CN106875424A - A kind of urban environment driving vehicle Activity recognition method based on machine vision - Google Patents
A kind of urban environment driving vehicle Activity recognition method based on machine vision Download PDFInfo
- Publication number
- CN106875424A CN106875424A CN201710027523.9A CN201710027523A CN106875424A CN 106875424 A CN106875424 A CN 106875424A CN 201710027523 A CN201710027523 A CN 201710027523A CN 106875424 A CN106875424 A CN 106875424A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- track
- image
- sample
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a kind of urban environment driving vehicle Activity recognition method based on machine vision, target detection, target following, feature extraction, Activity recognition are carried out by moving vehicle, draw the result of vehicle behavioural analysis.Vehicle target detection method first by background subtraction completes vehicle target detection;Then the vehicle target for detecting is tracked using optical flow method;After the completion of tracking, the basic act feature according to vehicle movement track draws the track of vehicle;Finally it is identified to exercising track using the SVM classifier for training, so as to judge that the behavior is to turn left, turn right or keep straight on.The inventive method performs speed soon, and accuracy rate is high, it is possible to achieve vehicle behavior is accurately identified, for the Vehicular real time monitoring of intelligent traffic monitoring system.
Description
Technical field
The invention belongs to technical field of visual navigation, and in particular to a kind of urban environment driving vehicle row based on machine vision
It is recognition methods.
Background technology
Currently, machine vision in artificial intelligence field development speed quickly, and in theoretical scientific and engineering application aspect
Extensive application prospect, the research of NI Vision Builder for Automated Inspection is several including target detection, image characteristics extraction and Activity recognition etc.
Key issue, and at medical video, image retrieval, multimedia signal processing and communication, fingerprint and recognition of face, image
Reason and each research field such as pretreatment, nature biotechnology category identification, traffic safety are all used widely.
Moving object detection is both a core technology in NI Vision Builder for Automated Inspection, is again image procossing, multimedia messages
Indispensable part in the every field such as treatment, intelligent video monitoring.Existed in the scene of various very complicateds various
Different information, but only partial information be people it is interested be it is effective, effective information and complex background success
Ground Split is opened and people's target interested is namely only extracted from background, here it is the basic task of moving object detection.From
The essential characteristics such as contour edge, the internal information of moving target can be at a glance observed in testing result, is conducive to spy
Extraction, the carrying out of Activity recognition work are levied, Research Significance is great.
The Activity recognition of moving target then contains the timely detection and feature extraction, behavior description, analysis and knowledge of target
Not etc..Monitoring device is installed in the public place such as factory, enterprise, market, station, airport, cell, is mostly to be with movement human
Research object, realize being monitored moving target and behavioural analysis, first detect target and moved according to the behavior for extracting
That target is analyzed as feature the behavior such as walks, runs, fighting, gathering, stealing.
At present, intelligent monitoring technology exists in terms of the robustness of target detection, feature extraction and Activity recognition method
Deficiency, the scope of application is restricted, and is not reaching to Activity recognition rate higher, so, for moving target feature extraction with
Activity recognition research is always the hot issue in intelligent safety system.The friendship of the environment of road traffic, especially intersection
Logical environment is extremely complex, but most of vehicle traffic accident all occurs at this, if intersection etc. can effectively be monitored
Road, if the act of violating regulations of vehicle can be automatically detected, can thus reduce the generation of accident in traffic route.
The content of the invention
The present invention provides a kind of urban environment driving vehicle Activity recognition method based on machine vision, by sport(s) car
Target detection, target following, feature extraction, Activity recognition are carried out, draw the result of vehicle behavioural analysis.First by background
The vehicle target detection method of calculus of finite differences completes vehicle target detection;Then the vehicle target for detecting is carried out using optical flow method
Tracking;After the completion of tracking, the basic act feature according to vehicle movement track draws the track of vehicle;Finally use what is trained
SVM classifier is identified to exercising track, so as to judge that the behavior is to turn left, turn right or keep straight on.
