CN105426811A - Crowd abnormal behavior and crowd density recognition method - Google Patents
Crowd abnormal behavior and crowd density recognition method Download PDFInfo
- Publication number
- CN105426811A CN105426811A CN201510640197.XA CN201510640197A CN105426811A CN 105426811 A CN105426811 A CN 105426811A CN 201510640197 A CN201510640197 A CN 201510640197A CN 105426811 A CN105426811 A CN 105426811A
- Authority
- CN
- China
- Prior art keywords
- crowd
- frame
- vector
- video
- crowd density
- 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
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a crowd abnormal behavior and crowd density recognition method. The method includes: calculating a displacement vector of each pixel point between the ith frame and the (i+1)th frame of a to-be-detected video; extracting a moving region foreground and performing binarization processing, and obtaining a moving region foreground image after binarization; calculating the displacement distance of each pixel point of the moving region foreground image after binarization, and removing the pixel points with smaller displacement; solving second-order differential in the x-direction and the y-direction for the remaining pixel points, and obtaining spatial characteristic points; calculating the average displacement vector along the directions of the x-axis and the y-axis of a neighborhood with a certain size taking each spatial characteristic point as the core; repeating the above steps for the (i+1)th frame and the (i+2)th frame of the video; calculating the video vector between the ith frame and the (i+1)th frame and a video vector difference between the (i+1)th frame and the (i+2)th frame, and generating a global characteristic vector via statistics; and inputting the global characteristic vector to a support vector machine for training, and performing classified training of crowd anomaly and crowd density.
Description
Technical field
The present invention relates to technical field of computer vision, be specifically related to a kind of crowd's abnormal behaviour and crowd density recognition methods.
Background technology
Computer vision technique is the technology how a kind of research uses computing machine and relevant device simulation biological vision thereof.Gather picture or video by the imaging device such as video camera to go forward side by side row relax, obtain the three-dimensional information of corresponding scene, then transfer to computer generation to replace brain complete process and understand.This technology relates to multiple subject, comprises image procossing, pattern-recognition, graphical analysis and image understanding etc.At present, computer vision technique is widely used in every field, as medical image process, video monitoring, electronics bayonet socket, virtual reality, intelligent transportation etc.
Along with the development of computer vision technique and monitoring instrument the public place such as subway station, airport universal, crowd's abnormal behaviour and crowd density recognition technology successfully realize.But relevant theory and product are still little, current technology is mainly by sampling particle moving direction and particle translational speed etc. index comprehensive evaluation crowd exception and crowd density.
For crowd density evaluation, concrete scheme comprises:
1. utilize three-dimensional Hessian matrix to detect the unique point that in crowd's video, change in time and space is violent;
2. extract the dynamic texture of unique point with space-time local binarization mode method, then analysis of spectrum is carried out to dynamic texture thus obtains the local feature of image;
3. utilize the local feature of statistics with histogram to entire image to analyze, obtain crowd's global characteristics vector, then the crowd's global characteristics obtained vector inputted support vector machine thus estimate the global density grade of crowd;
4. local feature is carried out processing and being transformed into color space, the chromatogram display of crowd local density distribution can be obtained.
For crowd's anomaly evaluation, concrete scheme comprises:
1., for the video of different size, adopt the grid of different size to sample, each sampling particle is considered as body one by one, calculate the weighting interreaction force between particle in conjunction with local density's information;
2. the local feature (weighting interreaction force) of pair each sample point carries out statistics with histogram, and force direction is divided into N number of interval, and size is divided into M interval, and statistics generates the global characteristics vector of N × M dimension, inputs support vector machine training and carries out tagsort;
3. the classification results of pair support vector machine carries out medium filtering to remove the trip point in classification results.
Although such scheme can carry out the identification of crowd's abnormal behaviour and crowd density, be unfavorable for the restriction of the factor identified because saltus step between, frame of video uneven by the interference of extraneous complex environment, Crowds Distribute is large etc., there is more defect, comprising:
1. cannot extract prospect and background, to whole frame of video uniform sampling, can only cause there is a lot of invalid information in the information obtained, make algorithm whole efficiency not high.
2. be that employing two kinds of incoherent algorithms calculate extremely for crowd density and crowd, need to carry out two-wheeled machine learning training: the first round carries out learning training to crowd density for this reason, this crowd density recognition training of taking turns must complete and obtain good achievement, so just can carry out the training that second takes turns crowd's anomalous identification, cause whole process CIMS loaded down with trivial details, in actual environment, practicality is poor, and owing to adopting algorithm more, operand is larger, cause cpu load high, algorithm length consuming time.
3. the crowd that calculates proposes " pressure model " time abnormal based on " social force model " theory, essence is merged crowd density and crowd's abnormal behaviour, the judgement of crowd density higher crowd abnormal behaviour is more responsive, more sensitivity is lower in the judgement of contrary crowd density lower crowd behaviour exception, so broadly crowd density and abnormal combination of crowd is increased false drop rate (FN) and loss (FP).
Summary of the invention
The object of the invention is to address the deficiencies of the prior art, provide crowd's abnormal behaviour and the crowd density recognition methods of a kind of high-level efficiency and degree of accuracy, the technical scheme of employing is as follows:
A kind of crowd's abnormal behaviour and crowd density recognition methods, comprising:
S1. the displacement vector of video i-th frame to be detected and each pixel of the i-th+1 interframe is calculated;
S2. extract moving region prospect and carry out binary conversion treatment, obtaining the moving region foreground image after binaryzation;
S3. calculate the shift length of each pixel of the moving region foreground image after binaryzation, reject the pixel that displacement is less;
S4. for remaining pixel, ask the second-order differential in x direction and y direction, obtain space characteristics point;
S5. calculate with the average displacement vector along the x-axis direction and along the y-axis direction of each space characteristics point certain specification neighborhood that is core;
S6. for the i-th+1 frame and the i-th+2 frame video, step S1 to S5 is repeated;
S7. calculate the difference of the vector of vector i-th+1 frame to the i-th+2 inter-frame video of the i-th frame to the i-th+1 inter-frame video, and carry out statistics generation global characteristics vector;
S8. by the training of global characteristics vector input support vector machine, respectively classification based training is carried out to crowd's exception and crowd density, to realize identifying while crowd density and crowd's exception.
As preferably, in described step S1, optical flow method is used to calculate the displacement vector of video i-th frame to be detected and each pixel of the i-th+1 interframe.
As preferably, in described step S2, frame-to-frame differences method is used to extract moving region prospect.
As preferably, in described step S5, calculate with the average displacement vector along the x-axis direction and along the y-axis direction of each space characteristics point 3 × 3 neighborhoods that are core.
As preferably, in described step S7, the difference of histogram to the vector of vector i-th+1 frame to the i-th+2 inter-frame video of the i-th frame to the i-th+1 inter-frame video is used to add up.
As preferably, in described step S8, be divided into normal and abnormal two class results to crowd is abnormal, extremely low, basic, normal, high and high five class results are divided into crowd density.
Compared with prior art, beneficial effect of the present invention: invention has been moving region foreground extraction, eliminates a large amount of garbage and interference; By the calculating adopting a kind of algorithm to carry out crowd's exception and crowd density simultaneously, greatly reduce operand, improve efficiency, and do not disturb each other; By the sense of displacement of each sampled point and displacement size two parameters are converted into, continuous two inter-frame video picture displacement phasor differences are compared again, be converted to one-dimensional data by 2-D data, make the robustness of algorithm better.Conventional art is added up the sense of displacement of sampled point and displacement size two features, be divided into 9 different directions, 7 different velocity magnitude shift values are divided in each direction, cause the error that caused by thin small angle variation near the separator bar in each direction larger, and conventional art is divided into 7 grades to the displacement size method for normalizing in each direction, and displacement size is divided into 50 intervals by this method, and be the absolute value in units of pixel, make statistics more accurate, and eliminate the similarity caused by normalization.
Accompanying drawing explanation
Fig. 1 is process flow diagram of the present invention;
Fig. 2 is the present embodiment experiment figure;
Fig. 3 carries out to the moving region prospect extracted the image that binary conversion treatment obtains;
Fig. 4 is the schematic diagram of second-order differential pixel being asked to x direction and y direction, the space characteristics obtained point;
Fig. 5 is the difference of the vector of vector i-th+1 frame to the i-th+2 inter-frame video of calculating i-th frame to the i-th+1 inter-frame video, and uses feature histogram to carry out adding up the image obtained.
Embodiment
Below in conjunction with drawings and Examples, the present invention is described in further detail.
Embodiment: as shown in Figure 1, uses method of the present invention to Fig. 2 process, comprising:
S1. the displacement vector of video i-th frame to be detected and each pixel of the i-th+1 interframe is calculated;
S2. extract moving region prospect and carry out binary conversion treatment, obtaining the moving region foreground image after binaryzation, as shown in Figure 3;
S3. calculate the shift length of each pixel of the moving region foreground image after binaryzation, reject the pixel that displacement is less;
S4. for remaining pixel, ask the second-order differential in x direction and y direction, obtain space characteristics point, as shown in Figure 4;
S5. calculate with the average displacement vector along the x-axis direction and along the y-axis direction of each space characteristics point certain specification neighborhood that is core;
S6. for the i-th+1 frame and the i-th+2 frame video, step S1 to S5 is repeated;
S7. calculate the difference of the vector of vector i-th+1 frame to the i-th+2 inter-frame video of the i-th frame to the i-th+1 inter-frame video, and carry out statistics generation global characteristics vector;
S8. by the training of global characteristics vector input support vector machine, respectively classification based training is carried out to crowd's exception and crowd density, to realize identifying while crowd density and crowd's exception.
In described step S1, optical flow method is used to calculate the displacement vector of video i-th frame to be detected and each pixel of the i-th+1 interframe.
In described step S2, frame-to-frame differences method is used to extract moving region prospect.
In described step S5, calculate with the average displacement vector along the x-axis direction and along the y-axis direction of each space characteristics point 3 × 3 neighborhoods that are core.
In described step S7, the difference of histogram to the vector of vector i-th+1 frame to the i-th+2 inter-frame video of the i-th frame to the i-th+1 inter-frame video is used to add up, as shown in Figure 5.
In described step S8, be divided into normal and abnormal two class results to crowd is abnormal, extremely low, basic, normal, high and high five class results are divided into crowd density.
The present embodiment has carried out moving region foreground extraction, eliminates a large amount of garbage and interference; By the calculating adopting a kind of algorithm to carry out crowd's exception and crowd density simultaneously, greatly reduce operand, improve efficiency, and do not disturb each other; By the sense of displacement of each sampled point and displacement size two parameters are converted into, continuous two inter-frame video picture displacement phasor differences are compared again, be converted to one-dimensional data by 2-D data, make the robustness of algorithm better.
Claims (6)
1. crowd's abnormal behaviour and a crowd density recognition methods, is characterized in that, comprising:
S1. the displacement vector of video i-th frame to be detected and each pixel of the i-th+1 interframe is calculated;
S2. extract moving region prospect and carry out binary conversion treatment, obtaining the moving region foreground image after binaryzation;
S3. calculate the shift length of each pixel of the moving region foreground image after binaryzation, reject the pixel that displacement is less;
S4. for remaining pixel, ask the second-order differential in x direction and y direction, obtain space characteristics point;
S5. calculate with the average displacement vector along the x-axis direction and along the y-axis direction of each space characteristics point certain specification neighborhood that is core;
S6. for the i-th+1 frame and the i-th+2 frame video, step S1 to S5 is repeated;
S7. calculate the difference of the vector of vector i-th+1 frame to the i-th+2 inter-frame video of the i-th frame to the i-th+1 inter-frame video, and carry out statistics generation global characteristics vector;
S8. by the training of global characteristics vector input support vector machine, respectively classification based training is carried out to crowd's exception and crowd density, to realize identifying while crowd density and crowd's exception.
2. a kind of crowd's abnormal behaviour according to claim 1 and crowd density recognition methods, is characterized in that, in described step S1, uses optical flow method to calculate the displacement vector of video i-th frame to be detected and each pixel of the i-th+1 interframe.
3. a kind of crowd's abnormal behaviour according to claim 1 and crowd density recognition methods, is characterized in that, in described step S2, uses frame-to-frame differences method to extract moving region prospect.
4. a kind of crowd's abnormal behaviour according to claim 1 and crowd density recognition methods, is characterized in that, in described step S5, calculates with the average displacement vector along the x-axis direction and along the y-axis direction of each space characteristics point 3 × 3 neighborhoods that are core.
5. a kind of crowd's abnormal behaviour according to claim 1 and crowd density recognition methods, it is characterized in that, in described step S7, the difference of histogram to the vector of vector i-th+1 frame to the i-th+2 inter-frame video of the i-th frame to the i-th+1 inter-frame video is used to add up.
