CN104463902B - A kind of static target removing method based on NMI features - Google Patents

A kind of static target removing method based on NMI features Download PDF

Info

Publication number
CN104463902B
CN104463902B CN201310452893.9A CN201310452893A CN104463902B CN 104463902 B CN104463902 B CN 104463902B CN 201310452893 A CN201310452893 A CN 201310452893A CN 104463902 B CN104463902 B CN 104463902B
Authority
CN
China
Prior art keywords
mrow
munderover
msub
sigma
msup
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.)
Active
Application number
CN201310452893.9A
Other languages
Chinese (zh)
Other versions
CN104463902A (en
Inventor
杨文佳
柴智
李亚鹏
李香祯
王楠
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.)
Beijing Institute of Environmental Features
Original Assignee
Beijing Institute of Environmental Features
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 Beijing Institute of Environmental Features filed Critical Beijing Institute of Environmental Features
Priority to CN201310452893.9A priority Critical patent/CN104463902B/en
Publication of CN104463902A publication Critical patent/CN104463902A/en
Application granted granted Critical
Publication of CN104463902B publication Critical patent/CN104463902B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence

Landscapes

  • Image Analysis (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)

Abstract

The invention belongs to photovoltaic applied technical field, and in particular to one kind is based on the static target removing method of NMI (Normalized Moment of Inertia) feature.The method that this algorithm accumulates frame difference first with more Gaussian Background modeling methods and three frames obtains moving target foreground area, is designated as Fgb and Fgd.Then foreground area mark is carried out to Fgb and Fgd respectively.This algorithm uses NMI (Normalized Moment of Inertia) invariant feature of image, i.e. the normalization rotary inertia of image is as characteristic matching parameter, NMI characteristic matchings accuracy is high, amount of calculation is small, speed is fast, and has anti-tonal distortion and have to geometric distortion and preferably maintain the invariance.