A kind of urban environment driving vehicle Activity recognition method based on machine vision, it is characterised in that step is as follows:
Step 1:Vehicle target is detected and tracking:The method modeled with mixed Gauss model using background subtraction is transported
Motor-car target detection, is then tracked, specially using the track algorithm of light stream to the moving vehicle target for detecting:
Step a:Respectively according toWithMeter
Calculate the mean μ of the pixel intensity of video sequence image in a period of time0(x, y) and varianceWith μ0(x, y) andThe respectively image B of pixel average and variance composition with Gaussian Profile0, B0As initial background estimating image;
Wherein, N is the totalframes of sequence image in initial background image access time section, 150≤N≤200;fi(x,
Y) it is that location of pixels of i-th two field picture in xth row, the pixel brightness value of y row, (x, y) expression image is x rows, y row;
Step b:Respectively according to μj(x, y)=(1- α) μj-1(x,y)+α·fj(x, y) andUpdate the mean μ of background estimating imagej(x, y) and side
DifferenceThe background estimating image B of the jth two field picture after being updatedj;
Wherein,δ is the constant between [0,1], and K is mixing
The number of Gauss model, 3≤K≤5;J >=1, fj(x, y) represents the pixel brightness value that jth two field picture is arranged in xth row, y;
Step c:According to dj(x, y)=| fj(x,y)-Bj(x, y) | it is calculated current frame image and present frame background estimating
The difference image of image, and according toBinary conversion treatment is carried out to difference image, is detected
The moving vehicle region for going out, i.e., pixel value is 1 region in image M after binary conversion treatment, and pixel value is 0 region in image M
It is background area;M in jth two field picturejThe moving region that (x, y)=1 represents;
Wherein, r is gray threshold, 50≤r≤60;
Step d:Corner Feature extraction is carried out to the moving vehicle region that each two field picture is detected, pyramid is reapplied
Lucas-Kanade sparse optical flows algorithm is tracked to the angle point in all frame video images, obtains moving vehicle motion rail
Mark;
Step 2:Track of vehicle feature extraction:The track characteristic extraction side being combined using matrix grid and two-way histogram
Method, constructs track of vehicle characteristic vector, for vehicle behavior classification provides feature foundation, specially:
Step a:On the basis of the moving vehicle trajectory coordinates obtained by step 1, with track abscissa as x-axis, track is vertical sits
Be designated as y-axis construction track of vehicle coordinate system Oxy, obtain respectively x-axis direction maximum x a littlemaxWith minimum value xmin, y-axis
Direction maximum ymaxWith minimum value ymin, create a width of in coordinate system OxyIt is a height of's
Matrix grid, initial value 0 is assigned to each sub-grid in grid;
Step b:Assignment is carried out to the sub-grid where trajectory coordinates, first along x-axis positive direction, then along y-axis positive direction, by suitable
Sequence assign incremental weighted value 1,2,3 ..., obtain trajectory coordinates matrix;
Step c:Row and column to trajectory coordinates matrix constructs histogram respectively, and vehicle target is estimated using two-way histogram
Basic act trend, i.e., according to trajectory coordinates matrix row and column histogram overall trend move towards, if successively decreased from left to right,
Then it is intended to turn right;If be incremented by from left to right, it is intended to turn left;If centre is successively decreased to both sides, trajectory is in
Straight trip trend, obtains moving vehicle behavior trend to turn left, turning right or keep straight on;
Wherein, histogram abscissa represents the line number or columns of trajectory coordinates matrix, and histogrammic ordinate counts track
Matrix element value is not 0 number in a certain row or column in coordinates matrix;
Step d:The trajectory coordinates matrix that the behavior trend and step b obtained by step c are obtained constitutes the rail of moving vehicle
Mark characteristic vector;
Step 3:Vehicle Behavioral training:Behavioral training is carried out to track of vehicle sample using SVM two layers of classified device structures, is obtained
To the recognition result of sample data, specially:
Step a:Create track sample:The simulation background of the video of collection in advance is imported with picture Core Generator, and is created
Including straight trip, turn left, each 100 of the track sample of right-hand bend three types;
Step b:Track characteristic extracting method according to step 2 carries out track characteristic extraction to track sample, obtains various
The track characteristic vector of type track sample;
Step c:The track characteristic vector of the sample obtained to step b using SVM two layers of classified device is trained, and obtains sample
The recognition result of notebook data;
Step 4:Vehicle Activity recognition:According to the SVM classifier that step 3 training is obtained, the moving vehicle obtained to step 2
Track characteristic vector carries out vehicle Activity recognition, finally gives vehicle behavior to turn left, turning right or keep straight on.
The beneficial effects of the invention are as follows:Vehicle is carried out due to carrying out vehicle target detection and optical flow method using background subtraction
Tracking, method performs speed soon, and accuracy rate is high;Because the track characteristic being combined using matrix grid and two-way histogram is extracted
Method, avoids redundancy condition to a certain extent, effectively extracts and expanded vehicle movement track characteristic vector, can be with
Avoid producing inseparable regional issue.
Brief description of the drawings
Fig. 1 is the basic flow sheet of the urban environment driving vehicle Activity recognition method based on machine vision of the invention
Fig. 2 is the result figure that the inventive method carries out vehicle target detection with background subtraction
Fig. 3 is the vehicle movement track schematic diagram that the inventive method is extracted
Specific embodiment
The present invention is further described with reference to the accompanying drawings and examples, and the present invention includes but are not limited to following implementations
Example.
The invention provides a kind of urban environment driving vehicle Activity recognition method based on machine vision, its basic procedure
Figure is as shown in figure 1, specifically include following steps:
Step 1:Vehicle target is detected and tracking
Vehicle target detection refers to that vehicle target is isolated from video sequence image, and the testing result of vehicle target is direct
The links such as the vehicle target tracking of influence later stage, behavioural characteristic extraction and behavior classification.Vehicle target tracking is to obtain car
The kinematic parameter (such as position, speed etc.) and movement locus of target, realize the behavior understanding to vehicle target.In traffic
In monitor video, the factor such as weather, vehicle target are interfered is the main original for influenceing vehicle target detect and track precision
Cause.
Regarding to the issue above, first, the method for being modeled using background subtraction and mixed Gauss model carries out target detection
With extraction, foreground image and background model are accurately distinguished from video image, extract foreground image and detect vehicle mesh
Mark.Then, moving vehicle is tracked using the track algorithm of light stream.
1st, the method modeled using background subtraction and mixed Gauss model is carried out target detection and extracted:
Background subtraction is a kind of most basic target identification method, and it is updated with reference to figure using according to certain background model
Picture, calculates the difference image of present image and reference picture, and then threshold values is partitioned into moving object, and this method calculates letter
Single, if reference picture chooses proper, the advantage of this method can be to be partitioned into moving object exactly.Implementation step is such as
Under:
(1) determine background model, and set up background image.Simplest background model is time the average image, but with
The passage of time, extraneous light can change, and this can cause the change of background image, thus using a width static background image
Method, is only suitable for being applied to the preferable occasion of external condition.In order to realize prolonged video monitor, the present invention is using based on height
The background image algorithm for estimating of this statistical model, the probability density distribution of each pixel color is described with Gaussian Profile.The calculation
Method is made up of the estimation and renewal two parts of background image.In the algorithm for estimating of background image, first, calculate in a period of time
The mean μ of the pixel intensity of video sequence image0(x, y) and varianceWith μ0(x, y) andComposition has height
The image B of this distribution0, B0As initial background estimating image;
Wherein,N is initialization
The totalframes of sequence image, 150≤N≤200 in background image access time section;fi(x, y) is the i-th two field picture in xth row, y
The pixel brightness value of row, (x, y) represents that the location of pixels in image is x rows, y row.
After the completion of the initialization of background estimating image, with the arrival of each frame new images, respectively according to μj(x, y)=
(1-α)·μj-1(x,y)+α·fj(x, y) andUpdate the back of the body
Scape estimates the mean μ of imagej(x, y) and varianceThe background estimating image B of the jth two field picture after being updatedj;
Wherein,δ is normal between one given [0,1]
Number, K is the number of mixed Gauss model, 3≤K≤5;J >=1, fj(x, y) represents that jth two field picture is bright in the pixel that xth row, y are arranged
Angle value.
(2) under voxel model, oneself is subtracted with present image and knows background image to obtain difference image, i.e., according to dj(x,
Y)=| fj(x, y)-Bj(x, y) | it is calculated the difference image of current frame image and present frame background estimating image;Then, it is right
Binary conversion treatment is done in difference image, the moving vehicle region for being detected, i.e.,:
Wherein, Mj(x, y) is any point in difference image, and r is Gray-scale value, 50≤r≤60.
If, Mj(x, y)=1, then it represents that pixel (x, y) belongs to moving region in jth frame, otherwise, pixel (x, y)
Belong to background area in jth frame.
2nd, moving vehicle is tracked using the track algorithm of light stream:
The method of optical flow field is that optical flow field is extracted from the image sequence of Real-time Collection, and hoof selects the larger motion mesh of light stream
Mark region and calculate the velocity of moving target, so as to realize the detect and track of moving target.Optical flow field is one two
N dimensional vector n, the information that it is included is the transient motion velocity information of each pixel.Study optical flow field purpose be exactly
For the sports ground that the approximate calculation from image sequence is not directly available.Any algorithm for estimating can obtain light stream, such as office
Portion loose her algorithm, multi-scale estimation algorithm and Hierarchical block matching algorithm etc..
Classical Horn-Schunck optical flow computation methods are mainly based upon spatial smoothness and assume and the vacation of brightness shape constancy
If.Assuming that it is gray scale (brightness) value of (x, y) point in t that I (x, y, t) is image coordinate.When the t+dt moment, the point is moved to
(x+dx, y+dy) point, assumes, it is believed that at the t+dt moment in the bright of picture point (x+dx, y+dy) place according to brightness shape constancy
The brightness that degree is put with (x, y) of moment t is identical, i.e. optical flow constraint equation is:
I (x, y, t)=I (x+dx, y+dy, t+dt) (2)
If it is considered to gradation of image is the consecutive variations function of position and time, then by above formula the right Taylor series expansion
And quadratic term and higher order term are omitted, obtain the fundamental equation of optical flow field:
The optical flow field that Horn and Schunck causes according to same moving object should be continuously smooth, it is proposed that space
Smoothness constraint is assumed.The hypothesis thinks that the movement velocity of object is local smooth in many cases, or with changing for putting
Become and it is slowly varying but very small in the change of regional area.Particularly target is each adjacent when undeformed rigid motion is made
Pixel should have identical movement velocity, i.e., the spatial variations rate of adjacent spot speed is zero, can be expressed as:
Wherein u and v is that coordinate is the pixel of (x, y) in direction and the instantaneous velocity in direction on the moment plane of delineation
Component, that is, light stream.Formula (3) and (4) are combined into solution, is formed being weighted two constraintss and is sought extreme value
Problem.
Light stream carries the abundant information about object of which movement and scenery three-dimensional structure, can be not only used in this way
Moving object detection, it might even be possible to be directly used in motion target tracking, and also can be just on the premise of camera has motion
Really detect moving target.But in actual applications, due to reasons such as multiple light courcess, blocking property, noises so that optical flow field is basic
The gray scale conservation assumed condition of equation tends not to meet, it is impossible to correct optical flow field is solved, while most of optical flow computations
Method is also considerably complicated, and amount of calculation is huge, it is impossible to meet requirement of real time.
The present invention carries out Corner Feature extraction to the moving vehicle region that each two field picture is detected first, reapplies golden word
Tower Lucas-Kanade sparse optical flows algorithm is tracked to the angle point in all frame video images, obtains moving vehicle motion rail
Mark.
Step 2:Track of vehicle feature extraction
The most basic data mode of track of vehicle, as position coordinates.It is all by can be calculated according to position coordinates
Such as target vehicle movement velocity, direction information.The movement locus of the moving vehicle obtained with the tracking of step 1 target detection is base
Plinth, the track characteristic extracting method being combined using matrix grid and two-way histogram constructs track of vehicle characteristic vector, can be with
For vehicle behavior classification provides feature foundation.Specially:
(1) on the basis of moving vehicle trajectory coordinates, with track abscissa as x-axis, track ordinate be y-axis construct vehicle
Trajectory coordinates system Oxy, obtain respectively x-axis direction maximum x a littlemaxWith minimum value xmin, y-axis direction maximum ymaxWith
Minimum value ymin, create a width of in coordinate system OxyIt is a height ofMatrix grid, to grid
In each sub-grid assign initial value 0;
(2) assignment is carried out to the sub-grid where trajectory coordinates, first along x-axis positive direction, then along y-axis positive direction, in order
Assign incremental weighted value 1,2,3 ..., obtain trajectory coordinates matrix;
(3) row and column to trajectory coordinates matrix constructs histogram respectively, and vehicle target is estimated using two-way histogram
Basic act trend, i.e., move towards according to the histogram overall trend of trajectory coordinates matrix row and column, if successively decreased from left to right,
It is intended to turn right;If be incremented by from left to right, it is intended to turn left;If centre is successively decreased to both sides, trajectory is in straight
Row trend, obtains moving vehicle behavior trend to turn left, turning right or keep straight on;
Wherein, histogram abscissa represents the line number or columns of trajectory coordinates matrix, and histogrammic ordinate counts track
Matrix element value is not 0 number in a certain row or column in coordinates matrix;
(4) the track characteristic vector of moving vehicle is then made up of moving vehicle behavior trend and trajectory coordinates matrix.
Step 3:Vehicle Behavioral training
Supporting vector machine (support vector machines, SVM) is a kind of classification mould maximum based on class class interval
Type.Because it is to small sample, the classification problem of higher-dimension has preferable effect to SVM, and training and test phase are relatively easy, therefore
Substantial amounts of application has been obtained in the problem of pattern-recognition and machine learning.Because support vector machine classifier is only capable of recognizing input
Into two classes, and it is often that more than two classes, use of the present invention can recognize the two layers of classified device knot of three classes to need the feature distinguished
Structure.
Before vehicle Activity recognition is carried out using SVM, track of vehicle sample is entered first with SVM two layers of classified device structures
Row Behavioral training, obtains the recognition result of sample data, specially:
(1) track sample is created:The simulation background of the video of collection in advance is imported with picture Core Generator, and establishment includes
Straight trip, left-hand bend, each 100 of the track sample of right-hand bend three types;
(2) the track characteristic extracting method according to step 2 carries out track characteristic extraction to track sample, obtains all kinds
The track characteristic vector of track sample;
(3) the track characteristic vector of sample is trained using SVM two layers of classified device, obtains the identification knot of sample data
Fruit and the SVM classifier for training;
Step 4:Vehicle Activity recognition
According to the SVM classifier that step 3 training is obtained, the moving vehicle track characteristic vector obtained to step 2 enters driving
Activity recognition, finally gives vehicle behavior to turn left, turning right or keep straight on.
The present embodiment carries out method emulation using VS2010 platforms and OpenCV technologies, realizes to vehicle detection, tracking
And Activity recognition.Test result shows, can be with complete using the vehicle target detection method and optical flow method algorithm of background subtraction
The detection of paired vehicle target, tracking (as shown in Figure 2), being combined using matrix grid and two-way histogram can extract vehicle
Track characteristic (as shown in Figure 3), the Classification and Identification to vehicle behavior, and experiment can be completed using the sorting algorithm based on SVM
Result is accurate.
Claims (1)
1. a kind of urban environment driving vehicle Activity recognition method based on machine vision, it is characterised in that step is as follows:
Step 1:Vehicle target is detected and tracking:The method modeled using background subtraction and mixed Gauss model carries out sport(s) car
Target detection, is then tracked, specially using the track algorithm of light stream to the moving vehicle target for detecting:
Step a:Respectively according toWithCalculate one
The mean μ of the pixel intensity of video sequence image in the section time0(x, y) and varianceWith μ0(x, y) andPoint
Wei not the image B of pixel average and variance composition with Gaussian Profile0, B0As initial background estimating image;
Wherein, N is the totalframes of sequence image in initial background image access time section, 150≤N≤200;fi(x, y) is the
The pixel brightness value that i two field pictures are arranged in xth row, y, (x, y) represents that the location of pixels in image is x rows, y row;
Step b:Respectively according to μj(x, y)=(1- α) μj-1(x,y)+α·fj(x, y) andUpdate the mean μ of background estimating imagej(x, y) and side
DifferenceThe background estimating image B of the jth two field picture after being updatedj;
Wherein,δ is the constant between [0,1], and K is mixed Gaussian
The number of model, 3≤K≤5;J >=1, fj(x, y) represents the pixel brightness value that jth two field picture is arranged in xth row, y;
Step c:According to dj(x, y)=| fj(x,y)-Bj(x, y) | it is calculated current frame image and present frame background estimating image
Difference image, and according toBinary conversion treatment is carried out to difference image, is detected
Pixel value is 1 region in image M behind moving vehicle region, i.e. binary conversion treatment, and pixel value is that 0 region is the back of the body in image M
Scene area;M in jth two field picturejThe moving region that (x, y)=1 represents;
Wherein, r is gray threshold, 50≤r≤60;
Step d:Corner Feature extraction is carried out to the moving vehicle region that each two field picture is detected, pyramid Lucas- is reapplied
Kanade sparse optical flows algorithm is tracked to the angle point in all frame video images, obtains moving vehicle movement locus;
Step 2:Track of vehicle feature extraction:The track characteristic extracting method being combined using matrix grid and two-way histogram,
Construction track of vehicle characteristic vector, for vehicle behavior classification provides feature foundation, specially:
Step a:On the basis of the moving vehicle trajectory coordinates obtained by step 1, with track abscissa as x-axis, track ordinate is as y
Axle constructs track of vehicle coordinate system Oxy, obtain respectively x-axis direction maximum x a littlemaxWith minimum value xmin, y-axis direction most
Big value ymaxWith minimum value ymin, create a width of in coordinate system OxyIt is a height ofMatrix net
Lattice, initial value 0 is assigned to each sub-grid in grid;
Step b:Assignment is carried out to the sub-grid where trajectory coordinates, first along x-axis positive direction, then along y-axis positive direction, is assigned in order
Give incremental weighted value 1,2,3 ..., obtain trajectory coordinates matrix;
Step c:Row and column to trajectory coordinates matrix constructs histogram respectively, and the base of vehicle target is estimated using two-way histogram
This behavior trend, i.e., move towards according to the histogram overall trend of trajectory coordinates matrix row and column, if successively decreased from left to right, becomes
To in right-hand bend;If be incremented by from left to right, it is intended to turn left;If centre is successively decreased to both sides, trajectory is in straight trip
Trend, obtains moving vehicle behavior trend to turn left, turning right or keep straight on;
Wherein, histogram abscissa represents the line number or columns of trajectory coordinates matrix, and histogrammic ordinate counts trajectory coordinates
Matrix element value is not 0 number in a certain row or column in matrix;
Step d:The track that the trajectory coordinates matrix that the behavior trend and step b obtained by step c are obtained constitutes moving vehicle is special
Levy vector;
Step 3:Vehicle Behavioral training:Behavioral training is carried out to track of vehicle sample using SVM two layers of classified device structures, sample is obtained
The recognition result of notebook data, specially:
Step a:Create track sample:The simulation background of the video of collection in advance is imported with picture Core Generator, and establishment includes
Straight trip, left-hand bend, each 100 of the track sample of right-hand bend three types;
Step b:Track characteristic extracting method according to step 2 carries out track characteristic extraction to track sample, obtains all kinds
The track characteristic vector of track sample;
Step c:The track characteristic vector of the sample obtained to step b using SVM two layers of classified device is trained, and obtains sample number
According to recognition result;
Step 4:Vehicle Activity recognition:According to the SVM classifier that step 3 training is obtained, the moving vehicle track obtained to step 2
Characteristic vector carries out vehicle Activity recognition, finally gives vehicle behavior to turn left, turning right or keep straight on.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710027523.9A CN106875424B (en) | 2017-01-16 | 2017-01-16 | A kind of urban environment driving vehicle Activity recognition method based on machine vision |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710027523.9A CN106875424B (en) | 2017-01-16 | 2017-01-16 | A kind of urban environment driving vehicle Activity recognition method based on machine vision |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106875424A true CN106875424A (en) | 2017-06-20 |
CN106875424B CN106875424B (en) | 2019-09-24 |
Family
ID=59158038
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710027523.9A Expired - Fee Related CN106875424B (en) | 2017-01-16 | 2017-01-16 | A kind of urban environment driving vehicle Activity recognition method based on machine vision |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106875424B (en) |
Cited By (28)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108229407A (en) * | 2018-01-11 | 2018-06-29 | 武汉米人科技有限公司 | A kind of behavioral value method and system in video analysis |
CN108830246A (en) * | 2018-06-25 | 2018-11-16 | 中南大学 | A kind of traffic environment pedestrian multi-dimensional movement characteristic visual extracting method |
CN109165602A (en) * | 2018-08-27 | 2019-01-08 | 成都华安视讯科技有限公司 | A kind of black smoke vehicle detection method based on video analysis |
CN109815856A (en) * | 2019-01-08 | 2019-05-28 | 深圳中兴网信科技有限公司 | Status indication method, system and the computer readable storage medium of target vehicle |
CN110414375A (en) * | 2019-07-08 | 2019-11-05 | 北京国卫星通科技有限公司 | Recognition methods, device, storage medium and the electronic equipment of low target |
CN110533687A (en) * | 2018-05-11 | 2019-12-03 | 深眸科技(深圳)有限公司 | Multiple target three-dimensional track tracking and device |
CN110675592A (en) * | 2019-08-16 | 2020-01-10 | 重庆特斯联智慧科技股份有限公司 | High-altitude parabolic early warning protection system based on target identification and control method |
CN110782485A (en) * | 2019-10-31 | 2020-02-11 | 广东泓胜科技股份有限公司 | Vehicle lane change detection method and device |
CN110954893A (en) * | 2019-12-23 | 2020-04-03 | 山东师范大学 | Method and system for identifying wall rear action based on wireless router |
CN111105437A (en) * | 2018-10-29 | 2020-05-05 | 西安宇视信息科技有限公司 | Vehicle track abnormity judgment method and device |
CN111126144A (en) * | 2019-11-20 | 2020-05-08 | 浙江工业大学 | Vehicle track abnormity detection method based on machine learning |
CN111754550A (en) * | 2020-06-12 | 2020-10-09 | 中国农业大学 | Method and device for detecting dynamic barrier in motion state of agricultural machine |
CN111746555A (en) * | 2019-03-26 | 2020-10-09 | 通用汽车环球科技运作有限责任公司 | Identification and avoidance of collision behavior |
CN111862142A (en) * | 2020-07-23 | 2020-10-30 | 软通智慧科技有限公司 | Motion trajectory generation method, device, equipment and medium |
CN111914627A (en) * | 2020-06-18 | 2020-11-10 | 广州杰赛科技股份有限公司 | Vehicle identification and tracking method and device |
CN112037250A (en) * | 2020-07-27 | 2020-12-04 | 国网四川省电力公司 | Target vehicle vector trajectory tracking and engineering view modeling method and device |
CN112085063A (en) * | 2020-08-10 | 2020-12-15 | 深圳市优必选科技股份有限公司 | Target identification method and device, terminal equipment and storage medium |
CN112084970A (en) * | 2020-09-14 | 2020-12-15 | 西安莱奥信息科技有限公司 | Vehicle identification method and device based on machine vision |
CN112132869A (en) * | 2020-11-02 | 2020-12-25 | 中远海运科技股份有限公司 | Vehicle target track tracking method and device |
CN112183252A (en) * | 2020-09-15 | 2021-01-05 | 珠海格力电器股份有限公司 | Video motion recognition method and device, computer equipment and storage medium |
CN112258462A (en) * | 2020-10-13 | 2021-01-22 | 广州杰赛科技股份有限公司 | Vehicle detection method and device and computer readable storage medium |
CN112416599A (en) * | 2020-12-03 | 2021-02-26 | 腾讯科技(深圳)有限公司 | Resource scheduling method, device, equipment and computer readable storage medium |
CN112418213A (en) * | 2020-11-06 | 2021-02-26 | 北京航天自动控制研究所 | Vehicle driving track identification method and device and storage medium |
CN113077494A (en) * | 2021-04-10 | 2021-07-06 | 山东沂蒙交通发展集团有限公司 | Road surface obstacle intelligent recognition equipment based on vehicle orbit |
CN113538891A (en) * | 2020-04-17 | 2021-10-22 | 无锡锦铖人工智能科技有限公司 | Intelligent vehicle counting system |
CN116071688A (en) * | 2023-03-06 | 2023-05-05 | 台州天视智能科技有限公司 | Behavior analysis method and device for vehicle, electronic equipment and storage medium |
CN117555333A (en) * | 2023-11-21 | 2024-02-13 | 深圳云程科技有限公司 | Dynamic travel track processing system and method |
CN111862142B (en) * | 2020-07-23 | 2024-08-02 | 软通智慧科技有限公司 | Motion trail generation method, device, equipment and medium |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104658006A (en) * | 2013-11-22 | 2015-05-27 | 上海宝康电子控制工程有限公司 | Method for achieving vehicle tracking based on variable split beam stream |
CN104951764A (en) * | 2015-06-17 | 2015-09-30 | 浙江工业大学 | Identification method for behaviors of high-speed vehicle based on secondary spectrum clustering and HMM (Hidden Markov Model)-RF (Random Forest) hybrid model |
-
2017
- 2017-01-16 CN CN201710027523.9A patent/CN106875424B/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104658006A (en) * | 2013-11-22 | 2015-05-27 | 上海宝康电子控制工程有限公司 | Method for achieving vehicle tracking based on variable split beam stream |
CN104951764A (en) * | 2015-06-17 | 2015-09-30 | 浙江工业大学 | Identification method for behaviors of high-speed vehicle based on secondary spectrum clustering and HMM (Hidden Markov Model)-RF (Random Forest) hybrid model |
Non-Patent Citations (1)
Title |
---|
XUAN NIE ET AL.: "A New Method on Vehicle Behavior Analysis of Intelligent Transportation Monitoring", 《LISS 2014》 * |
Cited By (35)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108229407A (en) * | 2018-01-11 | 2018-06-29 | 武汉米人科技有限公司 | A kind of behavioral value method and system in video analysis |
CN110533687B (en) * | 2018-05-11 | 2023-09-12 | 上海美城智能科技有限公司 | Multi-target three-dimensional track tracking method and device |
CN110533687A (en) * | 2018-05-11 | 2019-12-03 | 深眸科技(深圳)有限公司 | Multiple target three-dimensional track tracking and device |
CN108830246A (en) * | 2018-06-25 | 2018-11-16 | 中南大学 | A kind of traffic environment pedestrian multi-dimensional movement characteristic visual extracting method |
CN108830246B (en) * | 2018-06-25 | 2022-02-15 | 中南大学 | Multi-dimensional motion feature visual extraction method for pedestrians in traffic environment |
CN109165602A (en) * | 2018-08-27 | 2019-01-08 | 成都华安视讯科技有限公司 | A kind of black smoke vehicle detection method based on video analysis |
CN111105437B (en) * | 2018-10-29 | 2024-03-29 | 西安宇视信息科技有限公司 | Vehicle track abnormality judging method and device |
CN111105437A (en) * | 2018-10-29 | 2020-05-05 | 西安宇视信息科技有限公司 | Vehicle track abnormity judgment method and device |
CN109815856A (en) * | 2019-01-08 | 2019-05-28 | 深圳中兴网信科技有限公司 | Status indication method, system and the computer readable storage medium of target vehicle |
CN111746555A (en) * | 2019-03-26 | 2020-10-09 | 通用汽车环球科技运作有限责任公司 | Identification and avoidance of collision behavior |
CN110414375A (en) * | 2019-07-08 | 2019-11-05 | 北京国卫星通科技有限公司 | Recognition methods, device, storage medium and the electronic equipment of low target |
CN110675592A (en) * | 2019-08-16 | 2020-01-10 | 重庆特斯联智慧科技股份有限公司 | High-altitude parabolic early warning protection system based on target identification and control method |
CN110782485A (en) * | 2019-10-31 | 2020-02-11 | 广东泓胜科技股份有限公司 | Vehicle lane change detection method and device |
CN111126144A (en) * | 2019-11-20 | 2020-05-08 | 浙江工业大学 | Vehicle track abnormity detection method based on machine learning |
CN111126144B (en) * | 2019-11-20 | 2021-10-12 | 浙江工业大学 | Vehicle track abnormity detection method based on machine learning |
CN110954893A (en) * | 2019-12-23 | 2020-04-03 | 山东师范大学 | Method and system for identifying wall rear action based on wireless router |
CN113538891A (en) * | 2020-04-17 | 2021-10-22 | 无锡锦铖人工智能科技有限公司 | Intelligent vehicle counting system |
CN111754550A (en) * | 2020-06-12 | 2020-10-09 | 中国农业大学 | Method and device for detecting dynamic barrier in motion state of agricultural machine |
CN111754550B (en) * | 2020-06-12 | 2023-08-11 | 中国农业大学 | Method and device for detecting dynamic obstacle in movement state of agricultural machine |
CN111914627A (en) * | 2020-06-18 | 2020-11-10 | 广州杰赛科技股份有限公司 | Vehicle identification and tracking method and device |
CN111862142B (en) * | 2020-07-23 | 2024-08-02 | 软通智慧科技有限公司 | Motion trail generation method, device, equipment and medium |
CN111862142A (en) * | 2020-07-23 | 2020-10-30 | 软通智慧科技有限公司 | Motion trajectory generation method, device, equipment and medium |
CN112037250A (en) * | 2020-07-27 | 2020-12-04 | 国网四川省电力公司 | Target vehicle vector trajectory tracking and engineering view modeling method and device |
CN112037250B (en) * | 2020-07-27 | 2024-04-05 | 国网四川省电力公司 | Target vehicle vector track tracking and engineering view modeling method and device |
CN112085063A (en) * | 2020-08-10 | 2020-12-15 | 深圳市优必选科技股份有限公司 | Target identification method and device, terminal equipment and storage medium |
CN112085063B (en) * | 2020-08-10 | 2023-10-13 | 深圳市优必选科技股份有限公司 | Target identification method, device, terminal equipment and storage medium |
CN112084970A (en) * | 2020-09-14 | 2020-12-15 | 西安莱奥信息科技有限公司 | Vehicle identification method and device based on machine vision |
CN112183252A (en) * | 2020-09-15 | 2021-01-05 | 珠海格力电器股份有限公司 | Video motion recognition method and device, computer equipment and storage medium |
CN112258462A (en) * | 2020-10-13 | 2021-01-22 | 广州杰赛科技股份有限公司 | Vehicle detection method and device and computer readable storage medium |
CN112132869A (en) * | 2020-11-02 | 2020-12-25 | 中远海运科技股份有限公司 | Vehicle target track tracking method and device |
CN112418213A (en) * | 2020-11-06 | 2021-02-26 | 北京航天自动控制研究所 | Vehicle driving track identification method and device and storage medium |
CN112416599A (en) * | 2020-12-03 | 2021-02-26 | 腾讯科技(深圳)有限公司 | Resource scheduling method, device, equipment and computer readable storage medium |
CN113077494A (en) * | 2021-04-10 | 2021-07-06 | 山东沂蒙交通发展集团有限公司 | Road surface obstacle intelligent recognition equipment based on vehicle orbit |
CN116071688A (en) * | 2023-03-06 | 2023-05-05 | 台州天视智能科技有限公司 | Behavior analysis method and device for vehicle, electronic equipment and storage medium |
CN117555333A (en) * | 2023-11-21 | 2024-02-13 | 深圳云程科技有限公司 | Dynamic travel track processing system and method |
Also Published As
Publication number | Publication date |
---|---|
CN106875424B (en) | 2019-09-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106875424A (en) | A kind of urban environment driving vehicle Activity recognition method based on machine vision | |
Manh et al. | Scene-lstm: A model for human trajectory prediction | |
Zhang et al. | A traffic surveillance system for obtaining comprehensive information of the passing vehicles based on instance segmentation | |
CN114970321A (en) | Scene flow digital twinning method and system based on dynamic trajectory flow | |
CN107730881A (en) | Traffic congestion vision detection system based on depth convolutional neural networks | |
CN107730903A (en) | Parking offense and the car vision detection system that casts anchor based on depth convolutional neural networks | |
CN108830246B (en) | Multi-dimensional motion feature visual extraction method for pedestrians in traffic environment | |
CN108304798A (en) | The event video detecting method of order in the street based on deep learning and Movement consistency | |
CN102592138B (en) | Object tracking method for intensive scene based on multi-module sparse projection | |
CN101271527A (en) | Exception action detecting method based on athletic ground partial statistics characteristic analysis | |
CN101094413A (en) | Real time movement detection method in use for video monitoring | |
CN111402632B (en) | Risk prediction method for pedestrian movement track at intersection | |
CN110349192A (en) | A kind of tracking of the online Target Tracking System based on three-dimensional laser point cloud | |
CN105046206A (en) | Pedestrian detection method and apparatus based on moving associated prior information in videos | |
CN103268470A (en) | Method for counting video objects in real time based on any scene | |
CN107315998A (en) | Vehicle class division method and system based on lane line | |
Kavitha et al. | Pothole and object detection for an autonomous vehicle using yolo | |
CN109242019A (en) | A kind of water surface optics Small object quickly detects and tracking | |
CN103794050A (en) | Real-time transport vehicle detecting and tracking method | |
CN110032952A (en) | A kind of road boundary point detecting method based on deep learning | |
CN106056078A (en) | Crowd density estimation method based on multi-feature regression ensemble learning | |
Kim et al. | Semantic point cloud-based adaptive multiple object detection and tracking for autonomous vehicles | |
CN103473789A (en) | Human body video segmentation method fusing multi-cues | |
Khan | Estimating Speeds and Directions of Pedestrians in Real-Time Videos: A solution to Road-Safety Problem. | |
CN111145551A (en) | Intersection traffic planning system based on CNN detection follows chapter rate |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190924 Termination date: 20200116 |
|
CF01 | Termination of patent right due to non-payment of annual fee |