CN108492318B - Target tracking method based on bionic technology - Google Patents
Target tracking method based on bionic technology Download PDFInfo
- Publication number
- CN108492318B CN108492318B CN201810171787.6A CN201810171787A CN108492318B CN 108492318 B CN108492318 B CN 108492318B CN 201810171787 A CN201810171787 A CN 201810171787A CN 108492318 B CN108492318 B CN 108492318B
- Authority
- CN
- China
- Prior art keywords
- target
- characteristic
- probability
- visual
- coefficient
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Multimedia (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a target tracking method based on a bionic technology, which utilizes the motion mode and search strategy of eyes and five visual characteristics of a biological visual system in the target tracking process of the bionic technology, including light and shade characteristics, color characteristics, sensitivity characteristics, visual residue and memory characteristics, and utilizes the characteristics to carry out characteristic change on a target to design a search algorithm suitable for target tracking. The method comprises the three steps of constructing a target template, constructing a candidate target template and updating the target template. The target tracking method based on the bionic technology integrates the intelligent processing method of biological visual characteristics to the target, and provides a method for replacing the color characteristic gravity center by the maximum probability gravity center.
Description
Technical Field
The invention belongs to the field of machine vision, and relates to target feature detection and extraction by simulating biological visual characteristics, in particular to tracking and identifying a target based on bionic visual characteristics.
Background
The tracking problem of the moving target is always a research hotspot in the scientific research field, the effect factors influencing target tracking in real life are very many, most of the current target tracking algorithms can only be applied in specific environments, the traditional moving target tracking algorithm lacks the characteristic of biological intelligence and cannot adapt to complex environment transformation, the tracking effect is not satisfactory, great use limitation exists, the effect of biological tracking of the target is achieved, and the target tracking direction is always a challenge. The target tracking research based on the bionic technology is inspired by the high efficiency and intelligence of a biological vision system in the target tracking process, and is blended with the idea of human eye vision characteristics to make up the defects of the existing tracking algorithm.
The method is based on a human eye vision system, three image transformation methods including color feature transformation, sensitivity transformation and memory feature transformation are summarized by analyzing the effect of human eye vision characteristics in target feature extraction, and a target tracking method based on a bionic technology is provided.
Disclosure of Invention
The invention aims to simulate the biological vision to process the external visual information, the bionic vision characteristic in the target characteristic extraction and tracking process and the memory function of the target when the target is deformed, and provides a target tracking method simulating the biological vision characteristic.
The invention establishes five visual characteristics of the bionic visual system, including light and shade characteristics, color characteristics, sensitivity characteristics, visual residues and memory characteristics, and the movement mode and the search strategy of eyes of the bionic visual system in the process of tracking the target, and designs a search algorithm suitable for target tracking by using the characteristics to change the characteristics of the target.
The method carries out feature transformation on the target according to five visual characteristics, namely visual brightness characteristic, color characteristic, sensitivity characteristic, visual residue and memory characteristic, and uses the characteristics to change the feature of the target from RGB color space feature to target probability feature. In the target tracking method based on biological visual characteristics, the target template, the candidate target template and the background are subjected to characteristic transformation so as to search the optimal candidate model. The basic scheme is as follows:
1. first frame selection target y0Comprehensively transforming the visual characteristics of the target to obtain a target template P0And a target biological probability characteristic coefficient W.
2. In the next frame, a candidate target statistical model Z is calculated by using the target biological probability characteristic coefficient W1And calculating a candidate target probability center P1Setting the iteration times T of the probability characteristic coefficients, and selecting the T closest probability characteristic coefficients for iteration. And if the probability characteristic coefficient W is an empty set, judging that the coefficient is invalid, and selecting a first probability coefficient for intersection processing. Calculating background difference coefficient h and setting threshold h*If h is<h*Recalculating new probability characteristic coefficient and performing intersection processing with the previously calculated probability characteristic coefficient if h is more than or equal to h*Then proceed to step 3.
3. Setting iteration times N and iteration threshold epsilon*Calculating ∈ ═ P1-P0L. If epsilon>ε*Updating the target probability center P0=P1If ε is ≦ ε*Then the target area y is the optimal candidate target.
4. And calibrating the optimal candidate target as a tracking target.
The background information memorized by short-term memory can help us to better track the target. The background information used when the biological visual characteristics are used for extracting the target features is stored through short-time memory. In the face of complex environment, the background in the short-term memory is continuously changed along with the change of the environment, the most obvious characteristic with the highest sensitivity of the target is changed, and a transformed template is required to replace the former template. Color characteristics in biological visual characteristics also mention this, and the human eye can automatically adjust the perceived color according to the difference of background colors.
In a complex environment, the feature sensitivity of the target template changes along with the change of the environment, and when the sensitivity of the target template features relative to the background disappears or is weakened, the tracking of the target is disabled. Higher organisms utilize the memory function of the brain and the processing of the optic nervous system. The object is tracked by a method of constantly updating the target template to keep the target feature and the background always at the maximum sensitivity.
Compared with the prior art, the invention has the following advantages:
the movement mode of eyes of the bionic vision system is superior to a search strategy of a MeanShift target tracking algorithm in the tracking process of a target, in the MeanShift tracking algorithm, due to the influence of a kernel function and a weight function on the position of the target, the tracking effect of the MeanShift on a non-rigid object in the target tracking is poor, and once the target deforms, the tracking is easy to lose efficacy. The target tracking method based on the bionic visual characteristics integrates the intelligent processing method of the biological visual characteristics to the target, and provides a method for replacing the color characteristic gravity center with the maximum probability gravity center.
Drawings
FIG. 1 is a flow chart of a target tracking method based on a bionic technology.
Detailed Description
1. And constructing a target template, and converting the biological probability characteristic coefficient W of the target into a target probability graph. In the target probability map, the target characteristics can be obviously found. The gravity center position of the target probability map is calculated by using the target probability map,
i.e. at the center of maximum probability in the target template. The coordinate position of the x pixel point in the target template area is { xi *,i=1...n},
Target probability total:
target probability center:
target biological probability characteristic coefficient:
2. and constructing a candidate target template. The candidate target template is similar to the target template, namely, a target position area y of the previous frame is selected in the next frame, and the coordinate position of pixel points in the area y isAnd a target pixel probability value I of the current frameyiThen the candidate model:
total probability of candidate targets:
candidate target probability center:
3. finding the position of maximum probability of an object, i.e. finding the center of probability P of a candidate object1And the relative position of the image, calculating the offset vector of the candidate target and the target, and then solving the optimal candidate target position.
Offset vector:
bs=P1-P0 (6)
optimal candidate target area:
y=x+bs (7)
4. and calibrating the tracking target, calculating the target position by using the offset vector, calculating a more accurate position by using iterative processing, and calibrating the more accurate position to obtain the target position.
Claims (1)
1. A target tracking method based on bionic technology is characterized in that:
(1) the method comprises the steps that a biological vision system is used for carrying out eye movement mode and search strategy in the target tracking process, five visual characteristics of the biological vision system are used, wherein the five visual characteristics comprise light and shade characteristics, color characteristics, sensitivity characteristics, visual residues and memory characteristics, and the characteristics are used for carrying out characteristic change on a target to design a search algorithm suitable for target tracking;
(2) according to the five visual characteristics of the biological vision, namely the brightness characteristic, the color characteristic, the sensitivity characteristic, the visual residue and the memory characteristic, the characteristics of the target are transformed by applying the characteristics, so that the characteristics of the target are changed from the RGB color space characteristics into the target probability characteristics, and the target tracking method based on the bionic technology carries out the characteristic transformation on a target template, a candidate target template and a background so as to search an optimal candidate model, and specifically comprises the following steps:
first frame selection target y0Using the visual characteristic comprehensive transformation to the target to obtainTarget template P0Calculating a candidate target statistical model Z by using the target biological probability characteristic coefficient W1And calculating a candidate target probability center P1Setting the iteration times T of the probability characteristic coefficients, selecting the T closest probability characteristic coefficients for iteration, if the probability characteristic coefficient W is an empty set, judging that the coefficient is invalid, selecting the first probability coefficient for intersection processing, calculating the background difference coefficient h and setting a threshold h*If h is<h*Recalculating new probability characteristic coefficient and performing intersection processing with the previously calculated probability characteristic coefficient if h is more than or equal to h*Setting iteration number N and iteration threshold epsilon*Calculating ∈ ═ P1-P0If e.g. of>ε*Updating the target probability center P0=P1If ε is ≦ ε*Then the target area y is the optimal candidate target.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810171787.6A CN108492318B (en) | 2018-03-01 | 2018-03-01 | Target tracking method based on bionic technology |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810171787.6A CN108492318B (en) | 2018-03-01 | 2018-03-01 | Target tracking method based on bionic technology |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108492318A CN108492318A (en) | 2018-09-04 |