Description

A kind of static target removing method based on NMI features
Technical field
The invention belongs to photovoltaic applied technical field, and in particular to one kind is based on NMI (Normalized Moment Of Inertia) feature static target removing method.
Background technology
Intelligent Target is analyzed and video image searching system is that real-time intelligent video monitoring system further expands.The system Including：Man-machine interaction unit, dispatch control unit, intellectual analysis unit and resource database, complete the decoding of video, analysis, Pay close attention to information extraction and event trace retrieval.System receives User Format retrieval and inquisition condition, reads video by network and deposits Store up the related massive video data in equipment, decode the video of different coding form, by the intellectual analysis to video content and Concentration extraction, obtain user and pay close attention to target and realize target following and target association, fitting forms the space history of concern target Track, the query and search work of clue target, the time required to shortening target retrieval, improves effect of solving a case during mitigation case detecting Rate.
How object extraction interested to be come out from sequence of video images, be intelligent Target analysis and video image An initial and most important step in searching system.The validity of video frequency motion target detection algorithm directly influences follow-up system The treatment effect of system, in the application of reality, because moving target exists and from motion state is changed into static shape in complex scene The possibility of state.The change of this state easily causes the error detection of moving target, so as to be brought directly to the intellectual analysis of video The influence connect.
Therefore, a kind of effective ways that can eliminate static target are found, are the vital tasks for improving Design of System Software.
The content of the invention
The technical problem to be solved in the present invention is to provide a kind of image realized mobile target in complex background and accurately detected Processing Algorithm, intelligent Target analysis and video image searching system are mainly used in, in the moving target inspection based on background modeling In survey technology, eliminate moving target from motion state be changed into inactive state when background residual region, realize in complex background Under moving target accurately detect, so as to improve the accuracy of video frequency search system intellectual analysis.
In order to realize this purpose, the present invention adopts the technical scheme that：
A kind of static target removing method based on NMI features, comprises the following steps：
1) current video sequence image is gathered under capture apparatus inactive state；
2) using more gauss hybrid models algorithms detection moving target, moving target foreground picture Fgb is obtained；
3) using three frames accumulation difference algorithm detection moving target, moving target foreground picture Fgd is obtained；
4) barycenter (x of foreground area in Fgb figures is calculatedo,yo), if area size is (M, N), f (x, y) is in image position Put the value at (x, y) place, prospect 1, background 0：
5) calculate and surround barycenter (x in Fgb figureso,yo) rotary inertia be designated as
6) rotary inertia after being normalized in Fgb figures is calculated；
7) rotary inertia is normalized in the foreground area Fgd got using identical computational methods to three frame cumulative errors Calculating, obtain NMI2;
8) NMI1 and NMI2 alpha value calculateds are utilized：
9) when α values are less than threshold value Th, then regional background remnants regions are judged, without mark.
Further, a kind of static target removing method based on NMI features as described above, in step 9), threshold value Th Chosen according to specific video scene by testing.
Static target technology for eliminating of the technical solution of the present invention based on NMI features, by background modeling and multiple frame cumulation While difference, choose NMI features and the background residual of static target is differentiated, the technology can eliminate target from motion shape State is changed into background residual region during inactive state, realizes that the moving target under complex background accurately detects.By to street Visible light sequential image under environment is tested, it was demonstrated that the algorithm has a good effect.
Embodiment
Technical solution of the present invention is further elaborated below.
The technical principle of the present invention --- when being detected using more Gaussian Background modeling techniques to moving target, due to the back of the body The influence of scape model update method and renewal rate, in moving target foreground segmentation figure can using static target as move before Scape (i.e. background residual) and produce error detection problem.This algorithm is poor first with more Gaussian Background modeling methods and three frames accumulation frame The method divided obtains moving target foreground area, is designated as Fgb and Fgd.Then foreground area mark is carried out to Fgb and Fgd respectively. Due to target motion in itself and the influence of extraneous factor, shape, size, the light and shade of target can constantly change.Before being directly based upon The matching precision of scene area template matching method is not high, and the extracting method of traditional image invariant features is complicated, computationally intensive, right Have a great impact in the real-time of massive video searching system.This algorithm uses NMI (the Normalized Moment of image Of Inertia) invariant feature, i.e., the normalization rotary inertia of image is as characteristic matching parameter, NMI characteristic matching accuracy Height, amount of calculation is small, speed is fast, and has anti-tonal distortion and have to geometric distortion and preferably maintain the invariance.
The concrete application object of the present invention --- intelligent Target is analyzed and video image searching system, calculation proposed by the present invention Method uses C Plus Plus programming realization under VC6.0 platforms, in order to realize moving object detection and the static mesh under complex scene Mark eliminates, and comprises the following steps：
1) current video sequence image is gathered under capture apparatus inactive state；
2) using more gauss hybrid models algorithms detection moving target, moving target foreground picture Fgb is obtained；
3) using three frames accumulation difference algorithm detection moving target, moving target foreground picture Fgd is obtained；
4) barycenter (x of foreground area in Fgb figures is calculatedo,yo), if area size is (M, N), f (x, y) is in image position Put the value at (x, y) place, prospect 1, background 0：
5) calculate and surround barycenter (xo,yo) rotary inertia be designated as
6) rotary inertia, i.e. NMI1 after normalizing are calculated；
7) rotary inertia is normalized in the foreground area Fgd got using identical computational methods to three frame cumulative errors Calculating, obtain NMI2;
8) NMI1 and NMI2 alpha value calculateds are utilized：
9) when α values are less than threshold value Th（Threshold value Th is chosen according to specific video scene by testing）.Then judge the region Background residual region, without mark.

Claims (1)

  1. A kind of 1. static target removing method based on NMI features, it is characterised in that：Comprise the following steps：
    1) current video sequence image is gathered under capture apparatus inactive state；
    2) using more gauss hybrid models algorithms detection moving target, moving target foreground picture Fgb is obtained；
    3) using three frames accumulation difference algorithm detection moving target, moving target foreground picture Fgd is obtained；
    4) barycenter (x of foreground area in Fgb figures is calculatedo,yo), if area size is M × N, f (x, y) be picture position (x, Y) value at place, prospect 1, background 0：
    <mrow> <msub> <mi>x</mi> <mi>o</mi> </msub> <mo>=</mo> <mfrac> <mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>x</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>M</mi> </munderover> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>y</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <mi>x</mi> <mo>&amp;times;</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> </mrow> <mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>x</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>M</mi> </munderover> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>y</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> </mrow> </mfrac> <msub> <mi>y</mi> <mi>o</mi> </msub> <mo>=</mo> <mfrac> <mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>x</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>M</mi> </munderover> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>y</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <mi>y</mi> <mo>&amp;times;</mo> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> </mrow> <mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>x</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>M</mi> </munderover> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>y</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> </mrow> </mfrac> <mo>;</mo> </mrow>
    5) calculate and surround barycenter (x in Fgb figureso,yo) rotary inertia be designated as
    <mrow> <msub> <mi>J</mi> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>o</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>o</mi> </msub> <mo>)</mo> </mrow> </msub> <mo>=</mo> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>x</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>M</mi> </munderover> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>y</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <mo>|</mo> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>-</mo> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>o</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>o</mi> </msub> <mo>)</mo> </mrow> <msup> <mo>|</mo> <mn>2</mn> </msup> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>x</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>M</mi> </munderover> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>y</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <mrow> <mo>(</mo> <msup> <mrow> <mo>(</mo> <mrow> <mi>x</mi> <mo>-</mo> <msub> <mi>x</mi> <mi>o</mi> </msub> </mrow> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mo>+</mo> <msup> <mrow> <mo>(</mo> <mrow> <mi>y</mi> <mo>-</mo> <msub> <mi>y</mi> <mi>o</mi> </msub> </mrow> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mo>)</mo> </mrow> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>;</mo> </mrow>
    6) rotary inertia after being normalized in Fgb figures is calculated；
    <mrow> <mi>N</mi> <mi>M</mi> <mi>I</mi> <mn>1</mn> <mo>=</mo> <mfrac> <msqrt> <msub> <mi>J</mi> <mrow> <mo>(</mo> <msub> <mi>x</mi> <mi>o</mi> </msub> <mo>,</mo> <msub> <mi>y</mi> <mi>o</mi> </msub> <mo>)</mo> </mrow> </msub> </msqrt> <mi>m</mi> </mfrac> <mo>=</mo> <mfrac> <msqrt> <mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>x</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>M</mi> </munderover> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>y</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <mrow> <mo>(</mo> <msup> <mrow> <mo>(</mo> <mi>x</mi> <mo>-</mo> <msub> <mi>x</mi> <mi>o</mi> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mo>+</mo> <msup> <mrow> <mo>(</mo> <mi>y</mi> <mo>-</mo> <msub> <mi>y</mi> <mi>o</mi> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mo>)</mo> </mrow> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> </mrow> </msqrt> <mrow> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>x</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>M</mi> </munderover> <munderover> <mo>&amp;Sigma;</mo> <mrow> <mi>y</mi> <mo>=</mo> <mn>1</mn> </mrow> <mi>N</mi> </munderover> <mi>f</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> </mrow> </mfrac> </mrow>
    7) meter of rotary inertia is normalized in the foreground area Fgd got using identical computational methods to three frame cumulative errors Calculate, obtain NMI2；
    8) NMI1 and NMI2 alpha value calculateds are utilized：
    <mrow> <mi>&amp;alpha;</mi> <mo>=</mo> <mfrac> <mrow> <mo>|</mo> <mi>N</mi> <mi>M</mi> <mi>I</mi> <mn>1</mn> <mo>-</mo> <mi>N</mi> <mi>M</mi> <mi>I</mi> <mn>2</mn> <mo>|</mo> </mrow> <mrow> <mi>N</mi> <mi>M</mi> <mi>I</mi> <mn>1</mn> </mrow> </mfrac> </mrow>
    9) when α values are less than threshold value Th, then the region is judged for background residual region, without mark；Threshold value Th is according to specific Video scene is chosen by testing.
CN201310452893.9A 2013-09-25 2013-09-25 A kind of static target removing method based on NMI features Active CN104463902B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310452893.9A CN104463902B (en) 2013-09-25 2013-09-25 A kind of static target removing method based on NMI features