6. a kind of crowd's abnormal behaviour according to claim 1 and crowd density recognition methods, is characterized in that, in described step S8, is divided into normal and abnormal two class results, is divided into extremely low, basic, normal, high and high five class results to crowd density to crowd is abnormal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510640197.XA CN105426811B (en) | 2015-09-28 | 2015-09-28 | A kind of crowd's abnormal behaviour and crowd density recognition methods |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510640197.XA CN105426811B (en) | 2015-09-28 | 2015-09-28 | A kind of crowd's abnormal behaviour and crowd density recognition methods |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105426811A true CN105426811A (en) | 2016-03-23 |
| CN105426811B CN105426811B (en) | 2019-03-15 |
Family
ID=55505011
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510640197.XA Active CN105426811B (en) | 2015-09-28 | 2015-09-28 | A kind of crowd's abnormal behaviour and crowd density recognition methods |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105426811B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110796035A (en) * | 2019-10-14 | 2020-02-14 | 上海复瞰科技有限公司 | People entering and exiting counting method based on human shape detection and speed calculation |
| CN111079536A (en) * | 2019-11-18 | 2020-04-28 | 高新兴科技集团股份有限公司 | Behavior analysis method based on human body key point time sequence, storage medium and equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101159859A (en) * | 2007-11-29 | 2008-04-09 | 北京中星微电子有限公司 | Motion detection method, device and an intelligent monitoring system |
| CN102324016A (en) * | 2011-05-27 | 2012-01-18 | 郝红卫 | Statistical method for high-density crowd flow |
| CN103793920A (en) * | 2012-10-26 | 2014-05-14 | 杭州海康威视数字技术股份有限公司 | Retro-gradation detection method based on video and system thereof |
| US20140140583A1 (en) * | 2012-08-22 | 2014-05-22 | Canon Kabushiki Kaisha | Image recognition apparatus and image recognition method for identifying object |
| CN104123544A (en) * | 2014-07-23 | 2014-10-29 | 通号通信信息集团有限公司 | Video analysis based abnormal behavior detection method and system |
-
2015
- 2015-09-28 CN CN201510640197.XA patent/CN105426811B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101159859A (en) * | 2007-11-29 | 2008-04-09 | 北京中星微电子有限公司 | Motion detection method, device and an intelligent monitoring system |
| CN102324016A (en) * | 2011-05-27 | 2012-01-18 | 郝红卫 | Statistical method for high-density crowd flow |
| US20140140583A1 (en) * | 2012-08-22 | 2014-05-22 | Canon Kabushiki Kaisha | Image recognition apparatus and image recognition method for identifying object |
| CN103793920A (en) * | 2012-10-26 | 2014-05-14 | 杭州海康威视数字技术股份有限公司 | Retro-gradation detection method based on video and system thereof |
| CN104123544A (en) * | 2014-07-23 | 2014-10-29 | 通号通信信息集团有限公司 | Video analysis based abnormal behavior detection method and system |
Non-Patent Citations (1)
| Title |
|---|
| 陈禹: ""基于视频的人群数量统计及异常检测方法研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110796035A (en) * | 2019-10-14 | 2020-02-14 | 上海复瞰科技有限公司 | People entering and exiting counting method based on human shape detection and speed calculation |
| CN110796035B (en) * | 2019-10-14 | 2024-05-24 | 上海复瞰科技有限公司 | People entering and exiting number counting method based on human shape detection and speed calculation |
| CN111079536A (en) * | 2019-11-18 | 2020-04-28 | 高新兴科技集团股份有限公司 | Behavior analysis method based on human body key point time sequence, storage medium and equipment |
| CN111079536B (en) * | 2019-11-18 | 2023-08-29 | 高新兴科技集团股份有限公司 | Behavior analysis method, storage medium and device based on human body key point time sequence |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105426811B (en) | 2019-03-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110097115B (en) | Video salient object detection method based on attention transfer mechanism | |
| CN107967440A (en) | A kind of monitor video method for detecting abnormality based on multizone mutative scale 3D-HOF | |
| CN106815578A (en) | A kind of gesture identification method based on Depth Motion figure Scale invariant features transform | |
| CN101493943B (en) | Particle filtering tracking method and tracking device | |
| CN101179713A (en) | Method of detecting single moving target under complex background | |
| CN103839065A (en) | Extraction method for dynamic crowd gathering characteristics | |
| CN104091157A (en) | Pedestrian detection method based on feature fusion | |
| CN104063702A (en) | Three-dimensional gait recognition based on shielding recovery and partial similarity matching | |
| CN102821228B (en) | Low-rank video background reconstructing method | |
| Liu et al. | DLSE-Net: A robust weakly supervised network for fabric defect detection | |
| Xie et al. | Fabric defect detection method combing image pyramid and direction template | |
| CN103903238A (en) | Method for fusing significant structure and relevant structure of characteristics of image | |
| CN103699874A (en) | Crowd abnormal behavior identification method based on SURF (Speed-Up Robust Feature) stream and LLE (Locally Linear Embedding) sparse representation | |
| Liang et al. | Automatic defect detection of texture surface with an efficient texture removal network | |
| CN104200218B (en) | A kind of across visual angle action identification method and system based on timing information | |
| CN103413154A (en) | Human motion identification method based on normalized class Google measurement matrix | |
| CN106056135A (en) | Human body motion classification method based on compression perception | |
| CN102708589B (en) | Three-dimensional target multi-viewpoint view modeling method on basis of feature clustering | |
| Chen et al. | Multi-scale attention networks for pavement defect detection | |
| CN110458809B (en) | Yarn evenness detection method based on sub-pixel edge detection | |
| Zhang et al. | Zero-DD: Zero-sample defect detection for industrial products | |
| CN105426811A (en) | Crowd abnormal behavior and crowd density recognition method | |
| Wang et al. | 3d reconstruction and segmentation system for pavement potholes based on improved structure-from-motion (sfm) and deep learning | |
| CN102201060B (en) | Method for tracking and evaluating nonparametric outline based on shape semanteme | |
| Jiao et al. | A steerable pyramid autoencoder based framework for anomaly frame detection of water pipeline CCTV inspection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |