CN102222349B - Prospect frame detecting method based on edge model - Google Patents

Prospect frame detecting method based on edge model Download PDF

Info

Publication number
CN102222349B
CN102222349B CN 201110185415 CN201110185415A CN102222349B CN 102222349 B CN102222349 B CN 102222349B CN 201110185415 CN201110185415 CN 201110185415 CN 201110185415 A CN201110185415 A CN 201110185415A CN 102222349 B CN102222349 B CN 102222349B
Authority
CN
China
Prior art keywords
frame
edge
sequence
prospect
foreground
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
CN 201110185415
Other languages
Chinese (zh)
Other versions
CN102222349A (en
Inventor
朱伟兴
纪滨
李新城
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiangsu University
Original Assignee
Jiangsu University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Jiangsu University filed Critical Jiangsu University
Priority to CN 201110185415 priority Critical patent/CN102222349B/en
Publication of CN102222349A publication Critical patent/CN102222349A/en
Application granted granted Critical
Publication of CN102222349B publication Critical patent/CN102222349B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Abstract

The invention discloses a prospect frame detecting method based on an edge model, which is used for prospect frame determination in the analysis of a security monitoring video frame sequence. The method comprises the following steps of: extracting edge frames only containing edge images from sequence frames by means of a pseudo-sphere edge detection operator; counting the presence probability of edge points of the edge frames in frame sequence statistic time, marking a background attribute and a foreground attribute based on a determination condition of distinguishing according to the attributes of the edge pixel points within the current edge frame; in a foreground edge image, if the number of the edge points connected together in the current frame is less than or equal to 2, determining that the points are noise points and removing these points; and if the total number of pixels of the left foreground edge images is less than a noise determination threshold of the sequence frame, determining that the frames are background frames, otherwise, foreground frames. The method of the invention reduces the calculation cost, effectively adapts to the cases of background light change, slow object motion or short stagnation and the like, and creates favourable conditions for subsequent object motion analysis.

Description

A kind of prospect frame detection method based on edge model
Technical field
The present invention relates to machine vision technique, the prospect frame that is specifically related to be applied in the analysis of safety monitoring sequence of frames of video is judged.
Background technology
It is to detect whether have motion or the slow technology of the foreground target of mobile or short stay in the video that the prospect frame detects, at present in the video analysis process, widely used is that the less frame difference method of calculation cost and background method are (referring to Herrero S, Besc ó s J. Background Subtraction Techniques:Systematic Evaluation and Comparative Analysis[C] //Proceedings of the 11th International Conference on Advanced Concepts for Intelligent Vision Systems.Springer-Verlag. 2009,5807/2009:33-42).Frame difference method is responsive to noise ratio, is difficult for detecting slowly mobile or temporary transient static foreground target of fixed area; And the background method is to the relative robust of noise, but affected greatly by illumination variation, foreground target detects wayward, therefore, for addressing these problems, in recent years based on the algorithm of profile neighborhood information such as Snake model (referring to: Nie Xuan, Zhao Rongchun, Shen Yaping. the moving target profile based on the Snake technology extracts [J]. computer engineering. 2005,31 (23): 148-150), level set (referring to: Gong Yongyi, Luo Xiaonan, Huang Hui etc. the multiple goal profile based on single level set extracts [J]. Chinese journal of computers. 2007,30 (001): 120-128) etc., although it is better to be used for the foreground detection result, but computation complexity is high, is difficult to reach live effect.
Summary of the invention
The objective of the invention is: overcome the slow even temporary transient defective of stagnating of speed that general foreground detection method can not adapt to the variation of scene light and foreground moving target, a kind of simple, real-time prospect frame detection method based on edge model that calculates is provided.
Technical scheme of the present invention adopts following steps: (1) adopts under same fixedly camera position, the Same Scene picked-up to need that the prospect frame detects, resolution to be
Figure 2011101854157100002DEST_PATH_IMAGE002
The particular detection video, only contain the edge frame of edge image with pseudo-ball edge detection operator abstraction sequence frame; (2) current the of the described edge frame of statistics tThe frame border pixel ( I, j) at the frame sequence timing statistics
Figure 2011101854157100002DEST_PATH_IMAGE004
Interior probability of occurrence
Figure 2011101854157100002DEST_PATH_IMAGE006
,
Figure 2011101854157100002DEST_PATH_IMAGE008
Figure 2011101854157100002DEST_PATH_IMAGE010
Be kThe corresponding binaryzation edge image of frame; (3) Rule of judgment of distinguishing according to current edge frame inward flange pixel attribute
Figure 2011101854157100002DEST_PATH_IMAGE012
,
Figure 2011101854157100002DEST_PATH_IMAGE014
tEdge pixel point on the corresponding binaryzation edge image of frame, 0 is labeled as background attribute, and 1 is labeled as the prospect attribute,
Figure 2011101854157100002DEST_PATH_IMAGE016
Be background edge pixel decision threshold; (4) in the foreground edge image, if current tCounting and be less than or equal to 2 in the edge that links to each other in the frame, then removes for noise spot; To the sum of all pixels of last foreground edge image less than sequence frame noise edge judgment threshold
Figure 2011101854157100002DEST_PATH_IMAGE018
, be judged to and be background frames, otherwise be the prospect frame.
The invention has the beneficial effects as follows;
1, the present invention is based on edge model the prospect frame detection method theoretical foundation be that the stability of edge in frame sequence of foreground target is far away not as good as background edge, for the foreground moving analysis provides necessary, a small amount of motion pixel, greatly reduced calculation cost.
2, the present invention not only applicable non-rigid object prospect frame judge that the prospect frame that also is applicable to rigid-object judges, can adapt to effectively that background light changes and target travel is slow or the situation such as of short duration delay, for follow-up target motion analysis creates favorable conditions.
Description of drawings
Below in conjunction with the drawings and specific embodiments the present invention is described in further detail.
Fig. 1 is the process flow diagram of prospect frame detection method of the present invention.
Embodiment
Further specify technical scheme of the present invention below in conjunction with accompanying drawing:
Referring to shown in Figure 1, at first, extract video sample, adopt the particular detection video that picked-up needs the prospect frame to detect under same fixedly camera position, the Same Scene, resolution Be 320 * 240, the implication of scene certain fixing geographic area of referring to make a video recording herein, different scene video parameter values is different, but under the Same Scene, the frame sequence timing statistics
Figure 341713DEST_PATH_IMAGE004