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310452893.9A CN104463902B (en) 2013-09-25 2013-09-25 A kind of static target removing method based on NMI features

Publications (2)

Publication Number Publication Date
CN104463902A CN104463902A (en) 2015-03-25
CN104463902B true CN104463902B (en) 2017-12-22

Family

ID=52909884

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310452893.9A Active CN104463902B (en) 2013-09-25 2013-09-25 A kind of static target removing method based on NMI features

Country Status (1)

Country Link
CN (1) CN104463902B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113793533A (en) * 2021-08-30 2021-12-14 武汉理工大学 Collision early warning method and device based on vehicle front obstacle recognition
CN114372994B (en) * 2022-01-10 2022-07-22 北京中电兴发科技有限公司 Method for generating background image in video concentration

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1606039A (en) * 2004-11-18 2005-04-13 上海交通大学 Method for reducing false determination ratio of target image automatic identification and rapid tracking
CN102975659A (en) * 2012-11-14 2013-03-20 沈阳理工大学 Automobile night driving pre-warning system and method based on far infrared image

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1606039A (en) * 2004-11-18 2005-04-13 上海交通大学 Method for reducing false determination ratio of target image automatic identification and rapid tracking
CN102975659A (en) * 2012-11-14 2013-03-20 沈阳理工大学 Automobile night driving pre-warning system and method based on far infrared image

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
The Precise Recognition of Moving Object in Complex Background;Weiyao Huang 等;《Third International Conference on Natural Computation》;20070827;全文 *
基于Mean Shift算法和NMI特征的目标跟踪算法研究;甘明刚 等;《自动化学报》;20100930;第36卷(第9期);第1332-1336页 *
基于人运动的目标检测跟踪算法中有关问题的研究;童卫勇;《中国优秀硕士论文全文数据库信息科技辑》;20080515;第2008年卷(第5期);第4.2.2节，第4.4节 *
复杂背景下运动目标检测方法研究;张森悦;《中国优秀硕士论文全文数据库信息科技辑》;20091015;第2009年卷(第10期);第1.1.2节，第4章，第4.1节，第5.1节，第5.3.2节，图5.3 *

Also Published As

Publication number Publication date
CN104463902A (en) 2015-03-25

Similar Documents

Publication Publication Date Title
US20200160061A1 (en) Automatic ship tracking method and system based on deep learning network and mean shift
CN103440668B (en) Method and device for tracing online video target
CN104318263A (en) Real-time high-precision people stream counting method
CN105404857A (en) Infrared-based night intelligent vehicle front pedestrian detection method
JP2015506516A5 (en)
CN104866099A (en) Error compensation method for improving gesture identification precision of intelligent device based on motion sensor
CN102866237A (en) Water quality safety online biological early warning and monitoring system based on video identification
CN104574439A (en) Kalman filtering and TLD (tracking-learning-detection) algorithm integrated target tracking method
WO2014082480A1 (en) Method and device for calculating number of pedestrians and crowd movement directions
CN103679745B (en) A kind of moving target detecting method and device
CN105913454A (en) Pixel coordinate locus prediction method of motion object in video image
CN101853388B (en) Unchanged view angle behavior identification method based on geometric invariable
CN103617636A (en) Automatic video-target detecting and tracking method based on motion information and sparse projection
CN103426179A (en) Target tracking method and system based on mean shift multi-feature fusion
CN111798487A (en) Target tracking method, device and computer readable storage medium
CN105023013A (en) Target detection method based on local standard deviation and Radon transformation
CN104091350A (en) Object tracking method achieved through movement fuzzy information
CN106778570A (en) A kind of pedestrian&#39;s real-time detection and tracking
CN104766341A (en) Detection method for construction change of remote-sensing image based on DSM and kernel density estimation
CN104168444A (en) Target tracking method of tracking ball machine and tracking ball machine
CN105809718A (en) Object tracking method with minimum trajectory entropy
US20170103536A1 (en) Counting apparatus and method for moving objects
CN104683765A (en) Video concentration method based on mobile object detection
CN115311617A (en) Method and system for acquiring passenger flow information of urban rail station area
CN104463902B (en) A kind of static target removing method based on NMI features

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant