CN204347862U - A kind of vehicle vehicle-logo recognition system - Google Patents
A kind of vehicle vehicle-logo recognition system Download PDFInfo
- Publication number
- CN204347862U CN204347862U CN201420734567.7U CN201420734567U CN204347862U CN 204347862 U CN204347862 U CN 204347862U CN 201420734567 U CN201420734567 U CN 201420734567U CN 204347862 U CN204347862 U CN 204347862U
- Authority
- CN
- China
- Prior art keywords
- vehicle
- car
- logo
- image
- logo recognition
- 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.)
- Expired - Fee Related
Links
Landscapes
- Traffic Control Systems (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
Vehicle-logo recognition technology is supplemented as the gordian technique of vehicle recongnition technique, and be also that intelligent transportation field needs one of gordian technique solved, vehicle-logo recognition has important meaning for improving intelligent transportation system.At present, often there is the problems such as location is comparatively slow, accuracy is not high in the certain methods of vehicle-logo location.The utility model provides a kind of vehicle vehicle-logo recognition system, efficiently solve the existence due to radiator-grid around illumination, noise, inclination, deformation, stained, the partial occlusion in surface and car mark, affect vehicle car mark correctly to identify, thus affect the problem that vehicle distinguishes.
Description
Technical field
The present invention relates to a kind of recognition system, particularly relate to a kind of vehicle vehicle-logo recognition system.
Background technology
Since entering 21 century, along with the develop rapidly of China's economy, motor vehicles recoverable amount also grows with each passing day, and has caused the traffic problems be on the rise thus; Have also appeared the problem of other influences people normal life and property safety simultaneously, bring huge economic loss to the country and people.Therefore, become inevitable by computer information, intelligentized management vehicle.License plate recognition technology is widely used in traffic flow monitoring, and highway bayonet socket is charged, in the case detecting of red light violation vehicle monitoring and community automatic fare collection system and some public security systems.Vehicle-logo recognition technology is supplemented as the gordian technique of vehicle recongnition technique, is also one of gordian technique of intelligent transportation field.The identification of car target is divided into car mark coarse positioning substantially, car mark fine positioning, vehicle-logo recognition three steps, is utilize car plate and car target topological relation in the prior art, first determines car target approximate region, then extracts car mark accurately according to the rough position extracted; But, when positioning car mark, identifying, due to the factor that radiator-grid around illumination, noise, inclination, deformation, stained, the partial occlusion in surface and car mark exists, directly affect the accuracy of vehicle-logo location and vehicle-logo recognition.But car mark is a conspicuousness mark of vehicle, and it plays critical effect to the correct identification of vehicle.
For in prior art due to the existence of radiator-grid around illumination, noise, inclination, deformation, stained, the partial occlusion in surface and car mark, affect vehicle car mark and correctly identify, thus affect the problem that vehicle distinguishes, also not yet have effective solution at present.
Summary of the invention
For in prior art due to the existence of radiator-grid around illumination, noise, the stained and car mark in surface, affect vehicle car mark and correctly identify, thus affect the problem that vehicle distinguishes.The utility model provides a kind of vehicle vehicle-logo recognition system, to solve the problem.
A kind of vehicle vehicle-logo recognition system, includes: image collecting device, vehicle-logo location module, vehicle-logo recognition module and output module.
Described image collecting device adopts high-definition camera, is mainly used in the direct picture absorbing vehicle, for follow-up vehicle-logo recognition process provides vehicle essential information source clearly.
Described vehicle-logo location module includes coarse positioning submodule and fine positioning submodule, mainly completes and locates car target.
Coarse positioning submodule mainly completes vehicle car target coarse localization, and orient the rough region at car mark place, the process steps of coarse positioning is as follows.
The object of coarse positioning mainly determines the exact position of car plate and car light respectively by the first order difference technology of vehicle license location technique and image.First, need the image to image collecting device transmits to carry out color space conversion, by image by RGB color space conversion to HSV color space, and extract the image under V passage (i.e. luminance channel).
