CN107609491B - Vehicle illegal parking detection method based on convolutional neural network - Google Patents
Vehicle illegal parking detection method based on convolutional neural network Download PDFInfo
- Publication number
- CN107609491B CN107609491B CN201710729484.7A CN201710729484A CN107609491B CN 107609491 B CN107609491 B CN 107609491B CN 201710729484 A CN201710729484 A CN 201710729484A CN 107609491 B CN107609491 B CN 107609491B
- Authority
- CN
- China
- Prior art keywords
- vehicle
- size
- early warning
- current
- target vehicle
- 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.)
- Active
Links
Images
Landscapes
- Image Analysis (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention provides a vehicle illegal parking detection method based on a convolutional neural network, which comprises the following steps: step 1) respectively setting a parking forbidding area, a preset violation early warning threshold value and a preset violation threshold value in a video frame image; step 2) detecting a target vehicle in a set forbidden parking area through a convolutional neural network model, and recording vehicle information; step 3) judging whether the current vehicle set A and the historical vehicle set B in the step 2) have matchable vehicles or not; step 4) judging the current target vehicle AiWhether the time of the stay is longer than the preset violation early warning threshold in the step 1); step 5) judging the current target vehicle AiWhether the early warning detention time is longer than the preset violation threshold in the step 1) or not is judged; step 6) updating the vehicle information of the historical vehicle set B, inputting a next video frame image, and starting to perform vehicle violation detection on the next video frame image in the step 2; repeating the operations from the step 2) to the step 5).
Description
Technical Field
The invention relates to the technical field of intelligent traffic systems, deep learning and image recognition, in particular to a vehicle illegal parking detection method based on a convolutional neural network.
Background
With the progress and development of society, the number of urban motor vehicles is continuously increased, the economic loss and the number of casualties caused by road traffic accidents are also increased, the control of the occurrence of the traffic accidents becomes a problem which is more and more emphasized by traffic management departments, the leading cause of the traffic accidents is car violation behaviors, and the illegal parking is taken as a common violation behavior, so that traffic jam and traffic paralysis can be caused, and even serious traffic accidents can be caused. At present, the illegal parking detection mode is mainly manually detected. However, this requires a person to manually monitor all the places where parking is prohibited for a long time, which consumes a lot of manpower, makes it difficult to accurately measure the residence time of the vehicle against the traffic regulation, and is prone to cause missed detection due to the distractions of the monitoring person. In order to reduce the incidence of traffic accidents, traffic management departments continuously promote the construction of intelligent traffic management systems, and the intelligent traffic management systems are based on the illegal parking detection technology based on monitoring videos. Therefore, the method for detecting the overtime of the illegal parking of the vehicle based on the monitoring video has very important practical significance.
The existing illegal parking detection method has the premise that the position of a vehicle needs to be detected. The traditional video-based vehicle detection methods include a background subtraction method, a frame subtraction method, an optical flow method and the like, and the methods mainly extract a vehicle from a video image through the difference between the vehicle and a background image in terms of color, shape and the like so as to acquire information such as the position of the vehicle. In actual road scene application, the following problems often exist: 1) the outdoor illumination condition changes obviously along with time, and the video characteristics of the vehicle under different illumination conditions in the day and at night are greatly different; 2) interference factors such as shadow, light and reflection of a road surface after rain of the vehicle have similar motion characteristics with the vehicle, so that the interference factors are difficult to distinguish; 3) the video quality is seriously degraded under the severe weather conditions of rain, snow, fog, lightning and the like. These methods fail to perform effective vehicle detection in the foreground region by recognizing that the foreground region is the position where the vehicle appears, which makes it difficult to obtain the detection accuracy required for practical applications.
Disclosure of Invention
The invention aims to solve the problems of the conventional vehicle detection method based on video images, and provides a vehicle illegal parking detection method based on a convolutional neural network, which detects a target vehicle in a no-parking area by using a convolutional neural network model, acquires corresponding position information, and continuously compares and matches current vehicle information with historical vehicle information so as to judge whether the target vehicle is in overtime detention in the preset no-parking area; and comparing the image with the vehicle in the video frame image of the next period, and continuously circulating the operation to obtain the information of the detention time of the illegal parking of the target vehicle.
Different from the traditional vehicle detection method, the convolutional neural network has certain invariance to geometric transformation, deformation and illumination, effectively overcomes the difficulty brought by variable appearance of the target, can construct feature description in a self-adaptive manner under the drive of training data, and has higher flexibility and generalization capability. According to the invention, after the position of the target vehicle is detected by the convolutional neural network model, illegal parking detection is carried out, so that the detection precision and the detection speed are greatly improved.
The invention provides a vehicle illegal parking detection method based on a convolutional neural network, which comprises the following steps:
step 1) respectively setting a parking forbidding area, a preset violation early warning threshold value and a preset violation threshold value in a video frame image; obtaining a video frame image of a parking prohibition area;
step 2) detecting a target vehicle parked in the parking prohibition area set in the step 1) through a convolutional neural network model, and recording vehicle information detected in a current video frame image to form a current vehicle set A; i.e. the current vehicle set a ═ { a ═ a1,A2,...,AiAnd the area of a rectangular surrounding frame of the current vehicle set A is PAi；
Step 3) recording a historical vehicle set B ═ B in the detected video frame image in the previous period1,B2,...,BjAnd the rectangular surrounding frame area of the historical vehicle set B is PBj(ii) a Judging whether the current vehicle set A and the historical vehicle set B in the step 2) have matchable vehicles or not; if there are matchable vehicles, the target vehicle A is updatediHas a retention time ofCarrying out the next step; if no vehicles which can be matched exist, updating the target vehicle information in the video frame images recorded and detected in the current period, and directly entering the step 6);
step 4) judging the current target vehicle AiWhether the time of the stay is longer than the preset violation early warning threshold in the step 1); if the current target vehicle AiIf the time that the vehicle is detained is less than the preset violation early warning threshold value in the step 1), continuously updating and recording the current target vehicle AiThe time of retention and directly entering step 6); if the current target vehicle AiIf the time that the vehicle is detained is longer than the preset violation early warning threshold value in the step 1), updating and recording the current target vehicle AiThe time of the early warning detention and the next step;
step 5) judging the current target vehicle AiWhether the early warning detention time is longer than the preset violation threshold in the step 1) or not is judged; if the current target vehicle AiIf the early warning detention time is less than the preset violation threshold in the step 1), the current target vehicle A is continuously updatediThe time of the early warning detention and directly entering the step 6); if the current target vehicle AiIf the early warning detention time is longer than the preset violation threshold in the step 1), judging that the current target vehicle A isiIn the illegal parking state, recording and storing the time of illegal parking;
and 6) updating the vehicle information of the historical vehicle set B, and updating the vehicles which enter the illegal parking early warning state and have continuous undetected time less than the early warning judgment time in the current vehicle set A and the historical vehicle set, namely the vehicles with the early warning mark beginntpflag equal to true, into the new historical vehicle set B. Inputting a next video frame image, and entering the step 2 to start vehicle violation detection on the next video frame image according to target vehicle information in the image frame recorded and detected in the next period; repeating the operations from the step 2) to the step 5).
In the step 2), a specific process of recording the current vehicle set a is as follows:
step 2-1) inputting the video frame image in the no-parking area obtained in the step 1) into the convolutional neural network model, and detecting a rectangular surrounding frame of the current target vehicle parked in the no-parking area set in the step 1);
step 2-2) detects that the current vehicle set A ═ { A ═ A1,A2,...,Ai}, current target vehicle AiThe rectangular surrounding frame area is PAiCurrent vehicle AiThe parking delay time T, the matching flag beFoundFlag is false, the warning flag beginntpflag is false, the parking violation flag sureStopFlag is false, and the continuous interruption time break is 0.
In step 2), the convolutional neural network model includes: 1 input layer, 10 convolutional layers, 5 downsampling layers, 3 full-link layers and 1 output layer; the convolutional layer scanning boundary is automatically filled with 0, and neurons are activated by utilizing a Leaky-ReLu function; and the maximum pooling is adopted in the down-sampling layers.
In the step 2), the convolutional layer further includes: the convolution kernel size of convolution layer C1 is 7 × 7, 16 convolution kernels, the step size is 2, and the generated feature map size is 224 × 224; the convolution layer C2 has convolution kernel size of 1 × 1, 4 convolution kernels and step size of 1, and the generated feature map size is 56 × 56; the convolution layer C3 has convolution kernel size of 1 × 1, 4 convolution kernels and step size of 1, and the generated feature map size is 56 × 56; the convolution layer C4 has convolution kernel size of 3 × 3, 8 convolution kernels and step size of 1, and the generated feature map size is 56 × 56; the convolution layer C5 has convolution kernel size of 1 × 1, 8 convolution kernels and step size of 1, and the generated feature map size is 28 × 28; the convolution layer C6 has convolution kernel size of 1 × 1, 8 convolution kernels and step size of 1, and the generated feature map size is 28 × 28; the convolution layer C7 has convolution kernel size of 3 × 3, 16 convolution kernels and step size of 1, and the generated feature map size is 28 × 28; the convolution layer C8 has convolution kernel size of 3 × 3, 32 convolution kernels and step size of 1, and the generated feature map size is 14 × 14; the convolution layer C9 has convolution kernel size of 3 × 3, 64 convolution kernels and step length of 1, and the generated feature map size is 7 × 7; the window size of the downsampling layer S5 is 4 multiplied by 4, the step size is 4, and the size of the generated feature map is 2 multiplied by 2; convolutional layer C10 has a convolutional kernel size of 3 × 3, 64 convolutional kernels, a step size of 1, and a generated feature map size of 2 × 2.
In the step 2), the down-sampling layer further includes: the window size of the downsampling layer S1 is 4 multiplied by 4, the step size is 4, and the size of the generated feature map is 56 multiplied by 56; the window size of the downsampling layer S2 is 2 multiplied by 2, the step size is 2, and the size of the generated feature map is 28 multiplied by 28; the window size of the downsampling layer S3 is 2 multiplied by 2, the step size is 2, and the size of the generated feature map is 14 multiplied by 14; the downsampling layer S4 has a window size of 2 × 2, a step size of 2, and a generated feature map size of 7 × 7.
In the step 2), the fully-connected layer further includes: the fully-connected layer F1 includes 256 neurons, each of which is activated using the Relu function; the fully connected layer F2 includes 4096 neurons, each of which is activated using the leak-ReLu function.
In the step 2), the output layer further includes: the output layer includes 539 neurons that are each activated using the Relu function.
In step 3), the specific process of judging whether the matchable vehicles exist between the current vehicle set a and the historical vehicle set B in step 2) is as follows:
step 3-1) traversing a historical vehicle set B ═ B in the video frame image of the previous period according to the current vehicle set A obtained in the step 2-2)1,B2,...,BjComparing the historical vehicle set B with the current vehicle set A one by one, and judging the current vehicle AiWhether there are matchable vehicles in the historical vehicle set B;
for the detected target vehicle A in the current video frame imageiA vehicle A to be detectediRectangular bounding box area PAiAnd B in the historical vehicle setjDetected rectangular bounding box area PBjComparing one by one; the area overlapping rate IOU is two areas PAi、PBjThe intersection of (A) is divided into union, i.e. (PA)i∩PBj)/(PAi∪PBj) (ii) a Presetting a matching judgment threshold value R; if IOU is greater than or equal to R, the current target vehicle A is considerediHistorical vehicle B in the video frame image of the previous periodjIllegal parking behavior may occur for suspected stay of the same vehicleEntering the next step; if IOU is less than R, the current vehicle A is considerediIf the vehicle is newly detected, the process goes directly to step 5).
Step 3-2) extracting the current target vehicle AiDetected rectangular bounding box area PA ofiAnd corresponding history vehicle BjDetected rectangular bounding box area PBjNormalized to 48 x 48 size and grayed out, computing the LBP feature, i.e. LBPAiAnd LBPBj；
When the LBP characteristic is calculated, an LBP operator with the radius of 2 and the sampling point number of 8 is used, the LBP value of each pixel point of the whole image is calculated, a histogram of the LBP value of the whole image is counted to form the LBP characteristic of the image, and the dimension of the LBP characteristic is 59. Calculating PAiRegional image LBP feature LBPAiAnd PBjRegional image LBP feature LBPAiThe euclidean distance d between; if D is less than the Euclidean distance matching judgment threshold D, the current target vehicle AiAnd history vehicle BjMatching mutually, namely if the target vehicle Ai in the current video frame image and the vehicle Bj in the historical vehicle set are the same vehicle, entering the next step; if d is>D, then the current target vehicle AiAnd history vehicle BjNot matching, the current target vehicle AiTo newly detect a vehicle, step 6) is entered.
Step 3-3) Current target vehicle AiParking hold up time update Ai.stopTime＝BjstopTime + T, matching flag beFoundFlag true, continuous interrupt time break zero; corresponding matching vehicle BiThe matching flag befoundation flag is false.
In step 4), the current target vehicle A is judgediThe specific process of whether the time that the time is detained is greater than the preset violation early warning threshold in the step 1) is as follows:
if the current target vehicle A is updatediIf the residence time is more than the illegal parking early warning threshold value preset in the step 1), the current target vehicle A is representediEntering a violation parking early warning state, wherein an early warning mark beginnStopFlag is true; if the current target vehicle A is updatediThe residence time is smallAnd 6) entering into step 6) if the illegal parking early warning threshold value is preset in step 1).
In an actual road scene, there is a possibility that a vehicle parking violations is temporarily blocked by a running vehicle. In order to reduce the missing judgment, if all the detected target vehicles in the current video frame are not matched with the early-warned vehicle B in the historical vehiclesjEarly warning decision time TpStill-warned vehicle BjJudging to be in an early warning state, and keeping the early-warned vehicle BjMatching flag and early warning flag beginn flag true, and early warning vehicle BjResidence time update Bj.stopTime＝BjstopTime + T, pre-warned vehicle BjContinuous interrupt time update Bj.breakTimes＝BjBreak times + T. If B isj.breakTimes>TpI.e. continuously undetected early-warned vehicle BjThe time of the vehicle exceeds the early warning judgment time, and the vehicle B is considered to be the early warning vehicle BjAnd when the automobile is driven away, the early warning is relieved. The matching flag and the early warning flag beginnflag are false.
In step 5), the current target vehicle A is judgediThe specific process of whether the early warning detention time is longer than the preset violation threshold in the step 1) is as follows:
if the current target vehicle A is updatediThe early warning detention time is larger than the illegal parking threshold value set in the step 1), and the current target vehicle A is representediThe illegal parking detention state is entered, and the illegal mark sureStopFlag is true; at this time, the current target vehicle A of illegal parking is markediBackground saving of violation current target vehicle AiAnd using the picture as a judgment evidence; if the current target vehicle A is updatediAnd (4) if the early warning residence time is less than the illegal parking threshold set in the step 1), entering a step 6).
In order to avoid misjudgment, if all detected target vehicles in the current video frame are not matched with the violation vehicle B in the historical vehicle set BjJudgment time TpVehicle B still having violationjJudging the illegal parking state, and keeping the illegal vehicle BjMatching mark and violation markTsuestopplag true, violation vehicle BjResidence time update Bj.stopTime＝BjstopTime + T, offended vehicle BjContinuous interrupt time update Bj.breakTimes＝BjBreakthrough times + T; if B isj.breakTimes>TpI.e. continuously undetected against the offending vehicle BjIf the time exceeds the early warning judgment time, the violation vehicle is considered to be driven away; matching the mark, the early warning mark and the violation mark beFound flag ═ begestfplag ═ sureStopFlag ═ false, recording and storing the total violation residence time stopTime ═ stopTime-Tp。
The invention has the advantages that: the invention can realize uninterrupted detection of the no-parking area by automatically detecting the no-parking area appointed and set by the input video frame image, and overcomes the defects of large human resource consumption, incapability of detecting the illegal parking for a long time and the like caused by the current method for monitoring the video of the road by using manpower to detect the illegal parking. In addition, the target vehicle in the image frame can be effectively identified by using the convolutional neural network model, and compared with a method for extracting the target vehicle by using motion characteristics, the method has the advantages of higher accuracy and wider application range; the invention can avoid misjudging the vehicle staying in a short time as the illegal parking vehicle by utilizing the related judgment of the illegal parking early warning, can greatly reduce the missed judgment caused by the reason that the illegal parking vehicle is temporarily shielded and the like, and realizes the accurate evidence obtaining of the illegal parking behavior.
Drawings
FIG. 1 is a flow chart diagram generally describing a vehicle illegal parking detection method based on a convolutional neural network of the invention;
fig. 2(a) is a schematic diagram of the effect of setting a parking prohibition area in step 1 of the vehicle illegal parking detection method based on the convolutional neural network in the embodiment of the present invention;
fig. 2(b) is a schematic diagram of the effect of detecting the vehicle set a in step 2 of the vehicle illegal parking detection method based on the convolutional neural network in the embodiment of the invention;
fig. 2(c) is a schematic diagram of the effect of entering the illegal parking early warning state of the vehicle in step 4 of the vehicle illegal parking detection method based on the convolutional neural network in the embodiment of the invention;
fig. 2(d) is a schematic diagram of the effect of entering the illegal parking detention state of the vehicle by step 5 of the vehicle illegal parking detection method based on the convolutional neural network in the embodiment of the present invention;
fig. 2(e) is a schematic diagram of the effect of the illegal vehicle driving-away state of step 5 of the vehicle illegal parking detection method based on the convolutional neural network in the embodiment of the invention;
FIG. 3 is a block diagram of a vehicle detection convolutional neural network used in an embodiment of the present invention;
fig. 4 is a flow chart schematic diagram of a vehicle illegal parking detection method based on a convolutional neural network.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
Different from the traditional vehicle detection method, the convolutional neural network has certain invariance to geometric transformation, deformation and illumination, effectively overcomes the difficulty brought by variable appearance of the target, can construct feature description in a self-adaptive manner under the drive of training data, and has higher flexibility and generalization capability. According to the invention, after the position of the target vehicle is detected by the convolutional neural network model, illegal parking detection is carried out, so that the detection precision and the detection speed are greatly improved.
As shown in fig. 1, the invention provides a vehicle illegal parking detection method based on a convolutional neural network, which comprises the following brief steps:
initializing an illegal parking detection area and illegal parking parameters;
inputting a video and acquiring an image to be processed;
detecting vehicles in the video frame image area by using a convolutional neural network, and setting current vehicle information;
comparing the current vehicle information with the historical vehicle information, and updating the historical vehicle information;
and judging whether the illegal vehicle exists or not. If so, marking the illegal parking vehicle, storing the vehicle information and entering the next judgment; if not, directly entering the next step of judgment;
judging whether the current video frame is the last frame or not, if so, ending the algorithm; and if not, continuously inputting the video to acquire the image to be processed to detect the next frame of illegal parking vehicle.
As shown in fig. 4, the invention provides a vehicle illegal parking detection method based on a convolutional neural network, which comprises the following specific contents:
step 1) respectively setting a parking forbidding area, a preset violation early warning threshold value and a preset violation threshold value in a video frame image; obtaining a video frame image of a parking prohibition area; as shown in fig. 2 (a);
step 2) detecting a target vehicle parked in the parking prohibition area set in the step 1) through a convolutional neural network model, and recording vehicle information in a video frame image detected in the current period to form a current vehicle set A; i.e. the current vehicle set a ═ { a ═ a1,A2,...,AiAnd the area of a rectangular surrounding frame of the current vehicle set A is PAi(ii) a As shown in FIG. 2 (b);
step 3) assuming that the historical vehicle set B ═ B in the image frame detected in the last period record1,B2,...,BjAnd the rectangular surrounding frame area of the historical vehicle set B is PBj(ii) a Judging whether the current vehicle set A and the historical vehicle set B in the step 2) have matchable vehicles or not; if there are matchable vehicles, the target vehicle A is updatediThe time of retention and the next step; if no vehicles which can be matched exist, the information of the target vehicles detected in the current video frame image is updated, and the step 6) is directly carried out;
step 4) judging the current target vehicle AiWhether the time of the stay is longer than the preset violation early warning threshold in the step 1); if the current target vehicle AiIf the time that the vehicle is detained is less than the preset violation early warning threshold value in the step 1), continuously updating and recording the current target vehicle AiThe time of retention and directly entering step 6); if the current target vehicle AiHas been retained for a time greater than step1) The preset violation early warning threshold value in (1) indicates that the current target vehicle Ai enters a violation parking early warning state, and as shown in (c) of fig. 2, the current target vehicle A is updated and recordediThe time of the early warning detention and the next step;
step 5) judging the current target vehicle AiWhether the early warning detention time is longer than the preset violation threshold in the step 1) or not is judged; if the current target vehicle AiIf the early warning detention time is less than the preset violation threshold in the step 1), the current target vehicle A is continuously updatediThe time of the early warning detention and directly entering the step 6); if the current target vehicle AiIf the early warning detention time is longer than the preset violation threshold in the step 1), judging that the current target vehicle A isiIn the illegal parking state, recording and storing the time of illegal parking;
and 6) updating the vehicle information of the historical vehicle set B, and updating the vehicles which enter the illegal parking early warning state and have continuous undetected time less than the early warning judgment time in the current vehicle set A and the historical vehicle set, namely the vehicles with the early warning mark beginntpflag equal to true, into the new historical vehicle set B. Inputting a next video frame image, and entering the step 2 to start vehicle violation detection on the next video frame image according to target vehicle information in the image frame recorded and detected in the next period; repeating the operations from the step 2) to the step 5).
In the step 2), a specific process of recording the current vehicle set a is as follows:
step 2-1) inputting the video image in the no-parking area obtained in the step 1) into the convolutional neural network model, and detecting a rectangular surrounding frame of the current target vehicle parked in the no-parking area set in the step 1);
step 2-2) detecting a vehicle set A ═ { A ] from the current video frame image1,A2,...,Ai}, current target vehicle AiThe rectangular surrounding frame area is PAiAssume that the current vehicle AiIs in the set no-stop zone, and the current vehicle AiStop residence time of (1) ═ T, pThe matched flag beFoundFlag is equal to false, the early warning flag beginnstopflag is equal to false, the stop-violation flag sureStopFlag is equal to false, and the continuous interruption time break is equal to 0.
In step 2), the convolutional neural network model includes: 1 input layer, 10 convolutional layers, 5 downsampling layers, 3 full-link layers and 1 output layer; the convolutional layer scanning boundary is automatically filled with 0, and neurons are activated by utilizing a Leaky-ReLu function; and the maximum pooling is adopted in the down-sampling layers.
In the step 2), as shown in fig. 3, the convolutional layer further includes: the convolution kernel size of convolution layer C1 is 7 × 7, 16 convolution kernels, the step size is 2, and the generated feature map size is 224 × 224; the convolution layer C2 has convolution kernel size of 1 × 1, 4 convolution kernels and step size of 1, and the generated feature map size is 56 × 56; the convolution layer C3 has convolution kernel size of 1 × 1, 4 convolution kernels and step size of 1, and the generated feature map size is 56 × 56; the convolution layer C4 has convolution kernel size of 3 × 3, 8 convolution kernels and step size of 1, and the generated feature map size is 56 × 56; the convolution layer C5 has convolution kernel size of 1 × 1, 8 convolution kernels and step size of 1, and the generated feature map size is 28 × 28; the convolution layer C6 has convolution kernel size of 1 × 1, 8 convolution kernels and step size of 1, and the generated feature map size is 28 × 28; the convolution layer C7 has convolution kernel size of 3 × 3, 16 convolution kernels and step size of 1, and the generated feature map size is 28 × 28; the convolution layer C8 has convolution kernel size of 3 × 3, 32 convolution kernels and step size of 1, and the generated feature map size is 14 × 14; the convolution layer C9 has convolution kernel size of 3 × 3, 64 convolution kernels and step length of 1, and the generated feature map size is 7 × 7; the window size of the downsampling layer S5 is 4 multiplied by 4, the step size is 4, and the size of the generated feature map is 2 multiplied by 2; convolutional layer C10 has a convolutional kernel size of 3 × 3, 64 convolutional kernels, a step size of 1, and a generated feature map size of 2 × 2.
In the step 2), the down-sampling layer further includes: the window size of the downsampling layer S1 is 4 multiplied by 4, the step size is 4, and the size of the generated feature map is 56 multiplied by 56; the window size of the downsampling layer S2 is 2 multiplied by 2, the step size is 2, and the size of the generated feature map is 28 multiplied by 28; the window size of the downsampling layer S3 is 2 multiplied by 2, the step size is 2, and the size of the generated feature map is 14 multiplied by 14; the downsampling layer S4 has a window size of 2 × 2, a step size of 2, and a generated feature map size of 7 × 7.
In the step 2), the fully-connected layer further includes: the fully-connected layer F1 includes 256 neurons, each of which is activated using the Relu function; the fully connected layer F2 includes 4096 neurons, each of which is activated using the leak-ReLu function.
In the step 2), the output layer further includes: the output layer includes 539 neurons that are each activated using the Relu function.
In step 3), the specific process of judging whether the matchable vehicles exist between the current vehicle set a and the historical vehicle set B in step 2) is as follows:
step 3-1) traversing the historical vehicle set B ═ B of the previous cycle image frame according to the current vehicle set A obtained in the step 2-2)1,B2,...,BjComparing the historical vehicle set B with the current vehicle set A one by one, and judging the current vehicle AiWhether there are matchable vehicles in the historical vehicle set B;
for the detected target vehicle A in the current video frame imageiThe detected current target vehicle AiRectangular bounding box area PAiAnd B in the historical vehicle setjDetected rectangular bounding box area PBjComparing one by one; the area overlapping rate IOU is two areas PAi、PBjThe intersection of (A) is divided into union, i.e. (PA)i∩PBj)/(PAi∪PBj) (ii) a Presetting a matching judgment threshold value R; if IOU is greater than or equal to R, the current target vehicle A is considerediAnd the historical vehicle B of the previous periodjIf the same vehicle is suspected to be detained and illegal parking behaviors possibly occur, the next step is carried out; if IOU is less than R, the current vehicle A is considerediIf the vehicle is newly detected, the process goes directly to step 5).
Step 3-2) extracting the current target vehicle AiDetected rectangular bounding box area PA ofiAnd corresponding history vehicle BjDetected rectangular bounding box area PBjNormalized to 48 x 48 size and grayed out, computing the LBP feature, i.e. LBPAiAnd LBPBj；
When the LBP characteristic is calculated, an LBP operator with the radius of 2 and the sampling point number of 8 is used, the LBP value of each pixel point of the whole image is calculated, a histogram of the LBP value of the whole image is counted to form the LBP characteristic of the image, and the dimension of the LBP characteristic is 59. Calculating PAiRegional image LBP feature LBPAiAnd PBjRegional image LBP feature LBPAiThe euclidean distance d between; if D is less than the Euclidean distance matching judgment threshold D, the current target vehicle AiAnd history vehicle BjMatching mutually, namely if the target vehicle Ai in the current video frame image and the vehicle Bj in the historical vehicle set are the same vehicle, entering the next step; if d is>D, then the current target vehicle AiAnd history vehicle BjNot matching, the current target vehicle AiTo newly detect a vehicle, step 6) is entered.
Step 3-3) Current target vehicle AiParking hold up time update Ai.stopTime＝BjstopTime + T, matching flag beFoundFlag true, continuous interrupt time break zero; corresponding matching vehicle BiThe matching flag befoundation flag is false.
In step 4), the current target vehicle A is judgediThe specific process of whether the time that the time is detained is greater than the preset violation early warning threshold in the step 1) is as follows:
if the current target vehicle A is updatediIf the residence time is more than the illegal parking early warning threshold value preset in the step 1), the current target vehicle A is representediEntering a violation parking early warning state, wherein an early warning mark beginnStopFlag is true; if the current target vehicle A is updatediIf the residence time is less than the illegal parking early warning threshold value preset in the step 1), the step 6) is carried out.
In an actual road scene, there is a possibility that a vehicle parking violations is temporarily blocked by a running vehicle. In order to reduce the missing judgment, if all the detected target vehicles in the current video frame are not matched with the early-warned vehicle B in the historical vehiclesjEarly warning decision time TpStill-warned vehicle BjThe state of the early warning is judged,keep pre-warned vehicle BjMatching flag and early warning flag beginn flag true, and early warning vehicle BjResidence time update Bj.stopTime＝BjstopTime + T, pre-warned vehicle BjContinuous interrupt time update Bj.breakTimes＝BjBreak times + T. If B isj.breakTimes>TpI.e. continuously undetected early-warned vehicle BjThe time of the vehicle exceeds the early warning judgment time, and the vehicle B is considered to be the early warning vehicle BjAnd when the automobile is driven away, the early warning is relieved. The matching flag and the early warning flag beginnflag are false.
In step 5), the current target vehicle A is judgediThe specific process of whether the early warning detention time is longer than the preset violation threshold in the step 1) is as follows:
if the current target vehicle A is updatediThe early warning detention time is larger than the illegal parking threshold value set in the step 1), and the current target vehicle A is representediThe illegal parking stay state has been entered, as shown in fig. 2(d), the violation flag sureStopFlag is true; at this time, the current target vehicle A of illegal parking is markediBackground saving of violation current target vehicle AiAnd using the picture as a judgment evidence; if the current target vehicle A is updatediAnd (4) if the early warning residence time is less than the illegal parking threshold set in the step 1), entering a step 6).
In order to avoid misjudgment, if all detected target vehicles in the current video frame are not matched with the violation vehicle B in the historical vehicle set BjJudgment time TpVehicle B still having violationjJudging the illegal parking state, and keeping the illegal vehicle BjMatching the flag with the violation flag beFoundFlag (sureStopFlag) true, and the violated vehicle BjResidence time update Bj.stopTime＝BjstopTime + T, offended vehicle BjContinuous interrupt time update Bj.breakTimes＝BjBreakthrough times + T; if B isj.breakTimes>TpI.e. continuously undetected against the offending vehicle BjHas exceeded the warning decision time, as shown in fig. 2(e), the violation is consideredThe badge vehicle has driven away; matching the mark, the early warning mark and the violation mark beFound flag ═ begestfplag ═ sureStopFlag ═ false, recording and storing the total violation residence time stopTime ═ stopTime-Tp。
Finally, it should be noted that the above embodiments are only used for illustrating the technical solutions of the present invention and are not limited. Although the present invention has been described in detail with reference to the embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (7)
1. A vehicle illegal parking detection method based on a convolutional neural network comprises the following steps:
initializing an illegal parking detection area and illegal parking parameters;
inputting a video and acquiring an image to be processed;
detecting vehicles in the video frame image area by using a convolutional neural network, and setting current vehicle information;
comparing the current vehicle information with the historical vehicle information, and updating the historical vehicle information;
judging whether an illegal parking vehicle exists or not; if so, marking the illegal parking vehicle, storing the vehicle information and entering the next judgment; if not, directly entering the next step of judgment;
judging whether the current video frame image is the last frame or not, if so, ending the algorithm; if not, continuously inputting the video to acquire the image to be processed to detect the next frame of illegal parking vehicle;
the method comprises the following specific steps:
step 1) respectively setting a parking forbidding area, a preset violation early warning threshold value and a preset violation threshold value in a video frame image; obtaining a video frame image of a parking prohibition area;
step 2) detecting a target vehicle parked in the parking prohibition area set in the step 1) through a convolutional neural network model, and recording vehicle information detected in a current video frame image; forming a current vehicle set A;
step 3) judging whether the current vehicle set A and the historical vehicle set B in the step 2) have matchable vehicles or not; if there are matchable vehicles, the target vehicle A is updatediThe time of retention and the next step; if no vehicles which can be matched exist, updating the target vehicle information in the video frame images recorded and detected in the current period, and directly entering the step 6);
the specific process of the step 3) is as follows:
step 3-1) traversing a historical vehicle set B ═ B in the video frame image of the previous period according to the current vehicle set A obtained in the step 2-2)1,B2,...,BjComparing the historical vehicle set B with the current vehicle set A one by one, and judging the current vehicle AiWhether there are matchable vehicles in the historical vehicle set B;
for the detected target vehicle A in the current video frame imageiA vehicle A to be detectediRectangular bounding box area PAiAnd B in the historical vehicle setjDetected rectangular bounding box area PBjComparing one by one; the area overlapping rate IOU is two areas PAi、PBjIntersect disjunction set (PA) ofi∩PBj)/(PAi∪PBj) (ii) a Presetting a matching judgment threshold value R; if IOU is greater than or equal to R, the current target vehicle A is considerediHistorical vehicle B in the video frame image of the previous periodjIf the same vehicle is suspected to be detained and illegal parking behaviors possibly occur, the next step is carried out; if IOU is less than R, the current vehicle A is considerediIf the vehicle is newly detected, directly entering the step 5);
step 3-2) extracting the current target vehicle AiDetected rectangular bounding box area PA ofiAnd corresponding history vehicle BjDetected rectangular bounding box area PBjNormalized to 48 x 48 size, grayed out and calculated LBP characteristic LBPAiAnd LBPBj；
Calculating PAiRegional image LBP feature LBPAiAnd PBjRegional image LBP feature LBPAiEuclidean distance betweenD, separating; if D is less than the Euclidean distance matching judgment threshold D, the current target vehicle AiAnd history vehicle BjMatching the target vehicle Ai in the current video frame image with the vehicle Bj in the historical vehicle set, and entering the next step if the target vehicle Ai in the current video frame image and the vehicle Bj in the historical vehicle set are the same vehicle; if d is>D, then the current target vehicle AiAnd history vehicle BjNot matching, the current target vehicle AiEntering step 6) for newly detecting the vehicle;
step 3-3) Current target vehicle AiParking hold up time update Ai.stopTime＝BjstopTime + T, matching flag beFoundFlag true, continuous interrupt time break zero; corresponding matching vehicle BiThe matching flag beFoundFlag is false;
step 4) judging the current target vehicle AiWhether the time of the stay is longer than the preset violation early warning threshold in the step 1); if the current target vehicle AiIf the time that the vehicle is detained is less than the preset violation early warning threshold value in the step 1), continuously updating and recording the current target vehicle AiThe time of retention and directly entering step 6); if the current target vehicle AiIf the time that the vehicle is detained is longer than the preset violation early warning threshold value in the step 1), updating and recording the current target vehicle AiThe time of the early warning detention and the next step;
step 5) judging the current target vehicle AiWhether the early warning detention time is longer than the preset violation threshold in the step 1) or not is judged; if the current target vehicle AiIf the early warning detention time is less than the preset violation threshold in the step 1), the current target vehicle A is continuously updatediThe time of the early warning detention and directly entering the step 6); if the current target vehicle AiIf the early warning detention time is longer than the preset violation threshold in the step 1), judging that the current target vehicle A isiIn the illegal parking state, recording and storing the time of illegal parking;
step 6) updating the vehicle information of the historical vehicle set B, inputting a next video frame image, and starting to perform vehicle violation detection on the next video frame image in the step 2 according to the target vehicle information in the image frame recorded and detected in the next period; repeating the operations from the step 2) to the step 5).
2. The vehicle illegal parking detection method based on the convolutional neural network as claimed in claim 1, wherein in the step 2), a specific process of recording the current vehicle set a is as follows:
step 2-1) inputting the video frame image in the no-parking area obtained in the step 1) into the vehicle model based on the convolutional neural network, and detecting a rectangular surrounding frame of a current target vehicle parked in the no-parking area set in the step 1);
step 2-2) detects that the current vehicle set A ═ { A ═ A1,A2,...,Ai}, current target vehicle AiThe rectangular surrounding frame area is PAiCurrent vehicle AiThe parking delay time T, the matching flag beFoundFlag is false, the warning flag beginntpflag is false, the parking violation flag sureStopFlag is false, and the continuous interruption time break is 0.
3. The vehicle illegal parking detection method based on the convolutional neural network as claimed in claim 2, wherein in the step 2), the convolutional neural network model comprises: 1 input layer, 10 convolutional layers, 5 downsampling layers, 3 full-link layers and 1 output layer; the convolutional layer scanning boundary is automatically filled with 0, and neurons are activated by utilizing a Leaky-ReLu function; and the maximum pooling is adopted in the down-sampling layers.
4. The vehicle illegal parking detection method based on the convolutional neural network as claimed in claim 3, wherein in the step 2), the convolutional layer further comprises: the convolution kernel size of convolution layer C1 is 7 × 7, 16 convolution kernels, the step size is 2, and the generated feature map size is 224 × 224; the convolution layer C2 has convolution kernel size of 1 × 1, 4 convolution kernels and step size of 1, and the generated feature map size is 56 × 56; the convolution layer C3 has convolution kernel size of 1 × 1, 4 convolution kernels and step size of 1, and the generated feature map size is 56 × 56; the convolution layer C4 has convolution kernel size of 3 × 3, 8 convolution kernels and step size of 1, and the generated feature map size is 56 × 56; the convolution layer C5 has convolution kernel size of 1 × 1, 8 convolution kernels and step size of 1, and the generated feature map size is 28 × 28; the convolution layer C6 has convolution kernel size of 1 × 1, 8 convolution kernels and step size of 1, and the generated feature map size is 28 × 28; the convolution layer C7 has convolution kernel size of 3 × 3, 16 convolution kernels and step size of 1, and the generated feature map size is 28 × 28; the convolution layer C8 has convolution kernel size of 3 × 3, 32 convolution kernels and step size of 1, and the generated feature map size is 14 × 14; the convolution layer C9 has convolution kernel size of 3 × 3, 64 convolution kernels and step length of 1, and the generated feature map size is 7 × 7; the window size of the downsampling layer S5 is 4 multiplied by 4, the step size is 4, and the size of the generated feature map is 2 multiplied by 2; convolutional layer C10 has a convolutional kernel size of 3 × 3, 64 convolutional kernels, a step size of 1, and a generated feature map size of 2 × 2.
5. The convolutional neural network-based vehicle illegal parking detection method as recited in claim 3, wherein in the step 2), the down-sampling layer further comprises: the window size of the downsampling layer S1 is 4 multiplied by 4, the step size is 4, and the size of the generated feature map is 56 multiplied by 56; the window size of the downsampling layer S2 is 2 multiplied by 2, the step size is 2, and the size of the generated feature map is 28 multiplied by 28; the window size of the downsampling layer S3 is 2 multiplied by 2, the step size is 2, and the size of the generated feature map is 14 multiplied by 14; the downsampling layer S4 has a window size of 2 × 2, a step size of 2, and a generated feature map size of 7 × 7.
6. The vehicle illegal parking detection method based on the convolutional neural network as claimed in claim 1, characterized in that the specific process of step 4) is as follows:
if the current target vehicle A is updatediIf the residence time is more than the illegal parking early warning threshold value preset in the step 1), the current target vehicle A is representediEntering a violation parking early warning state, wherein an early warning mark beginnStopFlag is true; if the current target vehicle A is updatediIf the residence time is less than the illegal parking early warning threshold value preset in the step 1), entering a step 6);
if all detected target vehicles in the current video frame fail to be matched with the early-warned vehicle B in the historical vehiclesjEarly warning decision time TpStill-warned vehicle BjJudging to be in an early warning state, and keeping the early-warned vehicle BjMatching flag and early warning flag beginn flag true, and early warning vehicle BjResidence time update Bj.stopTime＝BjstopTime + T, pre-warned vehicle BjContinuous interrupt time update Bj.breakTimes＝BjBreakthrough times + T; if B isj.breakTimes>TpContinuously undetected early-warned vehicle BjThe time of the vehicle exceeds the early warning judgment time, and the vehicle B is considered to be the early warning vehicle BjWhen the vehicle is driven away, the early warning is removed; the matching flag and the early warning flag beginnflag are false.
7. The vehicle illegal parking detection method based on the convolutional neural network as claimed in claim 1, characterized in that the specific process of step 5) is as follows:
if the current target vehicle A is updatediThe early warning detention time is larger than the illegal parking threshold value set in the step 1), and the current target vehicle A is representediThe illegal parking detention state is entered, and the illegal mark sureStopFlag is true; at this time, the current target vehicle A of illegal parking is markediBackground saving of violation current target vehicle AiAnd using the picture as a judgment evidence; if the current target vehicle A is updatediIf the early warning residence time is less than the illegal parking threshold set in the step 1), entering a step 6);
if all detected target vehicles in the current video frame fail to be matched with the violation vehicle B in the historical vehicle set BjJudgment time TpVehicle B still having violationjJudging the illegal parking state, and keeping the illegal vehicle BjMatching the flag with the violation flag beFoundFlag (sureStopFlag) true, and the violated vehicle BjResidence time update Bj.stopTime＝BjstopTime + T, offended vehicle BjContinuous interrupt time update Bj.breakTimes＝BjBreakthrough times + T; if B isj.breakTimes>TpContinuously undetected against traffic violation BjIf the time exceeds the early warning judgment time, the violation vehicle is considered to be driven away; matching the mark, the early warning mark and the violation mark beFound flag ═ begestfplag ═ sureStopFlag ═ false, recording and storing the total violation residence time stopTime ═ stopTime-Tp。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710729484.7A CN107609491B (en) | 2017-08-23 | 2017-08-23 | Vehicle illegal parking detection method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710729484.7A CN107609491B (en) | 2017-08-23 | 2017-08-23 | Vehicle illegal parking detection method based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107609491A CN107609491A (en) | 2018-01-19 |
| CN107609491B true CN107609491B (en) | 2020-05-26 |
Family
ID=61065691
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710729484.7A Active CN107609491B (en) | 2017-08-23 | 2017-08-23 | Vehicle illegal parking detection method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107609491B (en) |
Families Citing this family (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109033950B (en) * | 2018-06-12 | 2020-07-17 | 浙江工业大学 | Vehicle illegal parking detection method based on multi-feature fusion cascade depth model |
| CN110659546B (en) * | 2018-06-29 | 2022-11-01 | 杭州海康威视数字技术股份有限公司 | Illegal booth detection method and device |
| CN109063612A (en) * | 2018-07-19 | 2018-12-21 | 中智城信息技术有限公司 | City intelligent red line management method and machine readable storage medium |
| CN109284752A (en) * | 2018-08-06 | 2019-01-29 | 中国科学院声学研究所 | A kind of rapid detection method of vehicle |
| CN109299672B (en) * | 2018-09-05 | 2022-02-15 | 重庆大学 | Parking event detection system and method capable of adaptively adjusting threshold value and algorithm structure |
| CN109118779B (en) * | 2018-10-12 | 2021-05-11 | 东软集团股份有限公司 | Traffic violation information identification method, equipment and readable storage medium |
| CN109598941A (en) * | 2018-11-21 | 2019-04-09 | 长沙小如信息科技有限公司 | The self-locking system stopped and method are disobeyed about vehicle |
| CN109902676B (en) * | 2019-01-12 | 2021-04-09 | 浙江工业大学 | Dynamic background-based violation detection algorithm |
| CN109658698A (en) * | 2019-01-25 | 2019-04-19 | 合肥市智信汽车科技有限公司 | A kind of detection of motor vehicle illegal running and grasp shoot method based on deep learning |
| CN110032947B (en) * | 2019-03-22 | 2021-11-19 | 深兰科技（上海）有限公司 | Method and device for monitoring occurrence of event |
| CN110135356A (en) | 2019-05-17 | 2019-08-16 | 北京百度网讯科技有限公司 | The detection method and device of parking offense, electronic equipment, computer-readable medium |
| CN110491132B (en) * | 2019-07-11 | 2022-04-08 | 平安科技（深圳）有限公司 | Vehicle illegal parking detection method and device based on video frame picture analysis |
| CN110633699B (en) * | 2019-10-09 | 2021-05-11 | 华南理工大学 | Visual detection method for parking behavior of interaction area of AGV intelligent parking system |
| CN111105621B (en) * | 2019-12-17 | 2021-11-16 | 浙江大华技术股份有限公司 | Method and device for detecting illegal parking |
| CN111147738A (en) * | 2019-12-19 | 2020-05-12 | 武汉裕众信息科技有限公司 | Police vehicle-mounted panoramic and coma system, device, electronic equipment and medium |
| CN111429726B (en) * | 2020-04-13 | 2021-11-26 | 城云科技（中国）有限公司 | Monitoring video illegal parking vehicle detection and management method and corresponding system |
| CN111754786A (en) * | 2020-07-15 | 2020-10-09 | 遵义同望智能科技有限公司 | System for identifying traffic vehicle passing events on highway |
| CN112289037B (en) * | 2020-10-29 | 2022-06-07 | 南通中铁华宇电气有限公司 | Motor vehicle illegal parking detection method and system based on high visual angle under complex environment |
| CN113096394A (en) * | 2021-03-29 | 2021-07-09 | 浙江大华技术股份有限公司 | Method and device for determining object of target type, storage medium and electronic device |
| FR3122019B1 (en) * | 2021-04-20 | 2024-01-05 | Urbaflux | System for detecting vehicles in a demarcated location, the stop terminal for parking management and its operating method |
| CN114664096B (en) * | 2022-03-24 | 2023-04-28 | 北京四象网讯科技有限公司 | Method and device for processing monitoring video of parking lot |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101183427A (en) * | 2007-12-05 | 2008-05-21 | 浙江工业大学 | Computer vision based peccancy parking detector |
| CN102568206A (en) * | 2012-01-13 | 2012-07-11 | 大连民族学院 | Video monitoring-based method for detecting cars parking against regulations |
| CN103093194A (en) * | 2013-01-07 | 2013-05-08 | 信帧电子技术（北京）有限公司 | Breach of regulation vehicle detection method and device based on videos |
| CN103116985A (en) * | 2013-01-21 | 2013-05-22 | 信帧电子技术（北京）有限公司 | Detection method and device of parking against rules |
| CN104036323A (en) * | 2014-06-26 | 2014-09-10 | 叶茂 | Vehicle detection method based on convolutional neural network |
| CN104376554A (en) * | 2014-10-16 | 2015-02-25 | 中海网络科技股份有限公司 | Illegal parking detection method based on image texture |
| CN105513371A (en) * | 2016-01-15 | 2016-04-20 | 昆明理工大学 | Expressway illegal parking detection method based on kernel density estimation |
| CN105809975A (en) * | 2016-05-30 | 2016-07-27 | 北京精英智通科技股份有限公司 | Abnormal parking judgment method and abnormal parking judgment device |
| CN105957352A (en) * | 2016-06-01 | 2016-09-21 | 深圳中兴力维技术有限公司 | Illegal parking detection method and apparatus |
| CN205609019U (en) * | 2016-05-10 | 2016-09-28 | 北京精英智通科技股份有限公司 | Warning system of parkking breaks law |
| CN106571039A (en) * | 2016-08-22 | 2017-04-19 | 中海网络科技股份有限公司 | Automatic snapshot system for highway traffic offence |
| CN106886755A (en) * | 2017-01-19 | 2017-06-23 | 北京航空航天大学 | A kind of intersection vehicles system for detecting regulation violation based on Traffic Sign Recognition |
| CN106935035A (en) * | 2017-04-07 | 2017-07-07 | 西安电子科技大学 | Parking offense vehicle real-time detection method based on SSD neutral nets |
-
2017
- 2017-08-23 CN CN201710729484.7A patent/CN107609491B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101183427A (en) * | 2007-12-05 | 2008-05-21 | 浙江工业大学 | Computer vision based peccancy parking detector |
| CN102568206A (en) * | 2012-01-13 | 2012-07-11 | 大连民族学院 | Video monitoring-based method for detecting cars parking against regulations |
| CN103093194A (en) * | 2013-01-07 | 2013-05-08 | 信帧电子技术（北京）有限公司 | Breach of regulation vehicle detection method and device based on videos |
| CN103116985A (en) * | 2013-01-21 | 2013-05-22 | 信帧电子技术（北京）有限公司 | Detection method and device of parking against rules |
| CN104036323A (en) * | 2014-06-26 | 2014-09-10 | 叶茂 | Vehicle detection method based on convolutional neural network |
| CN104376554A (en) * | 2014-10-16 | 2015-02-25 | 中海网络科技股份有限公司 | Illegal parking detection method based on image texture |
| CN105513371A (en) * | 2016-01-15 | 2016-04-20 | 昆明理工大学 | Expressway illegal parking detection method based on kernel density estimation |
| CN205609019U (en) * | 2016-05-10 | 2016-09-28 | 北京精英智通科技股份有限公司 | Warning system of parkking breaks law |
| CN105809975A (en) * | 2016-05-30 | 2016-07-27 | 北京精英智通科技股份有限公司 | Abnormal parking judgment method and abnormal parking judgment device |
| CN105957352A (en) * | 2016-06-01 | 2016-09-21 | 深圳中兴力维技术有限公司 | Illegal parking detection method and apparatus |
| CN106571039A (en) * | 2016-08-22 | 2017-04-19 | 中海网络科技股份有限公司 | Automatic snapshot system for highway traffic offence |
| CN106886755A (en) * | 2017-01-19 | 2017-06-23 | 北京航空航天大学 | A kind of intersection vehicles system for detecting regulation violation based on Traffic Sign Recognition |
| CN106935035A (en) * | 2017-04-07 | 2017-07-07 | 西安电子科技大学 | Parking offense vehicle real-time detection method based on SSD neutral nets |
Non-Patent Citations (2)
| Title |
|---|
| 《基于视频的高速公路违章停车实时检测算法》;张贝贝 等；;《电子科技》;20110915;第24卷(第9期);第20-23页； * |
| 《基于视频监控的违章停车检测算法》;姜明新 等；;《计算机工程》;20121031;第38卷(第19期);第151-158页； * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107609491A (en) | 2018-01-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107609491B (en) | Vehicle illegal parking detection method based on convolutional neural network | |
| CN110178167B (en) | Intersection violation video identification method based on cooperative relay of cameras | |
| CN106652465B (en) | Method and system for identifying abnormal driving behaviors on road | |
| CN109670376B (en) | Lane line identification method and system | |
| CN104537841B (en) | Unlicensed vehicle violation detection method and detection system thereof | |
| CN111368687A (en) | Sidewalk vehicle illegal parking detection method based on target detection and semantic segmentation | |
| US20130266187A1 (en) | Video-based method for parking angle violation detection | |
| CN111597905B (en) | Highway tunnel parking detection method based on video technology | |
| CN112339773B (en) | Monocular vision-based non-active lane departure early warning method and system | |
| Widyantoro et al. | Traffic lights detection and recognition based on color segmentation and circle hough transform | |
| Poddar et al. | Automated traffic monitoring system using computer vision | |
| CN114781479A (en) | Traffic incident detection method and device | |
| CN113903008A (en) | Ramp exit vehicle violation identification method based on deep learning and trajectory tracking | |
| Ren et al. | Automatic measurement of traffic state parameters based on computer vision for intelligent transportation surveillance | |
| CN110782485A (en) | Vehicle lane change detection method and device | |
| Sai Ganesh Naik et al. | Detecting potholes using image processing techniques and real-world footage | |
| Rouf et al. | Real-time vehicle detection, tracking and counting system based on YOLOv7 | |
| Dong et al. | Detection method for vehicles in tunnels based on surveillance images | |
| Al-Kadi et al. | Road scene analysis for determination of road traffic density | |
| Al Noman et al. | A Computer Vision-Based Lane Detection Approach for an Autonomous Vehicle Using the Image Hough Transformation and the Edge Features | |
| CN112686136B (en) | Object detection method, device and system | |
| Wangsiripitak et al. | Traffic light and crosswalk detection and localization using vehicular camera | |
| Deng et al. | A detection method of traffic parameters based on EPI | |
| Chen et al. | A forward collision avoidance system adopting multi-feature vehicle detection | |
| Bachtiar et al. | Vehicle classification and violation detection on traffic light area using BLOB and mean-shift tracking method |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20221206 Address after: 100190, No. 21 West Fourth Ring Road, Beijing, Haidian District Patentee after: INSTITUTE OF ACOUSTICS, CHINESE ACADEMY OF SCIENCES Patentee after: NANHAI RESEARCH STATION, INSTITUTE OF ACOUSTICS, CHINESE ACADEMY OF SCIENCES Address before: 100190, No. 21 West Fourth Ring Road, Beijing, Haidian District Patentee before: INSTITUTE OF ACOUSTICS, CHINESE ACADEMY OF SCIENCES |
|
| TR01 | Transfer of patent right |