, background edge pixel decision threshold
Figure 121450DEST_PATH_IMAGE016
With sequence frame noise edge judgment threshold
Figure 985501DEST_PATH_IMAGE018
These three parameter values are stable.The particular detection video that needs are detected manually intercepts.Then, the particular detection sequence of frames of video that needs are detected adopts pseudo-ball operator edge detection operator to process, and extracts resolution
Figure 268714DEST_PATH_IMAGE002
Be the edge image of 320 * 240 sequence frames, obtain only to contain the edge frame of edge image.
Before carrying out the detection of prospect frame, must obtain this three parameter values.
The frame sequence timing statistics
Figure 91177DEST_PATH_IMAGE004
Definite method be:
Comprise the of short duration delay of foreground target in the particular detection sequence of frames of video of artificial intercepting
Figure 2011101854157100002DEST_PATH_IMAGE020
Individual sample video-frequency band, and there are not the variation of scene light in these sample video-frequency bands is namely chosen and is not contained that light changes but prospect sample video-frequency band that of short duration delay campaign can be arranged.Adopt instantaneous frame difference method (referring to Herrero S to each sample video-frequency band, Besc ó s J. Background Subtraction Techniques:Systematic Evaluation and Comparative Analysis[C] //Proceedings of the 11th International Conference on Advanced Concepts for Intelligent Vision Systems. Springer-Verlag. 2009,5807/2009:33-42) carry out motion detection, utilize the characteristics of the static prospect of instantaneous frame difference method None-identified, obtain temporary transient actionless foreground target by the time period of background absorption
Figure 2011101854157100002DEST_PATH_IMAGE022
As the hold-up time.The present frame foreground target area that after the consecutive frame difference, obtains , it accounts for the two field picture total area
Figure 2011101854157100002DEST_PATH_IMAGE026
Ratio
Figure 2011101854157100002DEST_PATH_IMAGE028
Less than certain less threshold value
Figure 2011101854157100002DEST_PATH_IMAGE030
When (preferred 0.07), think that then prospect is detained or do not have a prospect, present frame attribute
Figure 2011101854157100002DEST_PATH_IMAGE032
Be labeled as 0, otherwise be 1, be i.e. the prospect frame of motion.Add up in every section video
Figure 681513DEST_PATH_IMAGE032
Be that 0 length is as the time period continuously In all video-frequency band samples, select the maximum time period
Figure 18134DEST_PATH_IMAGE022
As maximum time period (preferred 440), see formula (1):
Figure 2011101854157100002DEST_PATH_IMAGE036
（1）
In theory test Individual sample video hop count more better owing to can't obtain globally optimal solution, and there is local motion in non-rigid object, it is very short really to be detained the actionless time, so
Figure 172352DEST_PATH_IMAGE020
Value depend on that whether comprising maximum prospect in the video-frequency band that subjective judgement intercepts is detained the interval, in the practice
Figure 378205DEST_PATH_IMAGE020
Desirable 5～10(preferred 10); And the existence of camera system electronic noise affects the stability that real scene is mapped to two dimensional image edge pixel point, so calculate maximum time period
Figure 901590DEST_PATH_IMAGE034
Need to add again a modified value As final frame sequence timing statistics
Figure 767653DEST_PATH_IMAGE004
Value, see formula (2),
Figure 2011101854157100002DEST_PATH_IMAGE040
Relevant with camera system, the different video acquisition system of size of value, modified value Inconsistent, general
Figure 222085DEST_PATH_IMAGE040
Value is 10～20, and is preferred
Figure 865556DEST_PATH_IMAGE040
Value is 16(20FPS).It also is final frame sequence timing statistics
Figure 669564DEST_PATH_IMAGE004
Be 456, the motionless state of stopping of non-rigid body foreground target in the expression scene, the longest less than 22.8 seconds.If foreground target is rigid body, and it is in a single day static when just being counted as background, Can be made as 0.
Figure 2011101854157100002DEST_PATH_IMAGE042
（2）
Background edge pixel decision threshold
Figure 414983DEST_PATH_IMAGE016
Definite method be:
1 the sample video-frequency band that does not comprise foreground target and do not exist scene light to change in the particular video frequency frame sequence of artificial intercepting, adopt pseudo-ball edge detection operator (referring to Wang Zhiheng to the sample video-frequency band, Wu Fuchao. pseudosphere filter and rim detection [J]. Journal of Software. 2008,19 (4): 803-816) edge image of the sequence frame of extraction video sample, the herein scale parameter of pseudo-ball operator
Figure 2011101854157100002DEST_PATH_IMAGE044
Get 3.0, the edge keeps parameter
Figure 2011101854157100002DEST_PATH_IMAGE046
Get 0.1, template size is 5 * 5.At the frame sequence timing statistics
Figure 818020DEST_PATH_IMAGE004
Under the prerequisite of determining, the pixel that calculates on each frame border at the probability that same position occurs is
Figure DEST_PATH_IMAGE048
, establish when the tFrame background edge sum of all pixels is P, in the Making by Probability Sets that then each point occurs therein, get minimum threshold
Figure DEST_PATH_IMAGE050
For judging the threshold value of target context edge pixel, see formula (3):
Figure DEST_PATH_IMAGE052
(3)
For obtaining preferably threshold value of adaptability, need from the tFrame begins follow-on test nFrame (being taken as 15 here), until the T+nFrame, for avoiding exceptional value, nIndividual
Figure 996192DEST_PATH_IMAGE050
By from small to large ordering, get the median conduct
Figure 415672DEST_PATH_IMAGE016
(preferred 0.644) sees formula (4).
Figure DEST_PATH_IMAGE054
(4)
If video to be detected is current tThe edge pixel point of frame Greater than
Figure 235860DEST_PATH_IMAGE016
The time, edge and removing as a setting then, remaining being comprises noise at interior foreground edge image.Different video scenes is suitable
Figure 555721DEST_PATH_IMAGE016
And inconsistent.
Sequence frame noise edge judgment threshold
Figure 967111DEST_PATH_IMAGE018
Definite method be:
Artificial intercepting frame number is from the sample video-frequency band
Figure DEST_PATH_IMAGE058
(preferred
Figure DEST_PATH_IMAGE060
Frame) do not contain promising video-frequency band
Figure DEST_PATH_IMAGE062
(preferred Volume), right Individual sample video-frequency band is respectively through the edge image of pseudo-ball edge detection operator abstraction sequence frame, the edge that keeps non-background, calculate the image that obtains through formula (8), formula (9) and be the noise image that does not contain prospect, removal is connected to 2 and reaches isolated noise edge pixel again, add up at last the remaining edge pixel point quantity of every frame, obtain
Figure DEST_PATH_IMAGE066
Individual discrete value, statistics the iFrame border pixel quantity By from small to large ordering, noise edge pixel quantity statistical property meets Gaussian distribution, adopts median method approximate the iFrame border pixel quantity
Figure 169870DEST_PATH_IMAGE068
Expectation value
Figure DEST_PATH_IMAGE070
With
Figure 111281DEST_PATH_IMAGE018
Value is (preferred
Figure 192108DEST_PATH_IMAGE018
Value is 35), see formula (5) and formula (7).
Figure DEST_PATH_IMAGE072
（5）
Figure DEST_PATH_IMAGE074
（6）
Figure DEST_PATH_IMAGE076
（7）
Then, statistics frame sequence timing statistics
Figure 523863DEST_PATH_IMAGE004
The probability distribution of interior all edge pixels is namely added up the edge pixel point of edge frame at the frame sequence timing statistics
Figure 748171DEST_PATH_IMAGE004
In probability of occurrence, establish present frame and be the tFrame is at the frame sequence timing statistics
Figure 278509DEST_PATH_IMAGE004
Under the prerequisite of determining,
Figure 31702DEST_PATH_IMAGE048
Current tThe frame border pixel ( I, j) probability, computing formula is seen formula (8).
Figure 381912DEST_PATH_IMAGE006
，
Figure 409910DEST_PATH_IMAGE008
（8）
Carry out the judgement of prospect frame in the frame sequence according to formula (8), by background edge pixel decision threshold
Figure 558870DEST_PATH_IMAGE016
Extract the foreground edge image, the frame sequence timing statistics greater than Video-frequency band in carry out the sport foreground frame and judge, wherein, Be kThe corresponding binaryzation edge image of frame.The probability of occurrence of background edge pixel in theory
Figure 152159DEST_PATH_IMAGE048
Inevitable greater than active margin pixel probability, namely at the frame sequence timing statistics
Figure 922669DEST_PATH_IMAGE004
In, if greater than background edge pixel decision threshold
Figure 752085DEST_PATH_IMAGE016
, this edge pixel point
Figure DEST_PATH_IMAGE078
Be labeled as 0, show frame border
Figure 76887DEST_PATH_IMAGE010
On time of occurring in same position of point long belong to background; Otherwise be labeled as 1, what the time that shows appearance was relatively shorter is the activity prospect, sees formula (9).
Figure 446688DEST_PATH_IMAGE012
（9）
Reservation element marking value is 1 pixel in edge image, is the foreground edge image of extraction.
At last, in described foreground edge image, according to sequence frame noise edge Rule of judgment, the elimination noise pixel obtains the foreground target edge of frame.Whether be noise spot, if the edge pixel that links to each other in the present frame is counted if analyzing each foreground edge pixel according to neighborhood information
Figure DEST_PATH_IMAGE080
Be less than or equal to 2, then be considered as noise spot and remove; Linking to each other counts surpasses 2 noise possibility but still exist in theory, thus last foreground edge sum of all pixels and sequence frame noise edge judgment threshold are compared, to last foreground edge sum of all pixels threshold value
Figure 773502DEST_PATH_IMAGE018
Judge that this frame detects and is background frames when being worth less than this, otherwise is the prospect frame, keeps the prospect frame.
Even the present invention's scene light changes and foreground target speed is slow even time-out, do not affect the detection of prospect frame yet, when detecting present frame and be the prospect frame, do the time spent as the monitoring security protection and can send early warning, or when analyzing as foreground moving, for subsequent act research provides a small amount of, necessary foreground target information.

Claims (4)

1. prospect frame detection method based on edge model is characterized in that adopting following steps:
(1) picked-up needs that the prospect frame detects, resolution to be under same fixedly camera position, Same Scene
Figure 2011101854157100001DEST_PATH_IMAGE001
The particular detection video, adopt pseudo-ball edge detection operator to process to the frame sequence of described particular detection video, extract the edge image of described frame sequence, obtain only to contain the edge frame of edge image;
(2) current the of the described edge frame of statistics tThe frame border pixel ( I, j) at the frame sequence timing statistics Interior probability of occurrence
Figure 2011101854157100001DEST_PATH_IMAGE003
,
Figure 2011101854157100001DEST_PATH_IMAGE004
Figure 2011101854157100001DEST_PATH_IMAGE005
Be kThe corresponding binaryzation edge image of frame;
(3) distinguish current edge frame inward flange pixel, the Rule of judgment that current edge frame inward flange pixel attribute is distinguished is
Figure 2011101854157100001DEST_PATH_IMAGE006
,
Figure 2011101854157100001DEST_PATH_IMAGE007
tEdge pixel point on the corresponding binaryzation edge image of frame, 0 is labeled as background attribute, and 1 is labeled as the prospect attribute,
Figure 2011101854157100001DEST_PATH_IMAGE008
Be background edge pixel decision threshold;
(4) image that keeps the edge pixel point of described prospect attribute in edge image is the foreground edge image, in the foreground edge image, if current the tCounting and be less than or equal to 2 in the edge that links to each other in the frame, then removes for noise spot; If the sum of all pixels of last foreground edge image is less than sequence frame noise edge judgment threshold
Figure 2011101854157100001DEST_PATH_IMAGE009
, then this frame is background frames, otherwise is the prospect frame.
2. a kind of prospect frame detection method based on edge model according to claim 1 is characterized in that: the described frame sequence timing statistics of step (2)
Figure 408566DEST_PATH_IMAGE002
Determine by the following method:
1) manually intercepts and comprise the of short duration delay of foreground target in the particular detection sequence of frames of video and do not exist scene light to change
Figure 856865DEST_PATH_IMAGE010
Individual sample video-frequency band;
2) right
Figure 43126DEST_PATH_IMAGE010
Individual sample video-frequency band is obtained temporary transient actionless foreground target frame number through instantaneous frame difference method respectively
Figure 2011101854157100001DEST_PATH_IMAGE011
As the hold-up time, the selection maximum
Figure 2011101854157100001DEST_PATH_IMAGE012
As
Figure 2011101854157100001DEST_PATH_IMAGE013
3) frame sequence timing statistics
Figure 2011101854157100001DEST_PATH_IMAGE014
, wherein
Figure 2011101854157100001DEST_PATH_IMAGE015
,
Figure 2011101854157100001DEST_PATH_IMAGE016
The modified value relevant with camera system.
3. a kind of prospect frame detection method based on edge model according to claim 1 is characterized in that: the described background edge pixel of step (3) decision threshold
Figure 2011101854157100001DEST_PATH_IMAGE018
Determine by the following method:
1) manually intercepts 1 the sample video-frequency band that does not comprise foreground target in the particular detection sequence of frames of video and do not exist scene light to change;
2) to the edge image of sample video-frequency band through pseudo-ball edge detection operator abstraction sequence frame, at the frame sequence timing statistics
Figure 2011101854157100001DEST_PATH_IMAGE020
Under the prerequisite of determining, calculate the probability that the pixel on each frame border occurs at same position
Figure 2011101854157100001DEST_PATH_IMAGE021
, when tFrame background edge sum of all pixels is P, get the probability minimum value
Figure 2011101854157100001DEST_PATH_IMAGE022
3) from tFrame begins follow-on test nFrame, until the T+nFrame, nIndividual
Figure 2011101854157100001DEST_PATH_IMAGE024
By from small to large ordering, median is background edge pixel decision threshold
Figure 2011101854157100001DEST_PATH_IMAGE025
4. a kind of prospect frame detection method based on edge model according to claim 1 is characterized in that: the described sequence frame noise edge of step (4) judgment threshold Determine by the following method:
1) manually intercepting the frame number that does not comprise prospect in the particular detection sequence of frames of video is
Figure 2011101854157100001DEST_PATH_IMAGE026
Figure DEST_PATH_IMAGE027
Individual sample video-frequency band is right
Figure 511DEST_PATH_IMAGE027
Individual sample video-frequency band keeps the edge of non-background respectively through the edge image of pseudo-ball edge detection operator abstraction sequence frame;
2) the remaining edge pixel point quantity of the every frame of statistics, the iFrame border pixel quantity
Figure 2011101854157100001DEST_PATH_IMAGE028
By from small to large ordering, obtain
Figure 2011101854157100001DEST_PATH_IMAGE029
Individual discrete value;
3) determine sequence frame noise edge judgment threshold
Figure 2011101854157100001DEST_PATH_IMAGE030
, wherein:
Figure 2011101854157100001DEST_PATH_IMAGE031
,
Figure 2011101854157100001DEST_PATH_IMAGE032
CN 201110185415 2011-07-04 2011-07-04 Prospect frame detecting method based on edge model Expired - Fee Related CN102222349B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201110185415 CN102222349B (en) 2011-07-04 2011-07-04 Prospect frame detecting method based on edge model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 201110185415 CN102222349B (en) 2011-07-04 2011-07-04 Prospect frame detecting method based on edge model

Publications (2)

Publication Number Publication Date
CN102222349A CN102222349A (en) 2011-10-19
CN102222349B true CN102222349B (en) 2013-01-23

Family

ID=44778893

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201110185415 Expired - Fee Related CN102222349B (en) 2011-07-04 2011-07-04 Prospect frame detecting method based on edge model

Country Status (1)

Country Link
CN (1) CN102222349B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106384348A (en) * 2016-09-09 2017-02-08 海信集团有限公司 Monitor image anomaly detection method and device

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102494976B (en) * 2011-11-18 2014-04-09 江苏大学 Method for automatic measurement and morphological classification statistic of ultra-fine grain steel grains
CN103679745B (en) * 2012-09-17 2016-08-17 浙江大华技术股份有限公司 A kind of moving target detecting method and device
CN104135597B (en) * 2014-07-04 2017-12-15 上海交通大学 A kind of video jitter automatic testing method
CN111726620A (en) * 2019-03-22 2020-09-29 浙江宇视科技有限公司 Encoding method and device for monitoring video background frame, electronic equipment and medium
CN114037633B (en) * 2021-11-18 2022-07-15 南京智谱科技有限公司 Infrared image processing method and device
CN113822879B (en) * 2021-11-18 2022-06-21 南京智谱科技有限公司 Image segmentation method and device

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101582160A (en) * 2009-06-18 2009-11-18 北京中星微电子有限公司 Foreground detection method and device as well as adaptive threshold adjusting method and device
CN101673404A (en) * 2009-10-19 2010-03-17 北京中星微电子有限公司 Target detection method and device

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060239506A1 (en) * 2005-04-25 2006-10-26 Objectvideo, Inc. Line textured target detection and tracking with applications to "Basket-run" detection

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101582160A (en) * 2009-06-18 2009-11-18 北京中星微电子有限公司 Foreground detection method and device as well as adaptive threshold adjusting method and device
CN101673404A (en) * 2009-10-19 2010-03-17 北京中星微电子有限公司 Target detection method and device

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
王志衡等.伪球滤波和边缘检测.《软件学报》.2008,第19卷(第4期),803-816. *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106384348A (en) * 2016-09-09 2017-02-08 海信集团有限公司 Monitor image anomaly detection method and device
CN106384348B (en) * 2016-09-09 2019-06-14 海信集团有限公司 The method for detecting abnormality and device of monitoring image

Also Published As

Publication number Publication date
CN102222349A (en) 2011-10-19

Similar Documents

Publication Publication Date Title
CN102222349B (en) Prospect frame detecting method based on edge model
CN107330372B (en) Analysis method of video-based crowd density and abnormal behavior detection system
CN102307274B (en) Motion detection method based on edge detection and frame difference
Tripathi et al. A probabilistic approach for detection and removal of rain from videos
US8355079B2 (en) Temporally consistent caption detection on videos using a 3D spatiotemporal method
CN103679745B (en) A kind of moving target detecting method and device
US20130021477A1 (en) Method and camera for determining an image adjustment parameter
CN105469105A (en) Cigarette smoke detection method based on video monitoring
US20180144476A1 (en) Cascaded-time-scale background modeling
CN102542289A (en) Pedestrian volume statistical method based on plurality of Gaussian counting models
CN102111532A (en) Camera lens occlusion detecting system and method
CN104166983A (en) Motion object real time extraction method of Vibe improvement algorithm based on combination of graph cut
CN103279737A (en) Fight behavior detection method based on spatio-temporal interest point
CN104835147A (en) Method for detecting crowded people flow in real time based on three-dimensional depth map data
CN106991418B (en) Winged insect detection method and device and terminal
CN104766079A (en) Remote infrared weak object detecting method
CN110880184A (en) Method and device for carrying out automatic camera inspection based on optical flow field
CN103473533A (en) Video motion object abnormal behavior automatic detection method
CN113706566A (en) Perfuming spray performance detection method based on edge detection
Furuya et al. Road intersection monitoring from video with large perspective deformation
CN103400395A (en) Light stream tracking method based on HAAR feature detection
CN102663776A (en) Violent movement detection method based on characteristic point analysis and device thereof
CN101877135B (en) Moving target detecting method based on background reconstruction
CN113920585A (en) Behavior recognition method and device, equipment and storage medium
KR20180096096A (en) Coastline monitoring apparatus and method using ocean color image

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20130123

Termination date: 20150704

EXPY Termination of patent right or utility model