Further, binaryzation and the horizontal forward difference processing of single order are done to the image after extraction V passage.
Further, do horizontal gray-level projection to the image after the horizontal forward difference processing of single order, according to the position of the connected domain determination car light of drop shadow curve, the position of car light is thought in the maximum position of connected domain.
Further, because car mark is above car plate and between car light, car target rough region thus can be oriented;
The groundwork of fine positioning submodule carries out reprocessing to the car mark rough region of coarse positioning module location, and orient car target exact position, its positioning step is as follows.
First, fine positioning module car mark rough region image gray processing that coarse positioning module is oriented; And select suitable correction parameter, Gamma is done to car mark rough region image and corrects, obtain the image after Gamma correction; Gamma corrects the effect that can strengthen image, inhibits again the introducing of noise in image enhancement processes simultaneously.
Further, the Laws operator of definition horizontal direction and vertical direction, utilizes the Laws operator of definition to carry out filtering process to the car mark region coarse image after Gamma correction, obtains horizontal edge image and vertical edge image respectively; Then, vertical gray-level projection is done to the horizontal edge image obtained, according to the vertical gray-level projection curve of image, judge whether car mark is positioned on radiator-grid and the type of radiator-grid (horizontal texture radiator-grid, vertical texture radiator-grid, rhombus texture radiator-grid).
Further, car target position is judged.If car mark is not on radiator-grid or on the radiator-grid of horizontal texture, using the filtering image of the Laws operator of horizontal direction as application drawing picture; If car is marked on the radiator-grid of vertical texture or rhombus texture, the filtering process image of the Laws operator of horizontal direction and the filtering image of vertical direction Laws operator are done and computing, with the result of computing as application drawing picture.
Further, the structural element of selector disc shape, does closed operation to application drawing picture, and does vertical gray-level projection to the image after closed operation; Judge whether comprise enough points of interest in the image after closed operation according to the connected domain in vertical gray-level projection curve; If point of interest is less, then the operation that isolated point (non-region-of-interest) removes is done to image; If point of interest is enough, then opening operation is done to image, and pass through the border of gray-level projection curve determination region-of-interest, the part (non-region-of-interest) beyond region-of-interest is removed.
Further, dilation operation is done to the image removing non-region-of-interest, weaken opening operation and non-region-of-interest and remove the impact operating and car mark size is caused.
Further, car target border is determined.To car target width with highly all set a suitable threshold value, if car target size in the threshold range of setting, is then thought and located successfully;
Further, if car target size is not in the threshold range of setting, then applies Sobel operator and reorientate.The Sobel operator of definition horizontal direction and vertical direction, Sobel filtering process is done to the image after Gamma corrects, then filtered image is done and computing and closed operation, obtain closed operation image, the operation that isolated point (non-region-of-interest) removes is done to closed operation image, remake dilation operation, weaken isolated point and remove the impact operating and car mark size is caused; Finally determine car target border, accurate positioning car mark.
The groundwork of described vehicle-logo recognition module has been the car mark feature identification to the car logo image that vehicle-logo recognition module transmits, and identifying is as follows.
First vehicle-logo recognition module carries out the conversion of SIFT scale invariant feature to the car logo image (hereinafter referred to as " tested object ") that vehicle-logo location module transmits, and by extracting the position of the key point in car logo image, the descriptor of yardstick and rotational invariants obtains proper vector (i.e. SIFT feature).
Further, the car mark template base after carrying out the conversion of SIFT scale invariant feature is transferred.
Further, the proper vector in the proper vector extracted from tested object and car mark template base is carried out characteristic matching, using the tolerance of the Euclidean distance of the proper vector of key point as measurement two width image similarity; For certain key point in tested object, two nearest with it key points are found out in car mark template base, if the ratio of minimum distance and time minimum distance is less than certain threshold value, then think that key point in the car logo image key point nearest with car mark template base middle distance is mated; Otherwise, not think and mate.
Further, add up the quantity of the key point of each secondary car mark template base image and tested images match, maximum the thinking of match point quantity is mated most, draws matching result.
Described output module, exports the recognition result that vehicle-logo recognition module transmits.
By the utility model, efficiently solve the existence due to radiator-grid around illumination, noise, inclination, deformation, stained, the partial occlusion in surface and car mark, affect vehicle car mark and correctly identify, thus affect the problem that vehicle distinguishes.
Accompanying drawing explanation
Fig. 1 is the system architecture schematic diagram of the utility model vehicle vehicle-logo recognition system.
Fig. 2 is the utility model vehicle vehicle-logo recognition system cloud gray model process flow diagram.
Embodiment
For making the utility model easy to understand, below in conjunction with accompanying drawing, embodiment of the present utility model is further elaborated.
embodiment 1:as shown in Figure 1, known in conjunction with system architecture schematic diagram of the present utility model, identifying includes: image acquisition, vehicle-logo location, vehicle-logo recognition and result export, concrete implementation step and implement object as follows.
(1) image acquisition
Image collecting device adopts high-definition camera, is mainly used in the direct picture absorbing vehicle, for follow-up vehicle-logo recognition process provides vehicle essential information source clearly.
(2) vehicle-logo location
Vehicle-logo location has come primarily of vehicle-logo location module, and vehicle-logo location module includes coarse positioning submodule and fine positioning submodule.
Coarse positioning submodule mainly completes vehicle car target coarse localization, and orient the rough region at car mark place, the process steps of coarse positioning is as follows.
The object of coarse positioning mainly determines the exact position of car plate and car light respectively by the first order difference technology of vehicle license location technique and image.First, need the image to image collecting device transmits to carry out color space conversion, by image by RGB color space conversion to HSV color space, and extract the image under V passage (i.e. luminance channel).
Further, binaryzation and the horizontal forward difference processing of single order are done to the image after extraction V passage.
Further, do horizontal gray-level projection to the image after the horizontal forward difference processing of single order, according to the position of the connected domain determination car light of drop shadow curve, the position of car light is thought in the maximum position of connected domain.
Further, because car mark is above car plate and between car light, car target rough region thus can be oriented;
The groundwork of fine positioning submodule carries out reprocessing to the car mark rough region of coarse positioning module location, and orient car target exact position, its positioning step is as follows.
First, fine positioning module car mark rough region image gray processing that coarse positioning module is oriented; And select suitable correction parameter, Gamma is done to car mark rough region image and corrects, obtain the image after Gamma correction; Gamma corrects the effect that can strengthen image, inhibits again the introducing of noise in image enhancement processes simultaneously.
Further, the Laws operator of definition horizontal direction and vertical direction, utilizes the Laws operator of definition to carry out filtering process to the car mark region coarse image after Gamma correction, obtains horizontal edge image and vertical edge image respectively; Then, vertical gray-level projection is done to the horizontal edge image obtained, according to the vertical gray-level projection curve of image, judge whether car mark is positioned on radiator-grid and the type of radiator-grid (horizontal texture radiator-grid, vertical texture radiator-grid, rhombus texture radiator-grid).
Further, car target position is judged.If car mark is not on radiator-grid or on the radiator-grid of horizontal texture, using the filtering image of the Laws operator of horizontal direction as application drawing picture; If car is marked on the radiator-grid of vertical texture or rhombus texture, the filtering process image of the Laws operator of horizontal direction and the filtering image of vertical direction Laws operator are done and computing, with the result of computing as application drawing picture.
Further, the structural element of selector disc shape, does closed operation to application drawing picture, and does vertical gray-level projection to the image after closed operation; Judge whether comprise enough points of interest in the image after closed operation according to the connected domain in vertical gray-level projection curve; If point of interest is less, then the operation that isolated point (non-region-of-interest) removes is done to image; If point of interest is enough, then opening operation is done to image, and pass through the border of gray-level projection curve determination region-of-interest, the part (non-region-of-interest) beyond region-of-interest is removed.
Further, dilation operation is done to the image removing non-region-of-interest, weaken opening operation and non-region-of-interest and remove the impact operating and car mark size is caused.
Further, car target border is determined.To car target width with highly all set a suitable threshold value, if car target size in the threshold range of setting, is then thought and located successfully;
Further, if car target size is not in the threshold range of setting, then applies Sobel operator and reorientate.The Sobel operator of definition horizontal direction and vertical direction, Sobel filtering process is done to the image after Gamma corrects, then filtered image is done and computing and closed operation, obtain closed operation image, the operation that isolated point (non-region-of-interest) removes is done to closed operation image, do dilation operation again, weaken isolated point and remove the impact operating and car mark size is caused; Finally determine car target border, accurate positioning car mark.
(3) vehicle-logo recognition
Vehicle-logo recognition has come primarily of vehicle-logo recognition module, and realize the car mark feature identification to the car logo image that vehicle-logo recognition module transmits, identifying is as follows.
First vehicle-logo recognition module carries out the conversion of SIFT scale invariant feature to the car logo image (hereinafter referred to as " tested object ") that vehicle-logo location module transmits, and by extracting the position of the key point in car logo image, the descriptor of yardstick and rotational invariants obtains proper vector (i.e. SIFT feature).
Further, the car mark template base after carrying out the conversion of SIFT scale invariant feature is transferred.
Further, the proper vector in the proper vector extracted from tested object and car mark template base is carried out characteristic matching, using the tolerance of the Euclidean distance of the proper vector of key point as measurement two width image similarity; For certain key point in tested object, two nearest with it key points are found out in car mark template base, if the ratio of minimum distance and time minimum distance is less than certain threshold value, then think that key point in the car logo image key point nearest with car mark template base middle distance is mated; Otherwise, not think and mate.
Further, add up the quantity of the key point of each secondary car mark template base image and tested images match, maximum the thinking of match point quantity is mated most, draws matching result.
(4) result exports
Described output module, exports vehicle-logo recognition module and transmits recognition result.
Embodiment 2: as shown in Figure 2, the identification process following steps of the utility model vehicle vehicle-logo recognition system.
S201 system starts identification process, performs step S202.
S202 system completes image acquisition, performs step S203.
S203 system carries out car plate detection & localization, is car mark coarse positioning zoning, performs step S204.
S204 system carries out car mark coarse positioning, performs step S205 after completing coarse positioning.
S205 system carries out car mark fine positioning, performs step S206 after completing fine positioning.
S206 system judges that whether location is successful, locates successfully, then performs step S210; Otherwise, perform step S207.
S207 system changeover algorithm carries out secondary location, performs step S208.
S208 system judges that whether secondary location is successful, locates successfully, then performs step S210; Otherwise, perform step S209.
S209 system prompt None-identified, performs step S215.
S210 system carries out vehicle-logo recognition, performs step S211.
Proper vector in car target proper vector to be identified and car mark template base is carried out characteristic matching by S211 system, performs step S212.
S212 system judges whether car target proper vector to be identified mates with the proper vector in car mark template base, and the match is successful, performs step S213; Otherwise, perform step S214.
S213 system exports recognition result, performs step S215.
S214 system prompt coupling is unsuccessful, performs step S215.
S215 system identification process terminates.
Claims (1)
1. a vehicle vehicle-logo recognition system, is characterized in that: system includes: image collecting device, vehicle-logo location module, vehicle-logo recognition module and output module;
Described image collecting device adopts high-definition camera, is mainly used in the direct picture absorbing vehicle, for follow-up vehicle-logo recognition process provides vehicle essential information source clearly;
Described vehicle-logo location module includes coarse positioning submodule and fine positioning submodule, mainly completes and locates car target; Wherein, coarse positioning submodule mainly completes vehicle car target coarse localization, orients the rough region at car mark place; The groundwork of fine positioning submodule carries out reprocessing to the car mark rough region of coarse positioning module location, orients car target exact position;
The groundwork of described vehicle-logo recognition module has been the car mark feature identification to the car logo image that vehicle-logo recognition module transmits;
Described output module, exports the recognition result that vehicle-logo recognition module transmits.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201420734567.7U CN204347862U (en) | 2014-12-01 | 2014-12-01 | A kind of vehicle vehicle-logo recognition system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201420734567.7U CN204347862U (en) | 2014-12-01 | 2014-12-01 | A kind of vehicle vehicle-logo recognition system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN204347862U true CN204347862U (en) | 2015-05-20 |
Family
ID=53231119
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201420734567.7U Expired - Fee Related CN204347862U (en) | 2014-12-01 | 2014-12-01 | A kind of vehicle vehicle-logo recognition system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN204347862U (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105426896A (en) * | 2015-11-16 | 2016-03-23 | 成都神州数码索贝科技有限公司 | Car logo automatic identification method and system |
| CN108614985A (en) * | 2016-12-12 | 2018-10-02 | 广西师范大学 | A kind of vehicle-logo location method |
-
2014
- 2014-12-01 CN CN201420734567.7U patent/CN204347862U/en not_active Expired - Fee Related
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105426896A (en) * | 2015-11-16 | 2016-03-23 | 成都神州数码索贝科技有限公司 | Car logo automatic identification method and system |
| CN108614985A (en) * | 2016-12-12 | 2018-10-02 | 广西师范大学 | A kind of vehicle-logo location method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Luvizon et al. | A video-based system for vehicle speed measurement in urban roadways | |
| CN103136528B (en) | A kind of licence plate recognition method based on dual edge detection | |
| Greenhalgh et al. | Recognizing text-based traffic signs | |
| CN103500322A (en) | Automatic lane line identification method based on low-altitude aerial images | |
| CN102902957A (en) | Video-stream-based automatic license plate recognition method | |
| CN102880863B (en) | Method for positioning license number and face of driver on basis of deformable part model | |
| CN103679146A (en) | Safety belt intelligent detection method based on high-pass filter and Hough conversion | |
| CN103310231A (en) | Auto logo locating and identifying method | |
| CN102799859A (en) | Method for identifying traffic sign | |
| CN103268489A (en) | Motor vehicle plate identification method based on sliding window searching | |
| Soomro et al. | Vehicle number recognition system for automatic toll tax collection | |
| CN104537360A (en) | Method and system for detecting vehicle violation of not giving way | |
| CN104102909A (en) | Vehicle characteristic positioning and matching method based on multiple-visual information | |
| CN107358236A (en) | A kind of number-plate number identifying system and method based on camera device | |
| CN104978746A (en) | Running vehicle body color identification method | |
| CN103065144A (en) | Automobile logo identification method and device | |
| CN105787475A (en) | Traffic sign detection and identification method under complex environment | |
| CN204347862U (en) | A kind of vehicle vehicle-logo recognition system | |
| CN105069419A (en) | Traffic sign detection method based on edge color pair and characteristic filters | |
| CN104408431A (en) | Vehicle model identification method under traffic monitoring | |
| Lee et al. | Vehicle model recognition in video | |
| Moizumi et al. | Traffic light detection considering color saturation using in-vehicle stereo camera | |
| Goel et al. | Vehicle registration plate recognition system using template matching | |
| CN109753880B (en) | Detection and identification method for natural scene vehicle-mounted video road speed limit sign | |
| Huang | Research on license plate image segmentation and intelligent character recognition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20150520 Termination date: 20161201 |