CN108492318B true CN108492318B (en) | 2022-04-26 |
Family
ID=63341146
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810171787.6A Active CN108492318B (en) | 2018-03-01 | 2018-03-01 | Target tracking method based on bionic technology |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108492318B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP5076971B2 (en) * | 2008-02-29 | 2012-11-21 | 富士通株式会社 | Image target detection and tracking apparatus, image target detection and tracking method, and image target detection and tracking program |
CN105335986A (en) * | 2015-09-10 | 2016-02-17 | 西安电子科技大学 | Characteristic matching and MeanShift algorithm-based target tracking method |
CN106446833A (en) * | 2016-09-27 | 2017-02-22 | 湖南商学院 | Multichannel bionic vision method for recognizing complex scene image |
CN106780560A (en) * | 2016-12-29 | 2017-05-31 | 北京理工大学 | A kind of feature based merges the bionic machine fish visual tracking method of particle filter |
CN107368802A (en) * | 2017-07-14 | 2017-11-21 | 北京理工大学 | Motion target tracking method based on KCF and human brain memory mechanism |
CN107657627A (en) * | 2017-08-24 | 2018-02-02 | 北京理工大学 | Space-time contextual target tracking based on human brain memory mechanism |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6999599B2 (en) * | 2002-06-07 | 2006-02-14 | Microsoft Corporation | System and method for mode-based multi-hypothesis tracking using parametric contours |
CN103824305A (en) * | 2014-03-17 | 2014-05-28 | 天津工业大学 | Improved Meanshift target tracking method |
-
2018
- 2018-03-01 CN CN201810171787.6A patent/CN108492318B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP5076971B2 (en) * | 2008-02-29 | 2012-11-21 | 富士通株式会社 | Image target detection and tracking apparatus, image target detection and tracking method, and image target detection and tracking program |
CN105335986A (en) * | 2015-09-10 | 2016-02-17 | 西安电子科技大学 | Characteristic matching and MeanShift algorithm-based target tracking method |
CN106446833A (en) * | 2016-09-27 | 2017-02-22 | 湖南商学院 | Multichannel bionic vision method for recognizing complex scene image |
CN106780560A (en) * | 2016-12-29 | 2017-05-31 | 北京理工大学 | A kind of feature based merges the bionic machine fish visual tracking method of particle filter |
CN107368802A (en) * | 2017-07-14 | 2017-11-21 | 北京理工大学 | Motion target tracking method based on KCF and human brain memory mechanism |
CN107657627A (en) * | 2017-08-24 | 2018-02-02 | 北京理工大学 | Space-time contextual target tracking based on human brain memory mechanism |
Non-Patent Citations (2)
Title |
---|
Motion Control in Saccade and Smooth Pursuit for Bionic Eye Based on Three-dimensional Coordinates;Qingbin Wang等;《JOURNAL OF BIONIC ENGINEERING》;20170430;第14卷(第2期);全文 * |
基于相关区域分层的改进Meanshift目标跟踪算法;董慧芬等;《计算机应用》;20141215;第34卷(第S2期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN108492318A (en) | 2018-09-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111209810B (en) | Boundary frame segmentation supervision deep neural network architecture for accurately detecting pedestrians in real time through visible light and infrared images | |
CN111797716B (en) | Single target tracking method based on Siamese network | |
CN110503680B (en) | Unsupervised convolutional neural network-based monocular scene depth estimation method | |
CN108986140B (en) | Target scale self-adaptive tracking method based on correlation filtering and color detection | |
CN113469356B (en) | Improved VGG16 network pig identity recognition method based on transfer learning | |
CN107016691B (en) | Moving target detecting method based on super-pixel feature | |
CN103413120B (en) | Tracking based on object globality and locality identification | |
CN108537147B (en) | Gesture recognition method based on deep learning | |
CN110889844B (en) | Coral distribution and health condition assessment method based on deep clustering analysis | |
CN110675462A (en) | Gray level image colorizing method based on convolutional neural network | |
CN111199556B (en) | Indoor pedestrian detection and tracking method based on camera | |
CN110111351B (en) | Pedestrian contour tracking method fusing RGBD multi-modal information | |
CN102831604A (en) | Two-dimensional Renyi entropic threshold segmentation method for grayscale images | |
CN110084201B (en) | Human body action recognition method based on convolutional neural network of specific target tracking in monitoring scene | |
CN110246154B (en) | Visual target tracking method based on ICA-R multi-feature fusion and self-adaptive updating | |
CN109215003B (en) | Image fusion method and device | |
CN104966274B (en) | A kind of On Local Fuzzy restored method using image detection and extracted region | |
CN107609571A (en) | A kind of adaptive target tracking method based on LARK features | |
CN101504771A (en) | Vision tracing method for non-parameterized model | |
CN108053418B (en) | Animal background modeling method and device | |
CN108492318B (en) | Target tracking method based on bionic technology | |
CN116935133B (en) | Cardiovascular disease classification method and system based on SPECT image recognition | |
CN109064490A (en) | A kind of motion target tracking method based on MeanShift | |
CN113192049A (en) | Visible light and infrared image fusion method based on LatLRR and Retinex enhancement | |
CN116363733A (en) | Facial expression prediction method based on dynamic distribution fusion |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |