CN106886757A - A kind of multiclass traffic lights detection method and system based on prior probability image - Google Patents

A kind of multiclass traffic lights detection method and system based on prior probability image Download PDF

Info

Publication number
CN106886757A
CN106886757A CN201710043802.4A CN201710043802A CN106886757A CN 106886757 A CN106886757 A CN 106886757A CN 201710043802 A CN201710043802 A CN 201710043802A CN 106886757 A CN106886757 A CN 106886757A
Authority
CN
China
Prior art keywords
traffic lights
image
sliding window
feature
positive
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
Application number
CN201710043802.4A
Other languages
Chinese (zh)
Other versions
CN106886757B (en
Inventor
陶文兵
梁福禄
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huazhong University of Science and Technology
Original Assignee
Huazhong University of Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Huazhong University of Science and Technology filed Critical Huazhong University of Science and Technology
Priority to CN201710043802.4A priority Critical patent/CN106886757B/en
Publication of CN106886757A publication Critical patent/CN106886757A/en
Application granted granted Critical
Publication of CN106886757B publication Critical patent/CN106886757B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/2431—Multiple classes

Abstract

The invention discloses a kind of multiclass traffic lights detection method and system based on prior probability image, belong to technical field of computer vision.The inventive method extracts the integrating channel feature of the positive and negative sample image of traffic lights first, one traffic lights detection model is trained using Adaboost learning algorithms, the regularity of distribution of traffic lights is recycled to build prior probability image, self adaptation setting is carried out to detection threshold value, traffic lights positioning is carried out to target image by the way of soft cascade, Classification and Identification finally is carried out to traffic lights again, the traffic lights sample first to different shape and state before identification carries out category label, trains a multi classifier；Type identification is carried out to traffic lights image-region again.The invention also discloses a kind of multiclass traffic lights detecting system based on prior probability image, technical solution of the present invention by great amount of samples training, it is not easy to influenceed by noise and illumination variation, it is adaptable to complicated urban transportation scene.

Description

A kind of multiclass traffic lights detection method and system based on prior probability image
Technical field
The invention belongs to technical field of computer vision, handed over more particularly, to a kind of multiclass based on prior probability image Logical lamp inspection surveys method and system.
Background technology
With increasing for automobile, traffic problems are increasingly highlighted.In order to solve traffic congestion, accident rate is reduced, it is more next More research institutions have been put into the research of intelligent transportation system, and the hub device that traffic lights is operated as traffic, it is A wherein essential part.If energy is timely, traffic light signals are automatically captured, either to unmanned or intelligence Can aid in driving, there is very important meaning.
Traffic lights detecting system generally uses junction service, global positioning system (Global Positioning System, GPS), the mode such as computer vision.Wherein, it is easily installed due to it using the detection method of computer vision, cost Cheap, effect is directly perceived, and is taken seriously the features such as need not increase extra infrastructure.The method on automobile by installing one Individual video camera, gathers the image of vehicle front road scene, and traffic lights is whether there is in analysis image, and to its position and type Judged.
The traffic lights detection method for being currently based on computer vision substantially uses didactic method, is generally divided into Following steps：The candidate region of traffic lights is obtained first, and image is split in specific color space typically, Or the speck region for being obtained using top cap algorithm in image；Secondly template matches or connected domain are adopted to the candidate region for obtaining The modes such as analysis are filtered, and obtain final traffic lights testing result.
Although above method speed ratio is very fast, because its detection mode is the rule that is manually set using some, should Scope is generally relatively more limited.Traffic light kind is numerous, it is difficult to design one can be accurate to all types of traffic lights The method of detection, so most of existing method can be only applied to the traffic lights of circle.And the above method is to illumination variation And the poor robustness of noise, due to other lamps in the blocking of traffic lights, illumination variation, halation, flicker and complex scene The presence of the problems such as interference, the model being calculated often may be only available for certain specific data set, it is impossible to be applied to complexity Real scene in.
The content of the invention
For the disadvantages described above or Improvement requirement of prior art, the invention provides a kind of multiclass based on prior probability image Traffic lights detection method and system, its object is to using the traffic lights detection method based on study, to different in collecting sample The traffic lights sample of shape is marked, and a multi classifier is trained using multi classifier, recycles the distribution of traffic lights Rule builds prior probability image, and classification thresholds are carried out with self adaptation setting, and using integrating channel feature as traffic lights mould The feature representation of type, thus solves often to may be only available for certain given shape traffic lights and anti-interference energy in prior art The technical problem of power.
To achieve the above object, according to one aspect of the present invention, there is provided a kind of multiclass based on prior probability image is handed over Logical lamp detection method, the method is comprised the following steps：
(1) multiple dimensioned scaling is entered to testing image according to the ratio for setting；Preferably, Image scaling coefficients ground floor is 1, each layer is the 0.95 of last layer afterwards, and 20 layers are scaled altogether；
(2) to scaling after all image zooming-out integrating channel features, formed feature pyramid；
(3) line slip is entered in the pyramidal each layer of feature using a window for fixed size, and utilizes traffic lamp inspection Survey model to detect the integrating channel feature in sliding window, obtain the sliding window position comprising traffic lights and window Traffic lights probability score；
(4) overlapping area of the statistics comprising traffic lights and overlapped sliding window, overlapping area exceedes the weight of setting Folded threshold value then chooses traffic lights probability score highest sliding window as window to be sorted；Preferably, anti-eclipse threshold is 0.5；
(5) window to be sorted is input to traffic lights disaggregated model, obtains the type and state of traffic lights and export.
Further, the training of traffic lights detection model specifically includes following steps in the step (3)：
(21) road scene image is gathered, all traffic light kinds that need to be detected is included in the sample image for collecting And state, and the number of different type and state traffic lights is evenly distributed；
(22) handmarking is carried out to image pattern, region of the cutting comprising complete traffic lights is simultaneously labeled as positive sample, with Machine is chosen not comprising traffic lights region as negative sample, by the normalization of positive and negative sample-size；
(23) the feature passage of positive negative sample is chosen, the characteristics of image of positive and negative sample characteristics passage is extracted afterwards, it is finally right The characteristics of image for obtaining enters the integrating channel feature that row set obtains positive negative sample；
(24) using the integrating channel feature of the positive negative sample of traffic lights, one Adaboost grader of training is used as traffic lights Detection model.
Further, the integrating channel feature in sliding window is detected in the step (3) and is specifically used soft level Connection Adaboost, Adaboost graders are made up of multiple Weak Classifiers, such as following formula
Wherein, x be sliding window in be extracted integrating channel characteristic vector；hi(x) be i-th Weak Classifier to feature to Measure the output of x；Since first Weak Classifier, compare first Weak Classifier and self-adapting detecting threshold value, if less than adaptive Detection threshold value is answered, then assert that in sliding window be background, continued sliding window and detected next time；If being more than self-adapting detecting Threshold value is then added up latter Weak Classifier, and accumulation result is continued and self-adapting detecting threshold comparison, if being examined less than self adaptation Threshold value is surveyed, then assert that sliding window is background, continued sliding window and detected next time；If more than self-adapting detecting threshold value Continue the latter Weak Classifier that add up, until by last Weak Classifier, then assert that the sliding window includes traffic lights mesh Mark, output accumulation result is used as traffic lights probability score；Output sliding window position.
Further, the asking for of described self-adapting detecting threshold value comprises the following steps：
(41) the traffic lights location-prior probability graph and traffic lights size prior probability image of all positive samples are calculated；
(42) location-prior probability graph I is calculatedpIntegrogram
Wherein, the value of each point of integrogram is equivalent to adding up in this upper left corner all values in artwork；By right Integrogram carries out simple arithmetical operation, obtains
Wherein, x and y represent the coordinate in the detection window upper left corner, and w and h represents the wide and height of detection window respectively；
(43) by size prior probability image Is, calculate candidate area size and the region occur traffic lights desired size it Between deviation Ps(x,y,w,h)
(44) self-adapting detecting threshold value is
T (x, y, w, h)=- Ps(x,y,w,h)·Pp(x,y,w,h)。
Further, the training of traffic lights disaggregated model is specifically divided into following steps in the step (5)：
(51) traffic lights to different shape different conditions in positive sample carries out manual sort's mark；
(52) the feature passage of positive sample is chosen, the characteristics of image of positive sample feature passage is extracted afterwards, finally to obtaining Characteristics of image enter the integrating channel feature that row set obtains positive negative sample；
(53) using positive sample of classifying, it is trained using multi classifier, obtains traffic lights disaggregated model；Preferably, Multi classifier uses Adaboost.MH algorithms.
It is another aspect of this invention to provide that there is provided a kind of multiclass traffic lights detecting system based on prior probability image, should System includes following part：
Testing image Zoom module, for entering multiple dimensioned scaling to testing image according to the ratio for setting；Preferably, scheme As zoom factor ground floor is 1, each layer is the 0.95 of last layer afterwards, and 20 layers are scaled altogether；
Integrating channel characteristic extracting module, for all image zooming-out integrating channel features after to scaling, forms feature Pyramid；
Detection module is slided, line slip is entered in the pyramidal each layer of feature for the window using a fixed size, And the integrating channel feature in sliding window is detected using traffic lights detection model, obtain the sliding window comprising traffic lights Mouth position and the traffic lights probability score of window；
Non-maxima suppression module, for counting the overlapping area comprising traffic lights and overlapped sliding window, weight The anti-eclipse threshold that folded area exceedes setting then chooses traffic lights probability score highest sliding window as window to be sorted；It is preferred that , anti-eclipse threshold is 0.5；
Sort module, for window to be sorted to be input into traffic lights disaggregated model, obtains the type and state of traffic lights And export.
Further, the training system for sliding traffic lights detection model in detection module is specifically included with lower unit：
Sampling unit, it is all comprising what need to be detected in the sample image for collecting for gathering road scene image Traffic light kind and state, and the number of different type and state traffic lights is evenly distributed；
Positive and negative sample labeling unit, for carrying out handmarking, region of the cutting comprising complete traffic lights to image pattern And labeled as positive sample, randomly select not comprising traffic lights region as negative sample, by the normalization of positive and negative sample-size；
Integrating channel feature extraction unit, the feature passage for choosing positive negative sample, extracts positive and negative sample characteristics afterwards The characteristics of image of passage, finally enters the integrating channel feature that row set obtains positive negative sample to the characteristics of image for obtaining；
Traffic lights detection model training unit, for the integrating channel feature using the positive negative sample of traffic lights, trains one Adaboost graders are used as traffic lights detection model.
Further, the integrating channel feature in sliding window is carried out in the slip detection module detecting specific use Soft cascade Adaboost, Adaboost grader is made up of multiple Weak Classifiers, such as following formula
Wherein, x be sliding window in be extracted integrating channel characteristic vector；hi(x) be i-th Weak Classifier to feature to Measure the output of x；Since first Weak Classifier, compare first Weak Classifier and self-adapting detecting threshold value, if less than adaptive Detection threshold value is answered, then assert that in sliding window be background, continued sliding window and detected next time；If being more than self-adapting detecting Threshold value is then added up latter Weak Classifier, and accumulation result is continued and self-adapting detecting threshold comparison, if being examined less than self adaptation Threshold value is surveyed, then assert that sliding window is background, continued sliding window and detected next time；If more than self-adapting detecting threshold value Continue the latter Weak Classifier that add up, until by last Weak Classifier, then assert that the sliding window includes traffic lights mesh Mark, output accumulation result is used as traffic lights probability score；Output sliding window position.
Further, the system of asking for of described self-adapting detecting threshold value is included with lower unit：
Prior probability image computing unit, traffic lights location-prior probability graph and traffic lights chi for calculating all positive samples Very little prior probability image；
Integrogram computing unit, for calculating location-prior probability graph IpIntegrogram
Wherein, the value of each point of integrogram is equivalent to adding up in this upper left corner all values in artwork；By right Integrogram carries out simple arithmetical operation, obtains
Wherein, x and y represent the coordinate in the detection window upper left corner, and w and h represents the wide and height of detection window respectively；
Deviation computing unit, for by size prior probability image Is, calculate candidate area size and traffic occur with the region Deviation P between the desired size of lamps(x,y,w,h)
Self-adapting detecting threshold computation unit, for calculating self-adapting detecting threshold value
T (x, y, w, h)=- Ps(x,y,w,h)·Pp(x,y,w,h)。
Further, the training system of traffic lights disaggregated model is specifically divided into lower unit in the sort module：
Key words sorting unit, manual sort's mark is carried out for the traffic lights to different shape different conditions in positive sample；
Classification integrating channel feature extraction unit, the feature passage for choosing positive sample extracts positive sample feature afterwards The characteristics of image of passage, finally enters the integrating channel feature that row set obtains positive negative sample to the characteristics of image for obtaining；
Traffic lights disaggregated model training unit, for using positive sample of classifying, being trained using multi classifier, is obtained Traffic lights disaggregated model, it is preferred that multi classifier uses Adaboost.MH algorithms.
In general, by the contemplated above technical scheme of the present invention compared with prior art, it is special with following technology Levy and beneficial effect：
(1) using the traffic lights detection method based on study, traffic lights sample of different shapes in collecting sample is carried out Mark, a multi classifier is trained using multi classifier, realizes the identification to different type traffic lights, it can be determined that inspection The shape and state of the traffic lights for measuring；
(2) present invention uses integrating channel feature as the feature representation of traffic lights model.It is special with simple use color Levy or Gradient Features carry out target detection comparatively, integrating channel feature is extracted various different types of features again first Merged, so as to the multivariate table for foring target reaches, enhanced the robustness of feature；
(3) by the way of adaptive threshold, with reference to the characteristic distributions of traffic lights, using traffic lights probability of occurrence and size And the threshold value of the relation setting detection between picture position, improve the precision of detection.
Brief description of the drawings
Fig. 1 overall flow figures of the present invention；
Image channel schematic diagram in Fig. 2 present invention；
Position prior probability schematic diagram in Fig. 3 a present invention；
Size prior probability schematic diagram in Fig. 3 b present invention；
Traffic lights positioning function schematic diagram in Fig. 4 a, 4b, 4c and 4d present invention.
Specific embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, it is right below in conjunction with drawings and Examples The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and It is not used in the restriction present invention.As long as additionally, technical characteristic involved in invention described below each implementation method Not constituting conflict each other can just be mutually combined.
It is as shown in Figure 1 the implementing procedure of the embodiment of the present invention, specifically includes following steps：
(1) multiple dimensioned scaling is entered to testing image according to the ratio for setting；Zoom factor ground floor is 1 in this implementation, Each layer is the 0.95 of last layer afterwards, and 20 layers are scaled altogether；
(2) to scaling after all image zooming-out integrating channel features, formed feature pyramid；Image after scaling is carried Integrating channel feature is taken, feature pyramid is formed；
Passage is the output response of original image, the certain type of feature that expression is extracted；For coloured image, Each Color Channel then corresponds to a passage, and other similar passages can carry out various linear processes conversion to original image Obtain；The color and gradient feature of this Cass collection traffic lights, constructs ten feature passages, as shown in Fig. 2 wherein including Tri- Color Channels of LUV, a gradient intensity passage and gradient projection are at 0 degree to equally distributed six direction between 180 degree The six gradient direction passages for being constituted；On feature passage extract Haar features, by the Haar features of all feature passages to Amount couples together the integrating channel feature for obtaining sample.
(3) line slip is entered in the pyramidal each layer of feature using a window for fixed size, and utilizes traffic lamp inspection Survey model to detect the integrating channel feature in sliding window, obtain the sliding window position comprising traffic lights and window Traffic lights probability score；
The training of wherein traffic lights detection model specifically includes following steps：
(21) road scene image is gathered, all traffic light kinds that need to be detected is included in the sample image for collecting And state, and the number of different type and state traffic lights is evenly distributed；
(22) handmarking is carried out to image pattern, region of the cutting comprising complete traffic lights is simultaneously labeled as positive sample, with Machine is chosen not comprising traffic lights region as negative sample, by the normalization of positive and negative sample-size；
(23) the feature passage of positive negative sample is chosen, the characteristics of image of positive and negative sample characteristics passage is extracted afterwards, it is finally right The characteristics of image for obtaining enters the integrating channel feature that row set obtains positive negative sample；
(24) using the integrating channel feature of the positive negative sample of traffic lights, one Adaboost grader of training is used as traffic lights Detection model.
In order to accelerate the speed of detection, present invention employs soft cascade Adaboost, Adaboost graders are by multiple Weak Classifier composition
Wherein, x be sliding window in be extracted integrating channel characteristic vector；hi(x) be i-th Weak Classifier to feature to Measure the output of x；Since first Weak Classifier, compare first Weak Classifier and self-adapting detecting threshold value, if less than adaptive Detection threshold value is answered, then assert that in sliding window be background, continued sliding window and detected next time；If being more than self-adapting detecting Threshold value is then added up latter Weak Classifier, and accumulation result is continued and self-adapting detecting threshold comparison, if being examined less than self adaptation Threshold value is surveyed, then assert that sliding window is background, continued sliding window and detected next time；If more than self-adapting detecting threshold value Continue the latter Weak Classifier that add up, until by last Weak Classifier, then assert that the sliding window includes traffic lights mesh Mark, output accumulation result is used as traffic lights probability score；Output sliding window position.
Wherein self-adapting detecting threshold value ask for comprise the following steps：
(41) the traffic lights location-prior probability graph and traffic lights size prior probability image of all positive samples are calculated；
(42) location-prior probability graph I is calculatedpIntegrogram
Wherein, the value of each point of integrogram is equivalent to adding up in this upper left corner all values in artwork；By right Integrogram carries out simple arithmetical operation, obtains
Wherein, x and y represent the coordinate in the detection window upper left corner, and w and h represents the wide and height of detection window respectively；
(43) by size prior probability image Is, calculate candidate area size and the region occur traffic lights desired size it Between deviation Ps(x,y,w,h)
(44) self-adapting detecting threshold value is
T (x, y, w, h)=- Ps(x,y,w,h)·Pp(x,y,w,h)。
(4) overlapping area of the statistics comprising traffic lights and overlapped sliding window, overlapping area exceedes the weight of setting Folded threshold value then chooses traffic lights probability score highest sliding window as window to be sorted；The present embodiment anti-eclipse threshold is 0.5.
(5) window to be sorted is input to traffic lights disaggregated model, obtains the type and state of traffic lights and export；
The training of wherein traffic lights disaggregated model is comprised the following steps：
(51) traffic lights to different shape different conditions in positive sample carries out manual sort's mark；
(52) the feature passage of positive sample is chosen, the characteristics of image of positive sample feature passage is extracted afterwards, finally to obtaining Characteristics of image enter the integrating channel feature that row set obtains positive negative sample；
(53) using positive sample of classifying, it is trained using multi classifier, obtains traffic lights disaggregated model；The present embodiment Multi classifier uses Adaboost.MH algorithms.
Presently preferred embodiments of the present invention is the foregoing is only, is not intended to limit the invention, it is all in essence of the invention Any modification, equivalent and improvement made within god and principle etc., should be included within the scope of the present invention.

Claims (10)

1. a kind of multiclass traffic lights detection method based on prior probability image, it is characterised in that the method is comprised the following steps：
(1) multiple dimensioned scaling is entered to testing image according to the ratio for setting；
(2) to scaling after all image zooming-out integrating channel features, formed feature pyramid；
(3) line slip is entered in the pyramidal each layer of feature using a window for fixed size, and mould is detected using traffic lights Type detects to the integrating channel feature in sliding window, obtains the traffic of the sliding window position comprising traffic lights and window Lamp probability score；
(4) overlapping area of the statistics comprising traffic lights and overlapped sliding window, overlapping area exceedes the overlap threshold of setting Value then chooses traffic lights probability score highest sliding window as window to be sorted；
(5) window to be sorted is input to traffic lights disaggregated model, obtains the type and state of traffic lights and export.
2. a kind of multiclass traffic lights detection method based on prior probability image according to claim 1, it is characterised in that institute The training for stating traffic lights detection model in step (3) specifically includes following steps：
(21) road scene image is gathered, all traffic light kinds and shape that need to be detected is included in the sample image for collecting State, and the number of different type and state traffic lights is evenly distributed；
(22) handmarking is carried out to image pattern, region of the cutting comprising complete traffic lights is simultaneously labeled as positive sample, random choosing Take not comprising traffic lights region as negative sample, by the normalization of positive and negative sample-size；
(23) the feature passage of positive negative sample is chosen, the characteristics of image of positive and negative sample characteristics passage is extracted afterwards, finally to obtaining Characteristics of image enter the integrating channel feature that row set obtains positive negative sample；
(24) using the integrating channel feature of the positive negative sample of traffic lights, one Adaboost grader of training is detected as traffic lights Model.
3. a kind of multiclass traffic lights detection method based on prior probability image according to claim 1, it is characterised in that institute To state in step (3) carry out the integrating channel feature in sliding window and detect specific using soft cascade Adaboost, Adaboost Grader is made up of multiple Weak Classifiers, such as following formula
C ( x ) = C k ( x ) = Σ i = 1 k h i ( x ) ,
Wherein, x be sliding window in be extracted integrating channel characteristic vector；hiX () is i-th Weak Classifier to characteristic vector x Output；Since first Weak Classifier, compare first Weak Classifier and self-adapting detecting threshold value, if being less than self-adapting detecting Threshold value, then assert that in sliding window be background, continues sliding window and is detected next time；If more than self-adapting detecting threshold value Added up latter Weak Classifier, and accumulation result is continued and self-adapting detecting threshold comparison, if being less than self-adapting detecting threshold value, Then assert that sliding window is background, continue sliding window and detected next time；Continue to tire out if more than self-adapting detecting threshold value Plus latter Weak Classifier, until by last Weak Classifier, then assert that the sliding window includes traffic lights target, output Accumulation result is used as traffic lights probability score；Output sliding window position.
4. a kind of multiclass traffic lights detection method based on prior probability image according to claim 1,2 or 3, its feature exists In, described self-adapting detecting threshold value ask for comprise the following steps：
(41) the traffic lights location-prior probability graph and traffic lights size prior probability image of all positive samples are calculated；
(42) location-prior probability graph I is calculatedpIntegrogram
ii p ( x , y ) = Σ x ′ ≤ x , y ′ ≤ y I p ( x ′ , y ′ ) ,
Wherein, the value of each point of integrogram is equivalent to adding up in this upper left corner all values in artwork；By to integration Figure carries out simple arithmetical operation, obtains
P p ( x , y , w , h ) = ii p ( x + w , y + h ) + ii p ( x , y ) - ii p ( x + w , y ) - ii p ( x , y + h ) w · h
Wherein, x and y represent the coordinate in the detection window upper left corner, and w and h represents the wide and height of detection window respectively；
(43) by size prior probability image Is, calculate candidate area size and the region and occur between the desired size of traffic lights Deviation Ps(x,y,w,h)
P s ( x , y , w , h ) = E x p ( - | | w h - I s ( x , y ) | | I s ( x , y ) ) ,
(44) self-adapting detecting threshold value is
T (x, y, w, h)=- Ps(x,y,w,h)·Pp(x,y,w,h)。
5. a kind of multiclass traffic lights detection method based on prior probability image according to claim 1 and 2, its feature exists In the training of traffic lights disaggregated model is specifically divided into following steps in the step (5)：
(51) traffic lights to different shape different conditions in positive sample carries out manual sort's mark；
(52) the feature passage of positive sample is chosen, the characteristics of image of positive sample feature passage is extracted afterwards, finally the figure to obtaining As feature enters the integrating channel feature that row set obtains positive negative sample；
(53) using positive sample of classifying, it is trained using multi classifier, obtains traffic lights disaggregated model.
6. a kind of multiclass traffic lights detecting system based on prior probability image, it is characterised in that the system is included with lower module：
Testing image Zoom module, for entering multiple dimensioned scaling to testing image according to the ratio for setting；
Integrating channel characteristic extracting module, for all image zooming-out integrating channel features after to scaling, forms feature gold word Tower；
Detection module is slided, for entering line slip, and profit in the pyramidal each layer of feature using a window for fixed size The integrating channel feature in sliding window is detected with traffic lights detection model, obtains the sliding window position comprising traffic lights Put the traffic lights probability score with window；
Non-maxima suppression module, for counting the overlapping area comprising traffic lights and overlapped sliding window, faying surface The anti-eclipse threshold that product exceedes setting then chooses traffic lights probability score highest sliding window as window to be sorted；
Sort module, for window to be sorted to be input into traffic lights disaggregated model, obtains the type and state of traffic lights and defeated Go out.
7. a kind of multiclass traffic lights detecting system based on prior probability image according to claim 6, it is characterised in that institute The training system for stating traffic lights detection model in slip detection module is specifically included with lower unit：
Sampling unit, for gathering road scene image, will include all traffic that need to be detected in the sample image for collecting Lamp type and state, and the number of different type and state traffic lights is evenly distributed；
Positive and negative sample labeling unit, for carrying out handmarking to image pattern, region of the cutting comprising complete traffic lights is simultaneously marked Positive sample is designated as, is randomly selected not comprising traffic lights region as negative sample, by the normalization of positive and negative sample-size；
Integrating channel feature extraction unit, the feature passage for choosing positive negative sample, extracts positive and negative sample characteristics passage afterwards Characteristics of image, the integrating channel feature that row set obtains positive negative sample is finally entered to the characteristics of image for obtaining；
Traffic lights detection model training unit, for the integrating channel feature using the positive negative sample of traffic lights, trains one Adaboost graders are used as traffic lights detection model.
8. a kind of multiclass traffic lights detecting system based on prior probability image according to claim 6, it is characterised in that institute State the integrating channel feature in sliding window detect in slip detection module it is specific use soft cascade Adaboost, Adaboost graders are made up of multiple Weak Classifiers, such as following formula
C ( x ) = C k ( x ) = Σ i = 1 k h i ( x ) ,
Wherein, x be sliding window in be extracted integrating channel characteristic vector；hiX () is i-th Weak Classifier to characteristic vector x Output；Since first Weak Classifier, compare first Weak Classifier and self-adapting detecting threshold value, if being less than self-adapting detecting Threshold value, then assert that in sliding window be background, continues sliding window and is detected next time；If more than self-adapting detecting threshold value Added up latter Weak Classifier, and accumulation result is continued and self-adapting detecting threshold comparison, if being less than self-adapting detecting threshold value, Then assert that sliding window is background, continue sliding window and detected next time；Continue to tire out if more than self-adapting detecting threshold value Plus latter Weak Classifier, until by last Weak Classifier, then assert that the sliding window includes traffic lights target, output Accumulation result is used as traffic lights probability score；Output sliding window position.
9. a kind of multiclass traffic lights detecting system based on prior probability image according to claim 6,7 or 8, its feature exists In the system of asking for of described self-adapting detecting threshold value is included with lower unit：
Prior probability image computing unit, traffic lights location-prior probability graph and traffic lights size elder generation for calculating all positive samples Test probability graph；
Integrogram computing unit, for calculating location-prior probability graph IpIntegrogram
ii p ( x , y ) = Σ x ′ ≤ x , y ′ ≤ y I p ( x ′ , y ′ ) ,
Wherein, the value of each point of integrogram is equivalent to adding up in this upper left corner all values in artwork；By to integration Figure carries out simple arithmetical operation, obtains
P p ( x , y , w , h ) = ii p ( x + w , y + h ) + ii p ( x , y ) - ii p ( x + w , y ) - ii p ( x , y + h ) w · h
Wherein, x and y represent the coordinate in the detection window upper left corner, and w and h represents the wide and height of detection window respectively；
Deviation computing unit, for by size prior probability image Is, calculate candidate area size and the phase of traffic lights occur with the region The deviation P hoped between sizes(x,y,w,h)
P s ( x , y , w , h ) = E x p ( - | | w h - I s ( x , y ) | | I s ( x , y ) ) ,
Self-adapting detecting threshold computation unit, for calculating self-adapting detecting threshold value
T (x, y, w, h)=- Ps(x,y,w,h)·Pp(x,y,w,h)。
10. a kind of multiclass traffic lights detecting system based on prior probability image according to claim 6 or 7, its feature exists In the training system of traffic lights disaggregated model is specifically divided into lower unit in the sort module：
Key words sorting unit, manual sort's mark is carried out for the traffic lights to different shape different conditions in positive sample；
Classification integrating channel feature extraction unit, the feature passage for choosing positive sample extracts positive sample feature passage afterwards Characteristics of image, the integrating channel feature that row set obtains positive negative sample is finally entered to the characteristics of image for obtaining；
Traffic lights disaggregated model training unit, for using positive sample of classifying, being trained using multi classifier, obtains traffic Lamp disaggregated model.
CN201710043802.4A 2017-01-19 2017-01-19 A kind of multiclass traffic lights detection method and system based on prior probability image Active CN106886757B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710043802.4A CN106886757B (en) 2017-01-19 2017-01-19 A kind of multiclass traffic lights detection method and system based on prior probability image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710043802.4A CN106886757B (en) 2017-01-19 2017-01-19 A kind of multiclass traffic lights detection method and system based on prior probability image

Publications (2)

Publication Number Publication Date
CN106886757A true CN106886757A (en) 2017-06-23
CN106886757B CN106886757B (en) 2019-06-11

Family

ID=59176706

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710043802.4A Active CN106886757B (en) 2017-01-19 2017-01-19 A kind of multiclass traffic lights detection method and system based on prior probability image

Country Status (1)

Country Link
CN (1) CN106886757B (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108009554A (en) * 2017-12-01 2018-05-08 国信优易数据有限公司 A kind of image processing method and device
CN109646796A (en) * 2019-01-17 2019-04-19 浙江大学 Channel wireless radio multi closed loop stimulation system for epilepsy therapy
CN110211650A (en) * 2019-05-30 2019-09-06 苏州爱医斯坦智能科技有限公司 Operation Rich Media's electronic health record monitors the method and device of identification record automatically
CN112633313A (en) * 2020-10-13 2021-04-09 北京匠数科技有限公司 Bad information identification method of network terminal and local area network terminal equipment
CN113156935A (en) * 2020-01-07 2021-07-23 动态Ad有限责任公司 System and method for traffic light detection

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103886308A (en) * 2014-04-15 2014-06-25 中南大学 Pedestrian detection method through soft cascade classifiers according to polymerization channel characteristics
CN105046197A (en) * 2015-06-11 2015-11-11 西安电子科技大学 Multi-template pedestrian detection method based on cluster
CN105184278A (en) * 2015-09-30 2015-12-23 深圳市商汤科技有限公司 Human face detection method and device
CN105550675A (en) * 2016-02-02 2016-05-04 天津大学 Binocular pedestrian detection method based on optimization polymerization integration channel
CN105760858A (en) * 2016-03-21 2016-07-13 东南大学 Pedestrian detection method and apparatus based on Haar-like intermediate layer filtering features

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103886308A (en) * 2014-04-15 2014-06-25 中南大学 Pedestrian detection method through soft cascade classifiers according to polymerization channel characteristics
CN105046197A (en) * 2015-06-11 2015-11-11 西安电子科技大学 Multi-template pedestrian detection method based on cluster
CN105184278A (en) * 2015-09-30 2015-12-23 深圳市商汤科技有限公司 Human face detection method and device
CN105550675A (en) * 2016-02-02 2016-05-04 天津大学 Binocular pedestrian detection method based on optimization polymerization integration channel
CN105760858A (en) * 2016-03-21 2016-07-13 东南大学 Pedestrian detection method and apparatus based on Haar-like intermediate layer filtering features

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
姚栖: ""基于车载摄像头的行人检测与跟踪方法研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108009554A (en) * 2017-12-01 2018-05-08 国信优易数据有限公司 A kind of image processing method and device
CN109646796A (en) * 2019-01-17 2019-04-19 浙江大学 Channel wireless radio multi closed loop stimulation system for epilepsy therapy
CN110211650A (en) * 2019-05-30 2019-09-06 苏州爱医斯坦智能科技有限公司 Operation Rich Media's electronic health record monitors the method and device of identification record automatically
CN113156935A (en) * 2020-01-07 2021-07-23 动态Ad有限责任公司 System and method for traffic light detection
CN112633313A (en) * 2020-10-13 2021-04-09 北京匠数科技有限公司 Bad information identification method of network terminal and local area network terminal equipment
CN112633313B (en) * 2020-10-13 2021-12-03 北京匠数科技有限公司 Bad information identification method of network terminal and local area network terminal equipment

Also Published As

Publication number Publication date
CN106886757B (en) 2019-06-11

Similar Documents

Publication Publication Date Title
CN108830188B (en) Vehicle detection method based on deep learning
CN109902806B (en) Method for determining target bounding box of noise image based on convolutional neural network
CN109977812B (en) Vehicle-mounted video target detection method based on deep learning
CN104050827B (en) A kind of traffic lights of view-based access control model detect recognition methods automatically
CN110163187B (en) F-RCNN-based remote traffic sign detection and identification method
CN106886757A (en) A kind of multiclass traffic lights detection method and system based on prior probability image
CN109447033A (en) Vehicle front obstacle detection method based on YOLO
CN105160309B (en) Three lanes detection method based on morphological image segmentation and region growing
CN110909666B (en) Night vehicle detection method based on improved YOLOv3 convolutional neural network
CN107122776A (en) A kind of road traffic sign detection and recognition methods based on convolutional neural networks
CN110210475B (en) License plate character image segmentation method based on non-binarization and edge detection
CN106203330A (en) A kind of vehicle classification method based on convolutional neural networks
CN108509954A (en) A kind of more car plate dynamic identifying methods of real-time traffic scene
CN105809121A (en) Multi-characteristic synergic traffic sign detection and identification method
CN103258213B (en) A kind of for the dynamic vehicle model recognizing method in intelligent transportation system
CN111553201B (en) Traffic light detection method based on YOLOv3 optimization algorithm
CN108171136A (en) A kind of multitask bayonet vehicle is to scheme to search the system and method for figure
CN103824081B (en) Method for detecting rapid robustness traffic signs on outdoor bad illumination condition
CN106778835A (en) The airport target by using remote sensing image recognition methods of fusion scene information and depth characteristic
CN107729801A (en) A kind of vehicle color identifying system based on multitask depth convolutional neural networks
CN105844257A (en) Early warning system based on machine vision driving-in-fog road denoter missing and early warning method
CN103810503A (en) Depth study based method for detecting salient regions in natural image
JPWO2020181685A5 (en)
CN106650731A (en) Robust license plate and logo recognition method
CN107273832A (en) Licence plate recognition method and system based on integrating channel feature and convolutional neural networks

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
GR01 Patent grant
GR01 Patent grant