generated；
The probability distribution that vehicle sample is observed in grid chart under different points of view is determined, to generate road structure probability viewpoint figure；
According to the road structure probability viewpoint figure, the view prediction when vehicle sample is in specific position is provided.
4. road vehicle detection method according to claim 2, which is characterized in that pass through current road image and the portion
The matching of part display model carries out target vehicle component detection, specifically includes：
The study of root filter characteristic is carried out to vehicle sample；
Component filter characteristic is obtained from described filter characteristic；
By the component filter characteristic, detects the target vehicle component position extracted in current road image and regard
Angle information is simultaneously added to the component display model；
By the component display model, space constraint is carried out to the target vehicle component.
5. road vehicle detection method according to claim 4, which is characterized in that detect to extract in current road image
Target vehicle component position and Viewing-angle information, specifically include：
When the target vehicle component extracted in current road image is not blocked, the detection point of the target vehicle component is calculated
Number；
When the target vehicle component extracted in current road image is blocked, a steady state value is assigned to substitute the target carriage blocked
Component.
6. a kind of road vehicle detecting device, which is characterized in that including：
Model building module, the different components under multiple viewpoints being vehicle sample establish component display model, and export
The overall appearance information of vehicle sample；
Viewpoint figure generation module is provided for generating road structure probability viewpoint figure according to road information structure and viewpoint distribution
View prediction when vehicle sample is in specific position；
Component detection module, for carrying out target vehicle component by the matching of current road image and the component display model
Detection；
Probability inference module, in conjunction with the overall appearance information, the view prediction result and the target vehicle component
Testing result infers the probability that target vehicle component constitutes vehicle.
7. a kind of road vehicle detection device, which is characterized in that including processor and memory, wherein the processor executes
Such as road vehicle detection method described in any one of claim 1 to 5 is realized when the computer program preserved in the memory.
8. a kind of computer readable storage medium, which is characterized in that for storing computer program, wherein the computer journey
Such as road vehicle detection method described in any one of claim 1 to 5 is realized when sequence is executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810419768.0A CN108647624A (en) | 2018-05-04 | 2018-05-04 | A kind of road vehicle detection method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810419768.0A CN108647624A (en) | 2018-05-04 | 2018-05-04 | A kind of road vehicle detection method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108647624A true CN108647624A (en) | 2018-10-12 |
Family
ID=63749386
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810419768.0A Pending CN108647624A (en) | 2018-05-04 | 2018-05-04 | A kind of road vehicle detection method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108647624A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102646199A (en) * | 2012-02-29 | 2012-08-22 | 湖北莲花山计算机视觉和信息科学研究院 | Motorcycle type identifying method in complex scene |
| CN104615998A (en) * | 2015-02-15 | 2015-05-13 | 武汉大学 | Vehicle search method based on multiple views |
| US20160267331A1 (en) * | 2015-03-12 | 2016-09-15 | Toyota Jidosha Kabushiki Kaisha | Detecting roadway objects in real-time images |
| US20170098127A1 (en) * | 2015-10-02 | 2017-04-06 | Seiko Epson Corporation | Measuring device, measuring system, measuring method, and program |
| CN107622226A (en) * | 2017-08-27 | 2018-01-23 | 南京理工大学 | Vehicle checking method and system based on improved deformable part model algorithm |
-
2018
- 2018-05-04 CN CN201810419768.0A patent/CN108647624A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102646199A (en) * | 2012-02-29 | 2012-08-22 | 湖北莲花山计算机视觉和信息科学研究院 | Motorcycle type identifying method in complex scene |
| CN104615998A (en) * | 2015-02-15 | 2015-05-13 | 武汉大学 | Vehicle search method based on multiple views |
| US20160267331A1 (en) * | 2015-03-12 | 2016-09-15 | Toyota Jidosha Kabushiki Kaisha | Detecting roadway objects in real-time images |
| US20170098127A1 (en) * | 2015-10-02 | 2017-04-06 | Seiko Epson Corporation | Measuring device, measuring system, measuring method, and program |
| CN107622226A (en) * | 2017-08-27 | 2018-01-23 | 南京理工大学 | Vehicle checking method and system based on improved deformable part model algorithm |
Non-Patent Citations (4)
| Title |
|---|
| CHAO WANG ET AL: "Probabilistic Inference for Occluded and Multiview On-road Vehicle Detection", 《IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS》 * |
| SAYANAN SIVARAMAN, ET AL: "Looking at Vehicles on the Road: A Survey of Vision-Based Vehicle Detection, Tracking, and Behavior Analysis", 《IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS》 * |
| 吴东承: "基于可变形部件模型的车辆检测方法研究", 《中国优秀硕士学位论文全文数据库信息科技辑（月刊）》 * |
| 陈远: "面向部件遮挡补偿的车辆检测模型", 《中国图象图形学报》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108875603A (en) | Intelligent driving control method and device, electronic equipment based on lane line | |
| CN105608417B (en) | Traffic lights detection method and device | |
| CN110136199A (en) | A kind of vehicle location based on camera, the method and apparatus for building figure | |
| CN103810717B (en) | A kind of human body behavioral value method and device | |
| CN103164711A (en) | Regional people stream density estimation method based on pixels and support vector machine (SVM) | |
| CN107431786A (en) | Image processing equipment, image processing system, image processing method and computer program | |
| CN102804231A (en) | Piecewise planar reconstruction of three-dimensional scenes | |
| CN112883820B (en) | Road target 3D detection method and system based on laser radar point cloud | |
| CN106251695A (en) | Destination's parking stall intelligent recommendation system and method based on parking space state monitoring | |
| CN102982341A (en) | Self-intended crowd density estimation method for camera capable of straddling | |
| EP3576007A1 (en) | Method for validation of obstacle candidate | |
| CN111191570B (en) | Image recognition method and device | |
| Hinz et al. | Car detection in aerial thermal images by local and global evidence accumulation | |
| CN106558051A (en) | A kind of improved method for detecting road from single image | |
| CN103593679A (en) | Visual human-hand tracking method based on online machine learning | |
| CN107330973A (en) | A kind of single-view method for reconstructing based on various visual angles supervision | |
| CN109948683A (en) | Difficulty division methods, device and its relevant device of point cloud data | |
| CN113034586B (en) | Road inclination angle detection method and detection system | |
| CN109740609A (en) | A kind of gauge detection method and device | |
| CN110378942A (en) | Barrier identification method, system, equipment and storage medium based on binocular camera | |
| CN104616317B (en) | Video vehicle tracking validity checking method | |
| Zhou et al. | Monoef: Extrinsic parameter free monocular 3d object detection | |
| CN109978919A (en) | A kind of vehicle positioning method and system based on monocular camera | |
| CN107610224A (en) | It is a kind of that algorithm is represented based on the Weakly supervised 3D automotive subjects class with clear and definite occlusion modeling | |
| CN112732860B (en) | Road extraction method, device, readable storage medium and equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181012 |