CN104866830A - Abnormal motion detection method and device - Google Patents
Abnormal motion detection method and device Download PDFInfo
- Publication number
- CN104866830A CN104866830A CN201510279178.9A CN201510279178A CN104866830A CN 104866830 A CN104866830 A CN 104866830A CN 201510279178 A CN201510279178 A CN 201510279178A CN 104866830 A CN104866830 A CN 104866830A
- Authority
- CN
- China
- Prior art keywords
- foreground object
- depth
- abnormal operation
- difference image
- depth difference
- 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
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
- G06V40/23—Recognition of whole body movements, e.g. for sport training
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/103—Static body considered as a whole, e.g. static pedestrian or occupant recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Image Analysis (AREA)
Abstract
The application provides an abnormal motion detection method and device. The method comprises the steps that foreground objects in a monitoring video are detected according to depth information; depth difference of the foreground objects between adjacent frames is calculated so that depth difference images are obtained; continuous multiple frames of depth difference images are calculated so that polymerized depth difference images are obtained; features of a histogram of oriented gradients HOG are calculated according to the polymerized depth difference images; and abnormal motions corresponding to the HOG features are predicted by a support vector machine SVM classifier with well grained abnormal motions, and whether the abnormal motions occur on the foreground objects is confirmed according to the prediction result. According to the scheme of the embodiment of the application, the foreground objects in the monitoring video are detected according to the depth information so that people with different distances to a lens of the same position in the frame can be accurately separated, and thus the abnormal motions of each person in the scene can be accurately judged.
Description
Technical field
The application relates to technical field of computer vision, particularly relates to a kind of abnormal operation detection method and device.
Background technology
The abnormal operation of people detects has very large using value in intelligent security guard field, in a lot of monitoring scene, can by carrying out calculation process to security protection camera picture, real-time early warning is provided, such as: the abnormality detection etc. of people on the abnormality detection of people, square in bank when there being anomalous event to occur.
It is based on colour picture information mostly that the abnormal operation of current people detects, and specific practice can be divided into following two stages:
Training stage: first to each two field picture extract minutiae of training video, again the unique point of same two field picture is integrated into proper vector by the mode of " word bag " (Bag of Words), the last label training classifier according to training data, learns different action rule characteristically;
Test phase: for the video of input, first obtains the degree of confidence that on each two field picture, required movement occurs, then this degree of confidence is accumulated in a period of time, if accumulation degree of confidence exceedes predetermined threshold value, then determine that required movement occurs.
Based on the abnormal operation of colour picture infomation detection foreground object, because picture can only display plane pixel, when multiple object that same position in picture has distance camera lens distance different, accurately cannot determine which foreground object concrete has abnormal operation, accuracy of detection is lower.
Prior art deficiency is:
Existing abnormal operation detection method accuracy of detection is lower.
Summary of the invention
The embodiment of the present application proposes a kind of abnormal operation detection method and device, to solve the technical matters that in prior art, abnormal operation detection method accuracy of detection is lower.
The embodiment of the present application provides a kind of abnormal operation detection method, comprises the steps:
Step 1: detect the foreground object in monitor video according to depth information;
Step 3: the depth difference calculating described foreground object between consecutive frame, obtains depth difference image;
Step 5: the depth difference image of continuous multiple frames is calculated, obtains depth of cure difference image;
Step 7: according to described depth of cure difference image calculated direction histogram of gradients HOG feature;
Step 9: predicting by the SVM classifier of the good abnormal operation of training in advance the abnormal operation that described HOG feature is corresponding, determining whether described foreground object described abnormal operation occurs according to predicting the outcome.
The embodiment of the present application provides a kind of abnormal operation detection device, comprising:
Detection module, for detecting the foreground object in monitor video according to depth information;
Depth difference computing module, for calculating the depth difference of described foreground object between consecutive frame, obtains depth difference image;
Depth of cure difference computing module, for calculating the depth difference image of continuous multiple frames, obtains depth of cure difference image;
HOG feature calculation module, for according to described depth of cure difference image calculated direction histogram of gradients HOG feature;
According to predicting the outcome, determination module, for predicting by the SVM classifier of the good abnormal operation of training in advance the abnormal operation that described HOG feature is corresponding, determines whether described foreground object described abnormal operation occurs.
Beneficial effect is as follows:
The technical scheme that the embodiment of the present application provides, detects the foreground object in monitor video according to depth information, calculate the depth difference of described foreground object between consecutive frame, obtain depth difference image; Again the depth difference image of continuous multiple frames is carried out calculating depth of cure difference image, HOG feature is calculated according to described depth of cure difference image, predicting by SVM classifier the abnormal operation that described HOG feature is corresponding, determining according to predicting the outcome whether foreground object described abnormal operation occurs.The scheme provided due to the embodiment of the present application is according to the foreground object in depth information test and monitoring video, very accurately the same position place far and near different foreground object of distance camera lens in picture can be separated, therefore, it is possible to accurately judge in scene, whether each foreground object has abnormal operation.
Accompanying drawing explanation
The specific embodiment of the application is described below with reference to accompanying drawings, wherein:
Fig. 1 shows the schematic flow sheet that in the embodiment of the present application, abnormal operation detection method is implemented;
Fig. 2 shows the motion detection block diagram of foreground object in the embodiment of the present application;
Fig. 3 shows the structural representation of abnormal operation detection device in the embodiment of the present application.
Embodiment
In order to the technical scheme and advantage that make the application are clearly understood, be described in more detail below in conjunction with the exemplary embodiment of accompanying drawing to the application, obviously, described embodiment is only a part of embodiment of the application, instead of all embodiments is exhaustive.And when not conflicting, the embodiment in this explanation and the feature in embodiment can be combined with each other.
Inventor notices in invention process:
Existing action identification method realizes based on colour picture (being also RGB image), carrys out identification maneuver according to pixel change.When there being two or more people to overlap in camera lens picture, specifically which people cannot being distinguished and create abnormal operation; And, when carrying out identification maneuver according to pixel change, easily be subject to the impact of other factors, such as, people institute habited color, decorative pattern etc., particularly when people wears pattern clothes, as long as people is action more slightly, will be there is larger change in the pixel of RGB image, thus cause erroneous judgement, and accuracy of detection is lower.
For above-mentioned deficiency, the embodiment of the present application proposes a kind of abnormal operation detection method and device, is described below.
Fig. 1 shows the schematic flow sheet that in the embodiment of the present application, abnormal operation detection method is implemented, and as shown in the figure, described abnormal operation detection method can comprise the steps:
Step 101, the foreground object detected according to depth information in monitor video;
Between step 102, calculating consecutive frame, the depth difference of described foreground object, obtains depth difference image;
Wherein, described depth difference image reflects the action at a time of described foreground object;
Step 103, the depth difference image of continuous multiple frames to be calculated, obtain depth of cure difference image;
Wherein, described depth of cure difference image reflects the action of described foreground object in section sometime;
Step 104, according to described depth of cure difference image calculated direction histogram of gradients (HOG, Histogramof Oriented Gradient) feature;
Wherein, described HOG feature represents the action vector of described foreground object;
Step 105, support vector machine (SVM by the good abnormal operation of training in advance, Support VectorMachine) sorter predicts the abnormal operation that described HOG feature is corresponding, determines whether described foreground object described abnormal operation occurs according to predicting the outcome.
Wherein, foreground object can be people, animal or other monitored object of specifying.
In the specific implementation, in the embodiment of the present application, can utilize background model that the depth value of non-foreground object position is set to infinite distance, thus reduce further interference that non-foreground object brings or operation inconvenience.
The scheme provided due to the embodiment of the present application relies on depth map (to be also, there is each two field picture in the monitor video of depth information) and depth map on human detection and tracking, according to depth information is very accurate, the same position place far and near different foreground object of distance camera lens in picture can be separated, therefore, it is possible to accurately judge in scene, whether each foreground object has abnormal operation.And, owing to depth map only showing the depth information of each point, motion detection is the change relying on depth information, instead of rely on pixel change, therefore, adopt the scheme that the embodiment of the present application provides, pattern and pure color do not have too many difference in depth map, eliminate a large amount of redundant informations compared to existing technology, and then further increase accuracy of detection.
Further, in order to solve the not high problem of the accuracy in detection that causes when foreground object is more in monitoring scene more complicated or picture, can also implement in the following manner.
In enforcement, when monitor video comprises N number of foreground object, after step 101, before step 102, described method can further include:
According to each foreground object, described monitor video is divided into N number of independently deep video, described deep video comprises the continuous action of each foreground object;
Described monitor video to be divided into several independently after deep video described by described method according to each foreground object, be specifically as follows:
Step 102 is performed to step 105 to the deep video of each foreground object.
In the embodiment of the present application, for one section of video, can first detect foreground object all in video, suppose that foreground object is behaved, everyone deep video is split, forms some sections of independently deep videos, in each section of deep video, only occur the series of actions of a people, deep video for different people carries out step such as successive depths difference calculatings etc., can more clear, facilitate and detect exactly.
In enforcement, describedly according to each foreground object, described monitor video is divided into several independently deep videos, is specifically as follows:
Determine the depth location of foreground object;
Fill (Flood fill) method by described depth location by unrestrained water, infect point adjacent with described depth location in preset range, foreground object described in monitor video is split, forms independently deep video.
In concrete enforcement, suppose that foreground object is behaved, first can determine the head position of people, by Flood fill method of the prior art, from head position, search point adjacent with head position within the scope of certain length, and these points are infected, the point distant apart from this people then can not be infected, operates thus, finally just foreground object can be split.
In enforcement, between described calculating consecutive frame, the depth difference of described foreground object, obtains depth difference image, can be specially:
The centre of gravity place of the foreground object in every two field picture is moved to picture centre, and after the foreground object between consecutive frame image is adjusted to same size, calculates the depth difference of described foreground object between consecutive frame, obtain depth difference image.
In the embodiment of the present application, can first to the pretreatment operation that two adjacent two field pictures align respectively and are out of shape.For the prospect people of every pictures, its centre of gravity place is moved to center picture, and people is adjusted to same size, and then calculate the depth difference between consecutive frame image, result is taken absolute value.
In enforcement, the described SVM classifier by the good abnormal operation of training in advance predicts the abnormal operation that described HOG feature is corresponding, determining whether described foreground object described abnormal operation occurs, can be specially according to predicting the outcome:
According to the HOG feature calculation of every section of successive frame and the degree of confidence of described foreground object generation abnormal operation, determine described foreground object generation abnormal operation when described degree of confidence is greater than predetermined threshold value.
In the embodiment of the present application, for any one section of depth of cure difference image generated, the HOG feature of this depth of cure difference image can be passed to the good SVM classifier of a training in advance, predict the degree of confidence of existing object generation abnormal operation.In concrete enforcement, the HOG feature that multistage successive frame can also be obtained carries out the calculating such as progressive mean, after obtaining accumulative degree of confidence, more accumulative degree of confidence and the threshold value preset are compared, if be greater than predetermined threshold value, determine that this object there occurs this action.
For the ease of the enforcement of the application, be described with example below.
Suppose that monitoring scene is that bank self-help is withdrawn the money business hall, due to the personnel handling self-help drawing money business in bank may have at one time multiple, there is queuing phenomena, the people before and after the people that now these are queued up in monitor video there will be partly overlap or people below by outrunner the situation of blocking.
Monitor video first can be detected all people according to depth information by the embodiment of the present application, and everyone deep video is separated, and forms independently deep video.Suppose total A, B, C tri-people handling self-help drawing money business, so just can form three independently deep videos, also, the deep video of the deep video of A, the deep video of B, C.Here suppose that video is 90 frame videos (about about 3 seconds).
Obtaining everyone independently after deep video, can utilize background model that the depth value of non-prospect people (as the object such as ATM (automatic teller machine) or door) position is set to infinite distance, for the deep video of different people, following steps can be performed respectively.
The motion detection block diagram of foreground object in the embodiment of the present application is shown in Fig. 2, as shown in the figure, has been detected as example with the abnormal operation of A below and is described.
(1) the depth difference image of consecutive frame is calculated
First the centre of gravity place of A in two two field pictures that are connected moved on to center picture and is adjusted to same size, calculate the depth difference between two two field pictures and result is taken absolute value, obtaining depth difference image.Depth difference image reflects A motion state at a time.
Such as:
The two field picture of the 1st frame and the 2nd frame obtains the 1st depth difference image;
The two field picture of the 2nd frame and the 3rd frame obtains the 2nd depth difference image;
…
The two field picture of the 89th frame and the 90th frame obtains the 89th depth difference image.
In concrete enforcement, each depth difference image can carry out the display of different depth color according to degree of depth difference size, such as, when A fastens suddenly the neck of B, can be there is action by a relatively large margin in arm, cause depth information that larger change occurs, at this moment, can by the arm of A with black or red display, other body parts of A then can show with grey.
(2) the depth of cure difference image of successive frame is calculated
Continuous print 15 depth difference image additions are averaged, depth of cure difference image can be obtained, thus reflect the operating state of A within a period of time.
Such as:
1st to the 15th depth difference image addition is averaged, obtains the 1st depth of cure difference image;
2nd to the 16th depth difference image addition is averaged, obtains the 2nd depth of cure difference image;
…
75th to the 89th depth difference image addition is averaged, obtains the 75th depth of cure difference image.
In concrete enforcement, other account form can also be adopted to obtain depth of cure difference image, be not limited in the account form being added and being averaged, the application is not restricted this.
For any frame depth difference image in video, can also by 15 two field picture regeneration depth of cure difference images after any frame depth difference image and its, also can be carried out being polymerized etc. by continuous print 20 frame, the application be all restricted the frame number which frame is polymerized and is polymerized.
(3) HOG feature is extracted to depth of cure difference image
The HOG feature of depth of cure difference image represents the quantification vector representation of current action, in concrete enforcement, the area of space of 8*8 can be used, and add up the frequency histogram in 32 directions, concrete HOG characteristic quantification can adopt mode of the prior art, does not repeat at this.
(4) predict by SVM classifier the action that HOG feature is corresponding
For the depth of cure difference image that any one section of video generates, its HOG feature the good SVM classifier of a training in advance be can be passed to, specific action occurs A degree of confidence S1, S2 ... Sn doped.
SVM classifier can the action of training in advance a lot of, carries out predicted operation separately.Can the better action thinking act of violence of training in advance, such as fasten neck, head etc. of fiercelying attack, can also the better action thinking emergency behavior of training in advance, such as wave, wave to be further subdivided into that left hand is brandished, the right hand is brandished, two hands intersect and to brandish or two hands swing in the same way.The application is that abnormal operation is not restricted for which action.
(5) judge whether accumulative degree of confidence exceedes predetermined threshold value
The embodiment of the present application can add up the degree of confidence of the above-mentioned each HOG feature calculated, and the most accumulative degree of confidence compares with the threshold value preset, when accumulative degree of confidence is greater than the threshold value preset, then judges that current event occurs.
Respectively above-mentioned (1) to (5) step is performed to B, C, the testing result that whether B abnormal operation occurs, whether C abnormal operation occurs can be obtained.
Such as: predict the HOG feature of A, the degree of confidence of the action of other people neck that finds to fasten in the HOG feature of A and SVM classifier is higher, then think that A there occurs violent action;
The HOG feature of B is predicted, finds that the degree of confidence of the action of waving in the HOG feature of B and SVM classifier is higher, then think that B there occurs emergency action.
In the embodiment of the present application, different actions (corresponding different events, such as violence and emergency) can judge separately, and the threshold value of setting also can be different, and the application is not restricted for the concrete setting of threshold value.
The embodiment of the present application can judge violence and emergency action based on deep video, and relative to the violence in traditional rgb video with wave to detect, the scheme that the embodiment of the present application provides can detect violence in picture and action of crying for help more accurately; And the embodiment of the present application can not only detect violence in video and emergency event, particular location and concrete people that violence and emergency event occur can also be oriented.
Based on same inventive concept, a kind of abnormal operation detection device is additionally provided in the embodiment of the present application, the principle of dealing with problems due to these equipment is similar to a kind of abnormal operation detection method, and therefore the enforcement of these equipment see the enforcement of method, can repeat part and repeat no more.
Fig. 3 shows the structural representation of abnormal operation detection device in the embodiment of the present application, and as shown in the figure, abnormal operation detection device can comprise:
Detection module 301, for detecting the foreground object in monitor video according to depth information;
Depth difference computing module 302, for calculating the depth difference of described foreground object between consecutive frame, obtains depth difference image;
Depth of cure difference computing module 303, for calculating the depth difference image of continuous multiple frames, obtains depth of cure difference image;
HOG feature calculation module 304, for according to described depth of cure difference image calculated direction histogram of gradients HOG feature;
According to predicting the outcome, determination module 305, for predicting by the SVM classifier of the good abnormal operation of training in advance the abnormal operation that described HOG feature is corresponding, determines whether described foreground object described abnormal operation occurs.
In enforcement, described device may further include:
Segmentation module, for when monitor video comprises N number of foreground object, according to each foreground object, described monitor video is divided into N number of independently deep video, described deep video comprises the continuous action of each foreground object;
Loop module, for being input to described depth difference computing module, depth of cure difference computing module, HOG feature calculation module and determination module successively by the deep video of described each foreground object.
In enforcement, described segmentation module is specifically for determining the depth location of foreground object; By Flood fill method by described depth location, infect point adjacent with described depth location in preset range, foreground object described in monitor video is split, forms independently deep video; Wherein, described deep video comprises the continuous action of described foreground object.
In enforcement, described depth difference computing module is specifically for moving to picture centre by the centre of gravity place of the foreground object in every two field picture, and after the foreground object between consecutive frame image is adjusted to same size, calculates the depth difference of described foreground object between consecutive frame, obtain depth difference image.
In enforcement, described determination module, specifically for according to the HOG feature calculation of every section of successive frame and the degree of confidence of described foreground object generation abnormal operation, determines described foreground object generation abnormal operation when described degree of confidence is greater than predetermined threshold value.
For convenience of description, each several part of the above device is divided into various module or unit to describe respectively with function.Certainly, the function of each module or unit can be realized in same or multiple software or hardware when implementing the application.
Those skilled in the art should understand, the embodiment of the application can be provided as method, system or computer program.Therefore, the application can adopt the form of complete hardware embodiment, completely software implementation or the embodiment in conjunction with software and hardware aspect.And the application can adopt in one or more form wherein including the upper computer program implemented of computer-usable storage medium (including but not limited to magnetic disk memory, CD-ROM, optical memory etc.) of computer usable program code.
The application describes with reference to according to the process flow diagram of the method for the embodiment of the present application, equipment (system) and computer program and/or block scheme.Should understand can by the combination of the flow process in each flow process in computer program instructions realization flow figure and/or block scheme and/or square frame and process flow diagram and/or block scheme and/or square frame.These computer program instructions can being provided to the processor of multi-purpose computer, special purpose computer, Embedded Processor or other programmable data processing device to produce a machine, making the instruction performed by the processor of computing machine or other programmable data processing device produce device for realizing the function of specifying in process flow diagram flow process or multiple flow process and/or block scheme square frame or multiple square frame.
These computer program instructions also can be stored in can in the computer-readable memory that works in a specific way of vectoring computer or other programmable data processing device, the instruction making to be stored in this computer-readable memory produces the manufacture comprising command device, and this command device realizes the function of specifying in process flow diagram flow process or multiple flow process and/or block scheme square frame or multiple square frame.
These computer program instructions also can be loaded in computing machine or other programmable data processing device, make on computing machine or other programmable devices, to perform sequence of operations step to produce computer implemented process, thus the instruction performed on computing machine or other programmable devices is provided for the step realizing the function of specifying in process flow diagram flow process or multiple flow process and/or block scheme square frame or multiple square frame.
Although described the preferred embodiment of the application, those skilled in the art once obtain the basic creative concept of cicada, then can make other change and amendment to these embodiments.So claims are intended to be interpreted as comprising preferred embodiment and falling into all changes and the amendment of the application's scope.
Claims (10)
1. an abnormal operation detection method, is characterized in that, comprises the steps:
Step 1: detect the foreground object in monitor video according to depth information;
Step 3: the depth difference calculating described foreground object between consecutive frame, obtains depth difference image;
Step 5: the depth difference image of continuous multiple frames is calculated, obtains depth of cure difference image;
Step 7: according to described depth of cure difference image calculated direction histogram of gradients HOG feature;
Step 9: predicting by the support vector machines sorter of the good abnormal operation of training in advance the abnormal operation that described HOG feature is corresponding, determining whether described foreground object described abnormal operation occurs according to predicting the outcome.
2. the method for claim 1, is characterized in that, when monitor video comprises N number of foreground object, after step 1, before step 3, comprises further:
Step 2: according to each foreground object, described monitor video is divided into N number of independently deep video, described deep video comprises the continuous action of each foreground object;
Described method, after described step 2, is specially:
Step 3, step 5, step 7 and step 9 are performed to the deep video of each foreground object.
3. method as claimed in claim 2, is characterized in that, describedly according to each foreground object, described monitor video is divided into several independently deep videos, is specially:
Determine the depth location of foreground object;
By Flood fill method by described depth location, infect point adjacent with described depth location in preset range, foreground object described in monitor video is split, forms independently deep video.
4. the method for claim 1, it is characterized in that, described step 3 is specially: the centre of gravity place of the foreground object in every two field picture is moved to picture centre, and after the foreground object between consecutive frame image is adjusted to same size, calculate the depth difference of described foreground object between consecutive frame, after depth difference is taken absolute value, obtain depth difference image.
5. the method for claim 1, it is characterized in that, described step 9 is specially: according to the HOG feature calculation of every section of successive frame and the degree of confidence of described foreground object generation abnormal operation, determine described foreground object generation abnormal operation when described degree of confidence is greater than predetermined threshold value.
6. an abnormal operation detection device, is characterized in that, comprising:
Detection module, for detecting the foreground object in monitor video according to depth information;
Depth difference computing module, for calculating the depth difference of described foreground object between consecutive frame, obtains depth difference image, and described depth difference image reflects the action at a time of described foreground object;
Depth of cure difference computing module, for calculating the depth difference image of continuous multiple frames, obtains depth of cure difference image, and described depth of cure difference image reflects the action of described foreground object in section sometime;
HOG feature calculation module, for according to described depth of cure difference image calculated direction histogram of gradients HOG feature, described HOG feature represents the action vector of described foreground object;
According to predicting the outcome, determination module, for predicting by the SVM classifier of the good abnormal operation of training in advance the abnormal operation that described HOG feature is corresponding, determines whether described foreground object described abnormal operation occurs.
7. device as claimed in claim 6, is characterized in that, comprise further:
Segmentation module, for when monitor video comprises multiple foreground object, according to each foreground object, described monitor video is divided into several independently deep videos, described deep video comprises the continuous action of each foreground object;
Loop module, for being input to described depth difference computing module, depth of cure difference computing module, HOG feature calculation module and determination module successively by the deep video of described each foreground object.
8. device as claimed in claim 7, it is characterized in that, described segmentation module is specifically for determining the depth location of foreground object; By Flood fill method by described depth location, infect point adjacent with described depth location in preset range, foreground object described in monitor video is split, forms independently deep video; Wherein, described deep video comprises the continuous action of described foreground object.
9. device as claimed in claim 6, it is characterized in that, described depth difference computing module is specifically for moving to picture centre by the centre of gravity place of the foreground object in every two field picture, and after the foreground object between consecutive frame image is adjusted to same size, calculate the depth difference of described foreground object between consecutive frame, obtain depth difference image.
10. device as claimed in claim 6, it is characterized in that, described determination module, specifically for according to the HOG feature calculation of every section of successive frame and the degree of confidence of described foreground object generation abnormal operation, determines described foreground object generation abnormal operation when described degree of confidence is greater than predetermined threshold value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510279178.9A CN104866830B (en) | 2015-05-27 | 2015-05-27 | A kind of abnormal operation detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510279178.9A CN104866830B (en) | 2015-05-27 | 2015-05-27 | A kind of abnormal operation detection method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104866830A true CN104866830A (en) | 2015-08-26 |
| CN104866830B CN104866830B (en) | 2019-01-25 |
Family
ID=53912651
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510279178.9A Active CN104866830B (en) | 2015-05-27 | 2015-05-27 | A kind of abnormal operation detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104866830B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106022284A (en) * | 2016-05-30 | 2016-10-12 | 重庆大学 | Living-alone old person abnormal behavior detecting method based on panoramic infrared camera |
| CN107424170A (en) * | 2017-06-17 | 2017-12-01 | 复旦大学 | Motion feature for detecting local anomaly behavior in monitor video automatically describes method |
| WO2018006834A1 (en) * | 2016-07-08 | 2018-01-11 | Huawei Technologies Co., Ltd. | Systems, processes and devices for occlusion detection for video-based object tracking |
| CN109309798A (en) * | 2018-08-15 | 2019-02-05 | 吴伟锋 | Data frame automatic catching mechanism |
| WO2020024691A1 (en) * | 2018-08-01 | 2020-02-06 | 阿里巴巴集团控股有限公司 | Abnormality detection method, apparatus and device in unmanned settlement scenario |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7421417B2 (en) * | 2003-08-28 | 2008-09-02 | Wisconsin Alumni Research Foundation | Input feature and kernel selection for support vector machine classification |
| CN102880444A (en) * | 2012-08-24 | 2013-01-16 | 浙江捷尚视觉科技有限公司 | Fighting detecting method based on stereoscopic vision motion field analysis |
| CN103208006A (en) * | 2012-01-17 | 2013-07-17 | 株式会社理光 | Object locomotion mode identification method and device based on depth image sequence |
-
2015
- 2015-05-27 CN CN201510279178.9A patent/CN104866830B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7421417B2 (en) * | 2003-08-28 | 2008-09-02 | Wisconsin Alumni Research Foundation | Input feature and kernel selection for support vector machine classification |
| CN103208006A (en) * | 2012-01-17 | 2013-07-17 | 株式会社理光 | Object locomotion mode identification method and device based on depth image sequence |
| CN102880444A (en) * | 2012-08-24 | 2013-01-16 | 浙江捷尚视觉科技有限公司 | Fighting detecting method based on stereoscopic vision motion field analysis |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106022284A (en) * | 2016-05-30 | 2016-10-12 | 重庆大学 | Living-alone old person abnormal behavior detecting method based on panoramic infrared camera |
| WO2018006834A1 (en) * | 2016-07-08 | 2018-01-11 | Huawei Technologies Co., Ltd. | Systems, processes and devices for occlusion detection for video-based object tracking |
| US10185877B2 (en) | 2016-07-08 | 2019-01-22 | Huawei Technologies Co., Ltd. | Systems, processes and devices for occlusion detection for video-based object tracking |
| CN107424170A (en) * | 2017-06-17 | 2017-12-01 | 复旦大学 | Motion feature for detecting local anomaly behavior in monitor video automatically describes method |
| CN107424170B (en) * | 2017-06-17 | 2020-07-28 | 复旦大学 | Motion characteristic description method for automatically detecting local abnormal behaviors in monitoring video |
| WO2020024691A1 (en) * | 2018-08-01 | 2020-02-06 | 阿里巴巴集团控股有限公司 | Abnormality detection method, apparatus and device in unmanned settlement scenario |
| US11132559B2 (en) | 2018-08-01 | 2021-09-28 | Advanced New Technologies Co., Ltd. | Abnormality detection method, apparatus, and device for unmanned checkout |
| CN109309798A (en) * | 2018-08-15 | 2019-02-05 | 吴伟锋 | Data frame automatic catching mechanism |
| CN109309798B (en) * | 2018-08-15 | 2019-05-10 | 上海极链网络科技有限公司 | Data frame automatic catching mechanism |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104866830B (en) | 2019-01-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102465532B1 (en) | Method for recognizing an object and apparatus thereof | |
| CN102073851B (en) | Method and system for automatically identifying urban traffic accident | |
| CN104866830A (en) | Abnormal motion detection method and device | |
| WO2014125882A1 (en) | Information processing system, information processing method, and program | |
| US11508157B2 (en) | Device and method of objective identification and driving assistance device | |
| CN111931599B (en) | High altitude parabolic detection method, equipment and storage medium | |
| KR101541272B1 (en) | Apparatus and method for detecting terrorism using irregular motion of peoples | |
| CN104123544A (en) | Video analysis based abnormal behavior detection method and system | |
| JP2016099941A (en) | System and program for estimating position of object | |
| CN106241534A (en) | Many people boarding abnormal movement intelligent control method | |
| KR102587046B1 (en) | Method and System for identify Vehicle number and detect their speed | |
| CN109446926A (en) | A kind of traffic monitoring method and device, electronic equipment and storage medium | |
| CN113012383A (en) | Fire detection alarm method, related system, related equipment and storage medium | |
| KR101454644B1 (en) | Loitering Detection Using a Pedestrian Tracker | |
| Carletti et al. | An efficient and effective method for people detection from top-view depth cameras | |
| CN103475800A (en) | Method and device for detecting foreground in image sequence | |
| Liu et al. | Abnormal crowd behavior detection based on optical flow and dynamic threshold | |
| CN109086725A (en) | Hand tracking and machine readable storage medium | |
| Hamdi et al. | Hybrid deep learning and HOF for Anomaly Detection | |
| Patel et al. | Human tracking in video surveillance | |
| CN109977796A (en) | Trail current detection method and device | |
| KR20150088613A (en) | Apparatus and method for detecting violence situation | |
| CN109800684A (en) | The determination method and device of object in a kind of video | |
| CN109523573A (en) | The tracking and device of target object | |
| CN107665495B (en) | Object tracking method and object tracking device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 100092 Beijing Haidian District Yongtaizhuang North Road No. 1 Tiandi Adjacent to Block B, Building 1, Fengji Industrial Park Applicant after: BEIJING DEEPGLINT INFORMATION TECHNOLOGY CO., LTD. Address before: 100091 No. 6 Yudai Road, Haidian District, Beijing Applicant before: BEIJING DEEPGLINT INFORMATION TECHNOLOGY CO., LTD. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address |
Address after: 100192 Block B, Building 1, Tiandi Adjacent to Maple Industrial Park, No. 1, North Yongtaizhuang Road, Haidian District, Beijing Patentee after: Beijing gelingshentong Information Technology Co.,Ltd. Address before: 100092 Beijing Haidian District Yongtaizhuang North Road No. 1 Tiandi Adjacent to Block B, Building 1, Fengji Industrial Park Patentee before: BEIJING DEEPGLINT INFORMATION TECHNOLOGY Co.,Ltd. |
|
| CP03 | Change of name, title or address |