CN102426785B - Traffic flow information perception method based on contour and local characteristic point and system thereof - Google Patents
Traffic flow information perception method based on contour and local characteristic point and system thereof Download PDFInfo
- Publication number
- CN102426785B CN102426785B CN201110367452XA CN201110367452A CN102426785B CN 102426785 B CN102426785 B CN 102426785B CN 201110367452X A CN201110367452X A CN 201110367452XA CN 201110367452 A CN201110367452 A CN 201110367452A CN 102426785 B CN102426785 B CN 102426785B
- Authority
- CN
- China
- Prior art keywords
- vehicle
- frame
- trail
- car
- unique point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Landscapes
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention, which belongs to the intelligent transport field, discloses a traffic flow information perception method based on a contour and a local characteristic point and a system thereof. In the invention, through remote control, a parameter of an analytic algorithm can be configured. According to a concrete scene, the various kinds of parameters of the analytic algorithm can be set or adjusted at any time, which gives great convenience to use a traffic flow information perception apparatus. In the invention, for tracking a vehicle, rough tracking based on contour information and evaluation are used, and the tracking based on the local characteristic point is selectively performed. A tracking strategy, which is from coarse to minute, can be adaptive to the different traffic scenes and an execution efficiency is high. The traffic flow information perception method and the apparatus in the invention can be directly used in an intelligent transport system (ITS).
Description
Technical field
The invention belongs to intelligent transportation field, relate to based on the transport information influenza of Long-distance Control, video signal and know technology, relate in particular to transport information influenza perception method and the system of the tracking of a kind of location based on the image object contour area, tracking and target local feature point.
Background technology
Along with the develop rapidly of national economy, the mobile vehicle fleet size in the whole nation is increasing, and traffic system also faces great pressure.Intelligent transportation system has obtained development at a high speed in recent years under such background, especially in freeway facility.Intelligent transportation system (Intelligent Transport System, be called for short ITS) be to apply to whole traffic management system by infotech, data communication transmission technology, Electronic transducer technology, electron controls technology and computing machine treatment technology etc. are effectively integrated, and set up a kind of on a large scale in, comprehensive playing a role, in real time, multi-transportation and management system accurately and efficiently.
At present nationwide highway and relevant Freeway Infrastructure, control measures are all very sound.Most highway deploy omnidistance supervisory system, this makes supervision department more comprehensive to the grasp of the car flow information of highway, highway operation situation, has also guaranteed high speed, safety, the operation management economically of highway.
But all there is the problem of a general character in these systems, that is exactly that their detection means mainly relies on annular magnetic test coil, video frequency pick-up head, and transmission cable is the equipment such as microwave detector, radar even.Wherein, annular magnetic test coil is for detection of quantity, size and the speed of a motor vehicle etc. of vehicular traffic, and lay in the whole cross section that generally need to be embedded in underground ，Dui highway, lays cost very high, rapid wear and troublesome maintenance; Video frequency pick-up head is used for taking the real-time traffic scene, and transfers to Surveillance center's demonstration by transmission cable.These expensive equipment are all inapplicable to China's Class I highway, Class II highway, because one-level, Class II highway coverage are large, there is no the economic return source, and lay and safeguard that the cost of checkout equipment and web-transporting device is too large.
In the face of traffic pressure and traditional solution poverty-stricken of weightening finish day by day, Class I highway, Class II highway need effective checkout equipment, the one-tenth of equipment should be as far as possible low simultaneously.In recent years, rise gradually based on the Traffic flow detection method of video, based on the detection method of video, lay on the one hand cost low, be easy to maintenance and repair; In addition, sensing range is large, and a general check point only needs a camera just can cover.On the other hand, by the video image of video analysis Algorithm Analysis camera collection, thereby can obtain telecommunication flow information equally.
In order to address this problem, based on the Traffic flow detection method of video analysis, take full advantage of ripe video processnig algorithms, accurately detect, analyze the motion conditions of each vehicle.Traditional analytical algorithm based on video generally comprises vehicle detection that based on motion detects and based on the vehicle speed detection of virtual detection line, wherein motion detection generally includes background subtraction method and frame-to-frame differences method.But these methods can't adapt to scene complicated and changeable, accuracy of detection is not high.
Summary of the invention
Given this, primary and foremost purpose of the present invention is as a kind of cognitive method of telecommunication flow information based on profile and local unique point is provided, the method is in conjunction with multiple common video analysis algorithm, the thought that employing is learnt from other's strong points to offset one's weaknesses and the strategy of multilevel policy decision, can adjust according to the traffic scene self-adaptation, have higher execution efficiency and robustness.The method specifically comprises the following steps:
The systematic parameter of Remote configuration telecommunication flow information sensory perceptual system, comprise on the region of interest ROI (region of interest, write a Chinese character in simplified form ROI) of video analysis and video image the transformational relation F between space length on pixel distance and real road;
Vision signal is carried out the frame-to-frame differences operation, the frame-to-frame differences image is carried out to shape filtering, and on this basis the position of each car is positioned;
According to the positional information of vehicle, each car is followed the trail of respectively, design point storage organization Memory records the dynamical state of each car;
To following the trail of result, estimate;
According to evaluation result, if follow the trail of the good not unique point of following the trail of failed vehicle of extracting of result, according to the unique point tracing algorithm, unique point is followed the trail of, and according to the state of following the trail of each car in update mode storage organization as a result; If follow the trail of enough states of each car in good direct update mode storage organization of result;
Calculate information of vehicles;
The information of vehicles gathered in a period of time is gathered, generates the transport information stream file, and regularly to server, send the transport information stream file.
Another object of the present invention is to provide a kind of sensory perceptual system of telecommunication flow information based on profile and local unique point, it is characterized in that described system comprises:
Far-end server 2000: the systematic parameter to long-range telecommunication flow information sensory perceptual system is configured, and generates parameter configuration files at node, and the video data gathered after video also will gather outputs to video analytics subsystem 2100;
Video analytics subsystem 2100: by video data is carried out to sequence of operations, calculate information of vehicles, and the information of vehicles gathered in a period of time is gathered, generates the transport information stream file, regularly send the transport information stream file to far-end server;
Wherein, video analytics subsystem 2100 comprises:
Inter-frame difference, morphological filter 2110: vision signal is carried out to the frame-to-frame differences operation, the frame-to-frame differences image is carried out to binaryzation, then carry out shape filtering;
Vehicle ' s contour zone location device 2120: the position to each car on the basis of filtered image positions;
Tracking performance decision device 2140: estimate following the trail of result;
Vehicle local feature point extraction apparatus 2150: extract the unique point of following the trail of failed vehicle;
Beneficial effect: the telecommunication flow information cognitive method of following the trail of based on image object contour area and local unique point of the present invention and system are adjusted the camera posture, are determined the area-of-interest of video analysis by Long-distance Control; Adopt traditional frame difference method and shape filtering to detect the rough position of vehicle, and carry out rough car tracing according to position; By a vital evaluation mechanism, rough tracking result is done to Performance Evaluation after this; The algorithm based on local feature point after optionally carrying out according to assessment result again.The present invention is configured by the parameter of Long-distance Control to analytical algorithm, can according to concrete scene, arrange at any time or adjust the various parameters of analytical algorithm, and this is very easy to the use of telecommunication flow information sensing device.In the present invention, at first the tracking of vehicle is adopted to the rough tracking based on profile information, evaluation, and optionally carry out the tracking based on local feature point, this different traffic scene of pursive strategy energy self-adaptation from coarse to fine, and improved execution efficiency.
The accompanying drawing explanation
Fig. 1 is the cognitive method of the telecommunication flow information based on profile and local unique point process flow diagram of the present invention;
Fig. 2 is the process flow diagram that the present invention realizes the Remote configuration systematic parameter;
Fig. 3 is that the present invention realizes to vision signal is carried out the frame-to-frame differences operation, the frame-to-frame differences image is carried out to shape filtering, and the process flow diagram on this basis position of each car positioned;
Fig. 4 is that the present invention realizes the positional information according to vehicle, and each car is followed the trail of respectively, and the design point storage organization records the process flow diagram of the dynamical state of each car;
Fig. 5 is that the present invention realizes estimating following the trail of result, and according to the process flow diagram of the state of each car in evaluation result update mode storage organization;
Fig. 6 is that the present invention realizes the information of vehicles gathered in a period of time is gathered, generates the transport information stream file, and regularly to server, sends the process flow diagram of transport information stream file;
Fig. 7 is the block scheme of the telecommunication flow information sensory perceptual system of following the trail of based on image object contour area and local unique point of the present invention, and in figure, each mark meaning is as follows: far-end server 2000, video analytics subsystem 2100, inter-frame difference, morphological filter 2110, vehicle ' s contour zone location device 2120, the tracker 2130 based on the vehicle ' s contour zone, tracking performance decision device 2140, vehicle local feature point extraction apparatus 2150, based on the tracker 2160 of vehicle local feature.
Embodiment
As shown in Figure 1, the telecommunication flow information cognitive method based on profile and local unique point comprises the steps:
The systematic parameter of Remote configuration telecommunication flow information sensory perceptual system, comprise on the region of interest ROI of video analysis and video image the transformational relation F between space length on pixel distance and real road;
Vision signal is carried out the frame-to-frame differences operation, the frame-to-frame differences image is carried out to binaryzation, then carry out shape filtering, and on this basis the position of each car is positioned;
According to the positional information of vehicle, each car is followed the trail of respectively, design point storage organization Memory records the dynamical state of each car;
To following the trail of result, estimate;
According to evaluation result, if follow the trail of the good not unique point of following the trail of failed vehicle of extracting of result, according to the unique point tracing algorithm, unique point is followed the trail of, and according to the state of following the trail of each car in update mode storage organization as a result; If follow the trail of enough states of each car in good direct update mode storage organization of result;
The information of vehicles that comprises the information such as vehicle size, the speed of a motor vehicle according to calculating such as following the trail of the information that vehicle obtains, the systematic parameter of Remote configuration;
The information of vehicles gathered in a period of time is gathered, generates the transport information stream file, and regularly to server, send the transport information stream file.
The realization flow of Remote configuration systematic parameter as shown in Figure 2, comprising:
Step 101: adjust the posture of video signal collective camera by Long-distance Control, make the observable zone reach best;
Step 102: after the fixing camera posture, the telecommunication flow information sensory perceptual system sends a video pictures to server end;
Step 103: at far-end server, determine the region of interest ROI of video analysis by the mode of manpower intervention;
Step 104: at far-end server, calculate on video image the transformational relation F between space length on pixel distance and real road by the mode of manpower intervention;
Step 105: the systematic parameter using ROI information and transformational relation F as perception and apparatus system is sent to telecommunication flow information perception and apparatus system.
The present invention realizes to vision signal is carried out the frame-to-frame differences operation, the frame-to-frame differences image is carried out to shape filtering, and the flow process on this basis position of each car positioned is as shown in Figure 3, comprising:
Step 201: video analytic system is from obtaining image sequence vision signal, image sequence is made to following frame-to-frame differences to be operated: dI (t)=| I (t)-I (t-1) |, the Time labeling of t representative graph picture frame wherein, I (t) means t image constantly, it is present image, I (t-1) is the former frame image, and dI (t) is the frame-to-frame differences image, is specially the image that the poor absolute value of two frames forms;
Step 202: the frame-to-frame differences result is carried out to binaryzation: with the threshold value T of image binaryzation
1=20 is example, if dI (t, i, j)>20, dI (t, i, j) ← 0, otherwise dI (t, i, j) ← 255, wherein, i presentation video pixel line number, j presentation video pixel column number, I (t, i, j) mean the pixel value of t time chart picture at position (i, j), dI (t, i, j) be the t pixel value of frame-to-frame differences image on position (i, j) constantly;
Step 203: on 202 basis as a result, dI (t) is carried out to shape filtering, i.e. traditional opening operation and closed operation;
Step 204: on 203 basis as a result, determine the position of each car, the steps include:
(1) in 203 results, finding pixel value is 0 connected region, and goes out each together with zone with circumscribed rectangle frame, and these rectangles are designated as to { S
1..., S
n, n is the number of connected region;
(2) with threshold value T
2=10, T
2'=100 are example, remove { S
1..., S
nIn area be less than 100 and be greater than 10 rectangle, that is, think that rectangle size corresponding to vehicle is moderate, remove excessive or too small rectangle;
(3) with threshold value T
3=10 is example, on the basis of (2) will between distance be less than two rectangles of 10 and merging into same rectangle, that is, think that the distance between vehicle is larger, two rectangles that lean on closelyer in fact corresponding same car;
The rectangle that (4) will finally stay is as the domain of the existence { V of vehicle
1..., V
k, k is the quantity of the vehicle that detects.
According to the positional information of vehicle, each car is followed the trail of respectively, the design point storage organization record each car dynamical state flow process as shown in Figure 4, comprise the steps:
Step 301: set up state storage organization Memory, for example in Memory, comprise 50 mnemons, be designated as { mem_cell_1, ..., mem_cell_50}, storage unit, for being stored in the dynamical state of the car in the ROI scope, comprises the position in picture of vehicle, the records such as size in picture, thinks that here the number that is in the vehicle in the ROI scope is less than 50;
Step 302: at first expect according to historic state prediction vehicle in present frame of the vehicle in each mnemon the position occurred, for example adopt linear prediction method: the two frame historical position (10 of supposing a car of record in mem_cell_1,20) and (11,30), the position that can calculate roughly this vehicle of present frame of the method by linear prediction is (12,40);
Step 303: the position of the vehicle detected in present frame is { V
1..., V
K (t); k (t) means the number of the vehicle detected in present frame; the predicted position of vehicle in each mnemon of these zones and Memory is mated; for example according to Euclidean distance, mate: supposing has vehicle registration in current mem_cell_1 and mem_cell_2; and predicted position is respectively (12; 40) and (22,70), the position of the vehicle that current detection arrives is (11; 42) and (22; 68), according to Euclidean distance, can judge (11,42) and (12; 40) coupling; and (22,68) are complementary with (22,70);
Step 304. is according to matching result, the information of vehicles that in its present frame, the match is successful is added in mnemon corresponding in Memory and goes, as in step 303, by (11,42) and (22,68) add to respectively in mem_cell_1 and mem_cell_2;
Fig. 5 is that the present invention realizes estimating following the trail of result, and according to evaluation result, if follow the trail of the good not unique point of following the trail of failed vehicle of extracting of result, according to the unique point tracing algorithm, unique point is followed the trail of, and according to the state of following the trail of each car in update mode storage organization as a result, if follow the trail of enough process flow diagrams of the state of each car in good direct update mode storage organization of result, comprise the steps:
It is k (t) that step 401. is worked as the vehicle in front number, supposes that the storage unit number of record is arranged in Memory be M ', M '<M, and wherein M is the storage unit number in Memory.If M '≤k (t), be evaluated as " enough good ", otherwise be evaluated as " good not ";
If step 402. M '<k (t), meaning now has new vehicle or noise to appear in ROI, emerging vehicle (or noise) is added in a new storage unit in Memory;
If step 403. M '=k (t), mean that in ROI, the vehicle number does not change, and each car all tracked on;
If step 404. M '>k (t), the vehicle that expressed possibility does not detect, or has vehicle to leave the ROI zone.This does further judgement by following step 410.
If it is good not that step 405. is followed the trail of result, extract Harris (Chris Harris has proposed the method for angle point in image in 1988, claim that this angular-point detection method the is the Harris angle point) unique point { P in the ROI zone of present frame
1... P
N (t), N (t) is the number of unique point in ROI;
Step 406: extract the Harris unique point { Pf that follows the trail of failed vehicle in Memory
1 1..., Pf
N1 1..., { Pf
1 L..., Pf
NL L, wherein L means to follow the trail of the number of failed vehicle, { Pf
1 i..., Pf
Ni iMean that i is followed the trail of the failed unique point of vehicle in previous frame;
Step 407: utilize unique point tracing algorithm KLT (Kanade-Lucas-Tomasi) algorithm, at { P
1... P
N (t)Middle unique point of following the trail of the vehicle that each tracking is failed;
Step 408: for { Pf
1 i..., Pf
Ni i, the number of establishing the unique point successfully tracked in step 407 is s
i, s
i≤ N
1, with threshold value T
4=0.6 is example, if s
i/ N
i>0.6, think that i car followed the trail of successfully by local feature point tracking method in Memory, otherwise follow the trail of unsuccessfully;
Step 409: according to following the trail of the unique point of successful vehicle in step 408, and the change in location situation of the unique point be complementary with it, infer the motion conditions of this vehicle, construct on this basis the state of vehicle in present frame, and the structure result is added in the corresponding stored unit in Memory;
Step 410: up to the present, to above two kinds of vehicles that method for tracing all lost efficacy in Memory, if the predicted position of vehicle is in the boundary vicinity of ROI, judge that this vehicle left the ROI zone, simultaneously this information of vehicles is moved to cache file; Otherwise think and have vehicle to fail correctly to track;
On this basis, the information according to following the trail of the information that vehicle obtains, the systematic parameter of Remote configuration etc. and calculate vehicle, comprise and comprise the steps: vehicle size, the speed of a motor vehicle etc.
According to size and the picture position obtained and the transformation relation F between physical location of the rectangular area of the vehicle recorded in cache file, calculate the actual size of vehicle in abovementioned steps 104;
According to the situation of change of the vehicle recorded in cache file position in ROI, the transformation relation F between combining image position and physical location, calculate the movement velocity of vehicle.
The present invention realizes the information of vehicles gathered in a period of time is gathered, generates the transport information stream file, and regularly to the flow process of server transmission transport information stream file as shown in Figure 6, comprising:
As shown in Figure 7, this system comprises telecommunication flow information sensory perceptual system block diagram based on image object contour area and the tracking of local unique point of the present invention
Far-end server 2000: the systematic parameter to long-range telecommunication flow information sensory perceptual system is configured, and generates parameter configuration files at node, and the video data gathered after video also will gather outputs to video analytics subsystem 2100;
Video analytics subsystem 2100: by video data being carried out to the operations such as inter-frame difference, shape filtering, vehicle ' s contour location, tracking performance judgement, the extraction of vehicle local feature and the tracking of vehicle local feature, calculate information of vehicles, comprise vehicle size, the speed of a motor vehicle etc., and the information of vehicles gathered in a period of time is gathered, generates the transport information stream file, regularly to far-end server, send the transport information stream file;
Wherein, video analytics subsystem 2100 comprises:
Inter-frame difference, morphological filter 2110: vision signal is carried out to the frame-to-frame differences operation, the frame-to-frame differences image is carried out to binaryzation, then carry out shape filtering;
Vehicle ' s contour zone location device 2120: the position to each car on the basis of filtered image positions;
Tracking performance decision device 2140: estimate following the trail of result;
Vehicle local feature point extraction apparatus 2150: extract the unique point of following the trail of failed vehicle;
Above content is in conjunction with optimum implementation further description made for the present invention, can not assert that specific embodiment of the invention is confined to these explanations.It should be appreciated by those skilled in the art, in the situation that do not break away from the spirit and scope of the present invention that limited by appended claims, can also carry out the various modifications on form and details to it, all should be considered as belonging to protection scope of the present invention.
Claims (9)
1. cognitive method of the telecommunication flow information based on profile and local unique point is characterized in that comprising the following steps:
The systematic parameter of Remote configuration telecommunication flow information sensory perceptual system, comprise on the region of interest ROI of video analysis and video image the transformational relation F between space length on pixel distance and real road;
Vision signal is carried out the frame-to-frame differences operation, the frame-to-frame differences image is carried out to binaryzation, then carry out shape filtering, and on this basis the position of each car is positioned;
According to the positional information of vehicle, each car is followed the trail of respectively, design point storage organization Memory records the dynamical state of each car;
To following the trail of result, estimate;
According to evaluation result, if follow the trail of the good not unique point of following the trail of failed vehicle of extracting of result, according to the unique point tracing algorithm, unique point is followed the trail of, and according to the state of following the trail of each car in update mode storage organization as a result; If follow the trail of enough states of each car in good direct update mode storage organization of result;
Calculate information of vehicles;
The information of vehicles gathered in a period of time is gathered, generates the transport information stream file, and regularly to server, send the transport information stream file.
2. method according to claim 1, is characterized in that, the method for the systematic parameter of Remote configuration telecommunication flow information sensory perceptual system is:
101: by Long-distance Control, adjust the posture of video signal collective camera, make the observable zone reach best;
102: after the fixing camera posture, system sends a video pictures to server end;
103: at far-end server, by the mode of manpower intervention, determine the region of interest ROI of video analysis;
104: at far-end server, calculate on video image the transformational relation F between space length on pixel distance and real road by the mode of manpower intervention;
105: ROI information and transformational relation F are sent to system as systematic parameter.
3. method according to claim 1, is characterized in that, vision signal is carried out the frame-to-frame differences operation, the frame-to-frame differences image is carried out to binaryzation, then carries out shape filtering, and the method on this basis position of each car positioned is:
201: from vision signal, obtaining image sequence, image sequence is made to following frame-to-frame differences to be operated: dI (t)=| I (t) – I (t-1) |, the Time labeling of t representative graph picture frame wherein, I (t) means t image constantly, it is present image, I (t-1) is the former frame image, and dI (t) is the frame-to-frame differences image, is specially the image of the absolute value formation of two frame-to-frame differencess;
202: the frame-to-frame differences result is carried out to binaryzation: if dI is (t; I, j)>T
1,
DI (t; I, j) ← 0, otherwise dI (t; I, j) ← 255, wherein, i presentation video pixel line number, j presentation video pixel column number, I (t; I, j) mean the pixel value of t time chart picture at position (i, j), dI (t; I, j) be the t pixel value of frame-to-frame differences image on position (i, j) constantly; T
1Threshold value for image binaryzation;
203: on 202 basis as a result, frame-to-frame differences image dI (t) is carried out to shape filtering, i.e. traditional opening operation and closed operation;
204: on 203 basis as a result, determine the position of each car, the steps include:
(1) in 203 results, finding pixel value is 0 connected region, and goes out each connected region with circumscribed rectangle frame, and these rectangles are designated as to { S
1..., S
n, n is the number of connected region;
(2) remove { S
1..., S
nIn area be less than T
2' and be greater than T
2Rectangle, T wherein
2,T
2' be the area threshold of the rectangular area of setting, T
2' larger, T
2Less
(3) distance is less than T between on the basis of (2), inciting somebody to action
3Two rectangles merge into same rectangle, wherein T
3Distance threshold for two centers, rectangular area that arrange;
The rectangle that (4) will finally stay is the position { V of vehicle as the domain of the existence of vehicle
1..., V
k, k is the quantity of the vehicle that detects.
4. method according to claim 1, is characterized in that, according to the positional information of vehicle, each car followed the trail of respectively, and the method that the design point storage organization records the dynamical state of each car is:
301: set up state storage organization Memory, in Memory, comprise M storage unit, be designated as { mem_cell_1;, mem_cell_M}, storage unit is for being stored in the dynamical state of the car in the ROI scope; the M here is constant, and is greater than the number that is in the vehicle in the ROI scope;
302: at first according to historic state prediction vehicle in present frame of the vehicle in each storage unit, expect the position occurred;
303: the position of the vehicle detected in present frame is { V
1..., V
K (t), k (t) means the quantity of the vehicle detected in present frame, and the predicted position of vehicle in each storage unit of these zones and Memory is mated;
304. according to matching result, the information of vehicles that in its present frame, the match is successful is added in storage unit corresponding in Memory and goes.
5. method according to claim 1, is characterized in that, the method that the tracking result is estimated is:
401. current vehicle fleet size is k (t), supposes that it is M ' that the storage unit number of record is arranged in Memory, M '<M, M is the contained storage unit number of Memory, if M '≤k (t), be evaluated as " enough good ", otherwise be evaluated as " good not ";
If 402. M '<k (t), meaning now has new vehicle or noise to appear in ROI, emerging vehicle is added in a new storage unit in Memory;
403. if M '=k (t) means that in ROI, the vehicle number does not change, and each car all tracked on;
If 404. M '>k (t), the vehicle that expressed possibility does not detect, or there is vehicle to leave the ROI zone.
6. method according to claim 1, it is characterized in that, according to evaluation result, if it is good not to follow the trail of result, extract the unique point of the vehicle that in Memory, tracking is failed, utilize unique point tracing algorithm tracking feature point in the ROI of present frame scope, and according to the state of following the trail of each car in update mode storage organization as a result, its method is:
405: the Harris unique point { P in the ROI zone of extraction present frame
1... P
N (t), N (t) is the number of unique point in ROI;
406: extract the Harris unique point { Pf that follows the trail of failed vehicle in Memory
1 1..., Pf
N1 1..., { Pf
1 L..., Pf
NL L, { Pf
1 i..., Pf
Ni iMean that i is followed the trail of the failed unique point of vehicle in previous frame, i=1 wherein ... L, L mean to follow the trail of the number of failed vehicle;
407: utilize unique point tracing algorithm KLT algorithm, at { P
1... P
N (t)Middle unique point of following the trail of the vehicle that each tracking is failed;
408: for { Pf
1 i..., P
FNi
i, the number of establishing the unique point successfully tracked in 407 is s
i, s
i≤ Ni, if s
i/ Ni>T
4Think that i car followed the trail of successfully by local feature point tracking method in Memory, otherwise follow the trail of unsuccessfully, wherein Ni is the unique point number of i car, T
4For the unique point that successfully tracks and the ratio of unique point number;
409: according in step 408, following the trail of the unique point of successful vehicle, and the change in location situation of the unique point be complementary with it, infer the motion conditions of this vehicle, construct on this basis the state of vehicle in present frame, and the structure result is added in the corresponding stored unit in Memory;
410: up to the present, the vehicle that above-mentioned method for tracing in Memory was all lost efficacy, if the predicted position of vehicle is in the boundary vicinity of ROI, judge that this vehicle left the ROI zone, simultaneously this information of vehicles moved to cache file; Otherwise think and have vehicle to fail correctly to track.
7. method according to claim 1, is characterized in that, the method for calculating information of vehicles is:
According to the size of the rectangular area of the vehicle recorded in cache file and the transformation relation F between picture position and physical location, calculate the actual size of vehicle;
According to the situation of change of the vehicle recorded in cache file position in ROI, the transformation relation F between combining image position and physical location, calculate the movement velocity of vehicle.
8. method according to claim 1, it is characterized in that, system arranges, adds additional information to the information of vehicles cache file in a period of time and forms the transport information stream file in a period of time, and timing to the process that server sends the transport information stream file is:
501. the information of vehicles cache file is arranged, and adds additional information, form the transport information stream file;
502. by network, this transport information stream file is sent to server end.
9. sensory perceptual system of the telecommunication flow information based on profile and local unique point is characterized in that described system comprises:
Far-end server (2000): the systematic parameter to long-range telecommunication flow information sensory perceptual system is configured, and generates parameter configuration files at node, and the video data gathered after video also will gather outputs to video analytics subsystem (2100);
Video analytics subsystem (2100): by video data is carried out to sequence of operations, calculate information of vehicles, and the information of vehicles gathered in a period of time is gathered, generates the transport information stream file, regularly send the transport information stream file to far-end server;
Wherein, video analytics subsystem (2100) comprising:
Inter-frame difference, morphological filter (2110): vision signal is carried out to the frame-to-frame differences operation, to the frame-to-frame differences image binaryzation, then carry out shape filtering;
Vehicle ' s contour zone location device (2120): the position to each car on the basis of filtered image positions;
Tracker (2130) based on the vehicle ' s contour zone: according to the positional information of vehicle, each car is followed the trail of respectively, the design point storage organization records the dynamical state of each car;
Tracking performance decision device (2140): estimate following the trail of result;
Vehicle local feature point extraction apparatus (2150): extract the unique point of following the trail of failed vehicle;
Tracker (2160) based on the vehicle local feature: according to the unique point tracing algorithm, unique point is followed the trail of, and according to the state of following the trail of each car in update mode storage organization as a result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110367452XA CN102426785B (en) | 2011-11-18 | 2011-11-18 | Traffic flow information perception method based on contour and local characteristic point and system thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110367452XA CN102426785B (en) | 2011-11-18 | 2011-11-18 | Traffic flow information perception method based on contour and local characteristic point and system thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102426785A CN102426785A (en) | 2012-04-25 |
| CN102426785B true CN102426785B (en) | 2013-11-27 |
Family
ID=45960762
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201110367452XA Expired - Fee Related CN102426785B (en) | 2011-11-18 | 2011-11-18 | Traffic flow information perception method based on contour and local characteristic point and system thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102426785B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103093528A (en) * | 2012-12-28 | 2013-05-08 | 中山大学 | Intelligent system for monitoring vehicles and people in community |
| CN103297754B (en) * | 2013-05-02 | 2016-06-01 | 上海交通大学 | A kind of monitor video self-adaptation encoding region of interest system |
| CN103295015B (en) * | 2013-06-25 | 2016-02-03 | 中国人民解放军国防科学技术大学 | The local feature point extracting method of partial occlusion target |
| CN104702982B (en) * | 2013-12-20 | 2018-01-05 | 杭州海康威视数字技术股份有限公司 | A kind of algorithm configuration method and system of video analysis equipment |
| CN106446790A (en) * | 2016-08-30 | 2017-02-22 | 上海交通大学 | Method for tracking and analyzing traffic video flow of fixed camera |
| CN107909032A (en) * | 2017-11-15 | 2018-04-13 | 重庆邮电大学 | A kind of behavioral value and recognition methods based on single sample |
| CN108758728B (en) * | 2018-03-29 | 2023-02-17 | 青岛海尔智能技术研发有限公司 | Head gesture-based range hood control method and range hood |
| CN108571928A (en) * | 2018-04-13 | 2018-09-25 | 桂林电子科技大学 | A kind of intermediate plate anchorage dimensional defects detection method based on machine vision |
| CN108337486A (en) * | 2018-04-19 | 2018-07-27 | 北京软通智城科技有限公司 | A kind of device and method of the video analysis of the algorithm configuration based on scene |
| CN110718061B (en) * | 2019-10-17 | 2021-09-28 | 长沙理工大学 | Traffic intersection vehicle flow statistical method and device, storage medium and electronic equipment |
| CN111126261B (en) * | 2019-12-23 | 2023-05-26 | 珠海深圳清华大学研究院创新中心 | Video data analysis method and device, raspberry group device and readable storage medium |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101308607A (en) * | 2008-06-25 | 2008-11-19 | 河海大学 | Moving target tracking method by multiple features integration under traffic environment based on video |
| CN101800890B (en) * | 2010-04-08 | 2013-04-24 | 北京航空航天大学 | Multiple vehicle video tracking method in expressway monitoring scene |
| CN102034355A (en) * | 2010-12-28 | 2011-04-27 | 丁天 | Feature point matching-based vehicle detecting and tracking method |
-
2011
- 2011-11-18 CN CN201110367452XA patent/CN102426785B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN102426785A (en) | 2012-04-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102426785B (en) | Traffic flow information perception method based on contour and local characteristic point and system thereof | |
| Mukojima et al. | Moving camera background-subtraction for obstacle detection on railway tracks | |
| CN105744232B (en) | A kind of method of the transmission line of electricity video external force damage prevention of Behavior-based control analytical technology | |
| Sina et al. | Vehicle counting and speed measurement using headlight detection | |
| CN104200657B (en) | A kind of traffic flow parameter acquisition method based on video and sensor | |
| CN102622886B (en) | Video-based method for detecting violation lane-changing incident of vehicle | |
| CN102324183B (en) | Method for detecting and shooting vehicle based on composite virtual coil | |
| CN101510356B (en) | Video detection system and data processing device thereof, video detection method | |
| CN107301776A (en) | Track road conditions processing and dissemination method based on video detection technology | |
| CN102393901B (en) | Traffic flow information perception method based on hybrid characteristic and system thereof | |
| CN103605967A (en) | Subway fare evasion prevention system and working method thereof based on image recognition | |
| CN102496281B (en) | Vehicle red-light violation detection method based on combination of tracking and virtual loop | |
| CN103208190B (en) | Traffic flow detection method based on object detection | |
| Pan et al. | Traffic surveillance system for vehicle flow detection | |
| CN104134222A (en) | Traffic flow monitoring image detecting and tracking system and method based on multi-feature fusion | |
| CN104378582A (en) | Intelligent video analysis system and method based on PTZ video camera cruising | |
| CN103077423A (en) | Crowd quantity estimating, local crowd clustering state and crowd running state detection method based on video stream | |
| CN103280052B (en) | Be applied to the intrusion detection method of long distance track circuit intelligent video monitoring | |
| CN102254394A (en) | Antitheft monitoring method for poles and towers in power transmission line based on video difference analysis | |
| CN104159088A (en) | System and method of remote monitoring of intelligent vehicle | |
| CN103794050A (en) | Real-time transport vehicle detecting and tracking method | |
| CN105574502A (en) | Automatic detection method for violation behaviors of self-service card sender | |
| CN205680289U (en) | A kind of parking detecting system | |
| CN204305214U (en) | Loading goods train state high-definition image system | |
| WO2018209470A1 (en) | License plate identification method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20131127 Termination date: 20161118 |