CN110348492A - A kind of correlation filtering method for tracking target based on contextual information and multiple features fusion - Google Patents
A kind of correlation filtering method for tracking target based on contextual information and multiple features fusion Download PDFInfo
- Publication number
- CN110348492A CN110348492A CN201910548179.7A CN201910548179A CN110348492A CN 110348492 A CN110348492 A CN 110348492A CN 201910548179 A CN201910548179 A CN 201910548179A CN 110348492 A CN110348492 A CN 110348492A
- Authority
- CN
- China
- Prior art keywords
- target
- correlation filtering
- information
- response
- contextual information
- 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.)
- Pending
Links
- 238000001914 filtration Methods 0.000 title claims abstract description 31
- 230000004927 fusion Effects 0.000 title claims abstract description 15
- 238000000034 method Methods 0.000 title claims abstract description 11
- 230000004044 response Effects 0.000 claims abstract description 31
- 230000003044 adaptive effect Effects 0.000 claims abstract description 6
- 238000013135 deep learning Methods 0.000 claims abstract description 6
- 239000011159 matrix material Substances 0.000 claims description 12
- 125000004122 cyclic group Chemical group 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000005267 amalgamation Methods 0.000 claims 1
- 238000010586 diagram Methods 0.000 abstract description 7
- 230000000903 blocking effect Effects 0.000 abstract description 3
- 238000005286 illumination Methods 0.000 abstract description 3
- 239000000284 extract Substances 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 238000007500 overflow downdraw method Methods 0.000 description 3
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- 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
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20016—Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20056—Discrete and fast Fourier transform, [DFT, FFT]
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
Abstract
The present invention relates to a kind of correlation filtering method for tracking target based on contextual information and multiple features fusion, belongs to video frequency object tracking technical field.Histograms of oriented gradients (HOG) feature and color of object histogram feature are extracted first from target and context, the correlation filtering response of both traditional characteristics is merged by fixed weight strategy, then the convolution feature that target and context are extracted with the convolutional network in deep learning is utilized, using the fusion traditional characteristic response of adaptive weighting convergence strategy and convolution characteristic response, estimation target position is obtained based on fused response diagram and target scale variation issue is solved using Scale Estimation Method;The present invention can effectively solve the problem that tracking target because of situations such as blocking, dimensional variation, illumination etc., tracking drift conditions caused by background clutter factor, realize accurate and robust target following.
Description
Technical field
The present invention relates to a kind of correlation filtering method for tracking target based on contextual information and multiple features fusion, belongs to view
Frequency target following technical field.
Background technique
In computer vision field, target following has broad application prospects, and mainly includes human-computer interaction, military system
Lead, athlete's analysis, intelligent vision navigation etc..Although Target Tracking Problem has been achieved for very big breakthrough in recent years,
But because of situations such as existing forms variation and background clutter during target is in tracking, accurate target following is completed, is still
One greatly challenge.
According to the modeling pattern difference of target appearance model, target following model can be divided into two classes: production model and
Discriminative model.Based on the target tracking algorism of production model, the appearance features of target are described using production model, are passed through
Sampling searches out the candidate target come and realizes that reconstructed error minimizes, then compares the similarity degree of candidate target and model, finds
Maximum similar purpose is as tracking result.Discriminate apparent model be then distinguished by the various classifiers of training by with
The target object of track and background area efficiently utilize the context information of target, since effect preferably obtains extensively
Using.What wherein tool represented is the Staple-CA algorithm based on correlation filtering, the algorithm effective integration target and context
HOG feature and color histogram feature are divided into two independent ridge regressions and solve Target Tracking Problem.The algorithm is able to solve one
As the variation of target external form, such as target appearance encounters the discontinuity of variation and illumination, becomes to the target under complex environment
Change and is easy tracking failure.Since the algorithm is using traditional characteristic, the semantic information of target cannot be effectively extracted, when tracking mesh
When mark blocks, it is easily lost target, the situation of model drift and failure occurs.
Summary of the invention
The technical problem to be solved in the present invention is to provide a kind of correlation filtering based on contextual information and multiple features fusion
Method for tracking target, for solving during visual target tracking, because blocking, dimensional variation, illumination etc., background clutter factor
Caused by track the situations of drift conditions, realize accurate and robust target following.
The technical scheme is that a kind of correlation filtering target following side based on contextual information and multiple features fusion
Method extracts histograms of oriented gradients (HOG) feature and color of object histogram feature first from target and context, leads to
The correlation filtering response that fixed weight strategy merges both traditional characteristics is crossed, the convolutional network used in deep learning is then utilized
The convolution feature for extracting target and context, it is special using the fusion traditional characteristic response of adaptive weighting convergence strategy and convolution
Sign response is obtained estimation target position and is solved target scale variation using Scale Estimation Method to ask based on fused response diagram
Topic, specifically:
Step1, the initial position message and dimensional information for obtaining target;
Step2, the traditional characteristic that target and context information are extracted according to the initial information that Step1 is obtained: HOG is special
It seeks peace color histogram feature;
Step3, according to the Step2 HOG feature extracted and color histogram feature and respectively correlation filtering model is established, and
Two kinds of traditional characteristics are merged by the way of fixed coefficient fusion obtains correlation filtering response;
Step4, the convolution feature that initial target and context information are obtained using the convolutional network in deep learning;
Step5, correlation filtering model is established using the convolution feature extracted according to Step4, and using adaptive fusion side
Formula fusion Step3 obtains traditional characteristic response to obtain filter response to the end, and future position;
Step6, the target position information obtained using Step5, the dimensional variation of addition scaling filter prediction target;
Step7, it after obtaining the location information and dimensional information of target, updates correlation filtering model and is tracked, until most
A later frame.
The traditional characteristic and convolution feature extracted in the step Step3 and Step5 using target and contextual information is built
Specific step is as follows for vertical correlation filtering model:
Note is current to extract image block characteristics x, obtains matrix X by cyclic shift, adopts up and down around target sample x
The identical n contextual information of size is taken to obtain xi, corresponding cyclic shift matrices are Xi, sample n context letter
Breath sample trains classifier as negative sample, so that filter response with higher at target sample, upper and lower
Response is close to zero at literary information, and the ridge regression of objective function after contextual information is added are as follows:
In formula (1), λ and λ1For regularization parameter, w is filter, and y indicates correlation filtering desired output, by the back in formula
The circular matrix of scape sample and target sample merges, and can obtain:
In formula (2), Β is block circulant matrix, and direct computation of DFT transformation can be used in all circular matrixes in Fu's formula space
(DFT) matrix carries out diagonalization, obtains following formula:
Filter w utilizes Fourier's rapid solving are as follows:
The beneficial effects of the present invention are: be effectively utilized the background information around target, using background information auxiliary positioning,
Tracking accuracy is promoted.Traditional characteristic and convolution feature efficiently have been merged, can have been effectively reduced during tracking because of part
The generation of model drift phenomenon caused by blocking.It, can be with higher while accurately tracking target position and dimensional variation
Execute speed tracing target.
Detailed description of the invention
Fig. 1 is flow diagram of the invention;
Fig. 2 is idiographic flow schematic diagram of the invention.
Specific embodiment
With reference to the accompanying drawings and detailed description, the invention will be further described.
Embodiment 1: as shown in Figs. 1-2, a kind of correlation filtering target following based on contextual information and multiple features fusion
Method extracts histograms of oriented gradients (HOG) feature and color of object histogram feature first from target and context,
The correlation filtering response that both traditional characteristics are merged by fixed weight strategy, then utilizes the convolution net used in deep learning
Network extracts the convolution feature of target and context, using the fusion traditional characteristic response of adaptive weighting convergence strategy and convolution
Characteristic response obtains estimation target position based on fused response diagram and solves target scale variation using Scale Estimation Method
Problem.
Specific step is as follows:
Step1, the initial position message and dimensional information for obtaining target.
Since the present invention is the verifying invention validity on open test collection OTB-2013 in the Step1, so tracking
The location information and dimensional information of target first frame have mark in test set.By the mark file of read test collection, i.e.,
Available target initial information.
Step2, the HOG feature and color histogram feature that target is extracted according to the initial information that Step1 is obtained.
Step3, HOG feature and color histogram feature and the convolution spy that target and contextual information are extracted according to Step2
Sign establishes correlation filtering model respectively, and the response of traditional characteristic correlation filtering is obtained by the way of fixed coefficient fusion.It establishes
Specific step is as follows for correlation filtering model:
Note is current to extract image block characteristics x, obtains matrix X by cyclic shift, adopts up and down around target sample x
The identical n contextual information of size is taken to obtain xi, corresponding cyclic shift matrices are Xi.Sample obtained n context letter
Breath sample trains classifier as negative sample, so that filter response with higher at target sample, upper and lower
Response is close to zero at literary information.The ridge regression of objective function after addition contextual information are as follows:
In formula (1), λ and λ1For regularization parameter, w is filter, and y indicates correlation filtering desired output.By the back in formula
The circular matrix of scape sample and target sample merges, and can obtain:
In formula (2), Β is block circulant matrix.Direct computation of DFT transformation can be used in all circular matrixes in Fu's formula space
(DFT) matrix carries out diagonalization, obtains following formula:
Filter w utilizes Fourier's rapid solving are as follows:
Step4, the convolution feature that initial target and context information are obtained using the convolutional network in deep learning.
The convolutional network structure used is VGG-19, and since low layer convolution includes more location informations, and deep layer convolution includes more
Semantic information.So extracting conv3-4 with VGG-19, this three layers feature of conv4-4, conv5-4 carries out linear weighted function and obtains
It is responded to final convolution.
Step5, the volume obtained using the obtained correlation filtering response of the method fusion Step3 adaptively merged and Step4
Product characteristic response obtains response to the end: flast=kconvfconv+ktradftrad, and future position.Utilize the peak of video frame
Value secondary lobe ratio (PSR) measures each model to the contribution degree of trace model, dynamically distributes model response diagram and merges weight.
Respective adaptive weighting is calculated by following formula:
ktrad=1-kconv
Wherein, CconvIndicate traditional characteristic PSR, CtradIndicate convolution feature PSR, calculation formula are as follows:
C=PSR (ft)-PSR(ft-1)
In formula, t is the sequence number of present frame, and μ is mean value, and δ is variance.
Weight ktMore new strategy are as follows:
kt=(1- ηk)kt-1+ηkkt
Wherein, ηkCoefficient is updated for weight.
Step6, the target position information obtained using step5, the dimensional variation of addition scaling filter prediction target.
Scale " pyramid " sampling is carried out in the tracing positional predicted before.Target sample size for scale assessment
Selection principle are as follows:
Wherein, P, R are respectively width, height of the target in previous frame, and a is scale factor, and S is the total series of scale.Then it incites somebody to action
To the target sample of different scale be uniformly scaled the size of P × R, then carry out relevant operation with unidimensional scale correlation filter
Scale response diagram is obtained, maximum response position is corresponding target best scale.
Step7, it after obtaining the location information and dimensional information of target, updates correlation filtering model and is tracked, until most
A later frame.
In conjunction with attached drawing, the embodiment of the present invention is explained in detail above, but the present invention is not limited to above-mentioned
Embodiment within the knowledge of a person skilled in the art can also be before not departing from present inventive concept
Put that various changes can be made.
Claims (2)
1. a kind of correlation filtering method for tracking target based on contextual information and multiple features fusion, it is characterised in that:
Step1, the initial position message and dimensional information for obtaining target;
Step2, the traditional characteristic that target and context information are extracted according to the initial information that Step1 is obtained: HOG feature and
Color histogram feature;
Step3, it establishes correlation filtering model according to the Step2 HOG feature extracted and color histogram feature and respectively, and uses
The mode of fixed coefficient fusion merges two kinds of traditional characteristics and obtains correlation filtering response;
Step4, the convolution feature that initial target and context information are obtained using the convolutional network in deep learning;
Step5, correlation filtering model is established using the convolution feature extracted according to Step4, and melted using adaptive amalgamation mode
It closes Step3 and obtains traditional characteristic response to obtain filter response to the end, and future position;
Step6, the target position information obtained using Step5, the dimensional variation of addition scaling filter prediction target;
Step7, it after obtaining the location information and dimensional information of target, updates correlation filtering model and is tracked, to the last one
Frame.
2. the correlation filtering method for tracking target according to claim 1 based on contextual information and multiple features fusion,
Be characterized in that: the traditional characteristic and convolution feature extracted in the step Step3 and Step5 using target and contextual information is built
Specific step is as follows for vertical correlation filtering model:
Note is current to extract image block characteristics x, obtains matrix X by cyclic shift, takes up and down around target sample x big
Small identical n contextual information obtains xi, corresponding cyclic shift matrices are Xi, n contextual information sample sampling
This trains classifier as negative sample, so that filter response with higher at target sample, believes in context
Response is close to zero at breath, and the ridge regression of objective function after contextual information is added are as follows:
In formula (1), λ and λ1For regularization parameter, w is filter, and y indicates correlation filtering desired output, by the background sample in formula
It merges, can obtain with the circular matrix of target sample:
In formula (2), Β is block circulant matrix, and direct computation of DFT transformation can be used in all circular matrixes in Fu's formula space
(DFT) matrix carries out diagonalization, obtains following formula:
Filter w utilizes Fourier's rapid solving are as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910548179.7A CN110348492A (en) | 2019-06-24 | 2019-06-24 | A kind of correlation filtering method for tracking target based on contextual information and multiple features fusion |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910548179.7A CN110348492A (en) | 2019-06-24 | 2019-06-24 | A kind of correlation filtering method for tracking target based on contextual information and multiple features fusion |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110348492A true CN110348492A (en) | 2019-10-18 |
Family
ID=68182853
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910548179.7A Pending CN110348492A (en) | 2019-06-24 | 2019-06-24 | A kind of correlation filtering method for tracking target based on contextual information and multiple features fusion |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110348492A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111812636A (en) * | 2020-06-01 | 2020-10-23 | 杭州电子科技大学 | Particle filter tracking-before-detection method based on weight fusion selection |
CN111862160A (en) * | 2020-07-23 | 2020-10-30 | 中国兵器装备集团自动化研究所 | Target tracking method, medium and system based on ARM platform |
CN112651999A (en) * | 2021-01-19 | 2021-04-13 | 滨州学院 | Unmanned aerial vehicle ground target real-time tracking method based on space-time context perception |
CN113538509A (en) * | 2021-06-02 | 2021-10-22 | 天津大学 | Visual tracking method and device based on adaptive correlation filtering feature fusion learning |
Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2016026370A1 (en) * | 2014-08-22 | 2016-02-25 | Zhejiang Shenghui Lighting Co., Ltd. | High-speed automatic multi-object tracking method and system with kernelized correlation filters |
CN106127776A (en) * | 2016-06-28 | 2016-11-16 | 北京工业大学 | Based on multiple features space-time context robot target identification and motion decision method |
WO2017088050A1 (en) * | 2015-11-26 | 2017-06-01 | Sportlogiq Inc. | Systems and methods for object tracking and localization in videos with adaptive image representation |
CN107403175A (en) * | 2017-09-21 | 2017-11-28 | 昆明理工大学 | Visual tracking method and Visual Tracking System under a kind of movement background |
CN107451601A (en) * | 2017-07-04 | 2017-12-08 | 昆明理工大学 | Moving Workpieces recognition methods based on the full convolutional network of space-time context |
CN107680119A (en) * | 2017-09-05 | 2018-02-09 | 燕山大学 | A kind of track algorithm based on space-time context fusion multiple features and scale filter |
KR101837407B1 (en) * | 2017-11-03 | 2018-03-12 | 국방과학연구소 | Apparatus and method for image-based target tracking |
US20180268559A1 (en) * | 2017-03-16 | 2018-09-20 | Electronics And Telecommunications Research Institute | Method for tracking object in video in real time in consideration of both color and shape and apparatus therefor |
CN108647694A (en) * | 2018-04-24 | 2018-10-12 | 武汉大学 | Correlation filtering method for tracking target based on context-aware and automated response |
WO2018208245A1 (en) * | 2017-05-12 | 2018-11-15 | Aselsan Elektronik Sanayi Ve Ticaret Anonim Sirketi | A method for correlation filter based visual tracking |
CN109285179A (en) * | 2018-07-26 | 2019-01-29 | 昆明理工大学 | A kind of motion target tracking method based on multi-feature fusion |
CN109461172A (en) * | 2018-10-25 | 2019-03-12 | 南京理工大学 | Manually with the united correlation filtering video adaptive tracking method of depth characteristic |
CN109741366A (en) * | 2018-11-27 | 2019-05-10 | 昆明理工大学 | A kind of correlation filtering method for tracking target merging multilayer convolution feature |
-
2019
- 2019-06-24 CN CN201910548179.7A patent/CN110348492A/en active Pending
Patent Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2016026370A1 (en) * | 2014-08-22 | 2016-02-25 | Zhejiang Shenghui Lighting Co., Ltd. | High-speed automatic multi-object tracking method and system with kernelized correlation filters |
WO2017088050A1 (en) * | 2015-11-26 | 2017-06-01 | Sportlogiq Inc. | Systems and methods for object tracking and localization in videos with adaptive image representation |
CN106127776A (en) * | 2016-06-28 | 2016-11-16 | 北京工业大学 | Based on multiple features space-time context robot target identification and motion decision method |
US20180268559A1 (en) * | 2017-03-16 | 2018-09-20 | Electronics And Telecommunications Research Institute | Method for tracking object in video in real time in consideration of both color and shape and apparatus therefor |
WO2018208245A1 (en) * | 2017-05-12 | 2018-11-15 | Aselsan Elektronik Sanayi Ve Ticaret Anonim Sirketi | A method for correlation filter based visual tracking |
CN107451601A (en) * | 2017-07-04 | 2017-12-08 | 昆明理工大学 | Moving Workpieces recognition methods based on the full convolutional network of space-time context |
CN107680119A (en) * | 2017-09-05 | 2018-02-09 | 燕山大学 | A kind of track algorithm based on space-time context fusion multiple features and scale filter |
CN107403175A (en) * | 2017-09-21 | 2017-11-28 | 昆明理工大学 | Visual tracking method and Visual Tracking System under a kind of movement background |
KR101837407B1 (en) * | 2017-11-03 | 2018-03-12 | 국방과학연구소 | Apparatus and method for image-based target tracking |
CN108647694A (en) * | 2018-04-24 | 2018-10-12 | 武汉大学 | Correlation filtering method for tracking target based on context-aware and automated response |
CN109285179A (en) * | 2018-07-26 | 2019-01-29 | 昆明理工大学 | A kind of motion target tracking method based on multi-feature fusion |
CN109461172A (en) * | 2018-10-25 | 2019-03-12 | 南京理工大学 | Manually with the united correlation filtering video adaptive tracking method of depth characteristic |
CN109741366A (en) * | 2018-11-27 | 2019-05-10 | 昆明理工大学 | A kind of correlation filtering method for tracking target merging multilayer convolution feature |
Non-Patent Citations (7)
Title |
---|
BIN ZHOU: "Adaptive Context-Aware and Structural Correlation Filter for Visual Tracking", 《DOI:10.3390/APP9071338》 * |
FENG TANG: "Spatial-aware correlation filters with adaptive weight maps for visual tracking", 《NEUROCOMPUTING》 * |
JIMMY T.MBELWA: "Visual tracking using objectness-bounding box regression and correlation filters", 《JOURNAL OF ELECTRONIC IMAGING》 * |
JOAO F.HENRIQUES: "Exploiting the Circulant Structure of Tracking-by-detection with Kernels", 《ECCV‘12》 * |
MATTHIAS MUELLER: "Context-Aware Correlation Filter Tracking", 《2017 IEEE CONFERENCE ON CVPR》 * |
李国友等: "一种基于多特征融合的核相关滤波目标跟踪算法", 《计算机与应用化学》 * |
杨洋: "空时上下文模型下基于多种特征融合的监控目标跟踪", 《计算机测量与控制》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111812636A (en) * | 2020-06-01 | 2020-10-23 | 杭州电子科技大学 | Particle filter tracking-before-detection method based on weight fusion selection |
CN111812636B (en) * | 2020-06-01 | 2023-06-13 | 杭州电子科技大学 | Particle filtering pre-detection tracking method based on weight fusion selection |
CN111862160A (en) * | 2020-07-23 | 2020-10-30 | 中国兵器装备集团自动化研究所 | Target tracking method, medium and system based on ARM platform |
CN111862160B (en) * | 2020-07-23 | 2023-10-13 | 中国兵器装备集团自动化研究所有限公司 | Target tracking method, medium and system based on ARM platform |
CN112651999A (en) * | 2021-01-19 | 2021-04-13 | 滨州学院 | Unmanned aerial vehicle ground target real-time tracking method based on space-time context perception |
CN113538509A (en) * | 2021-06-02 | 2021-10-22 | 天津大学 | Visual tracking method and device based on adaptive correlation filtering feature fusion learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110070074B (en) | Method for constructing pedestrian detection model | |
CN109145939B (en) | Semantic segmentation method for small-target sensitive dual-channel convolutional neural network | |
CN110348492A (en) | A kind of correlation filtering method for tracking target based on contextual information and multiple features fusion | |
CN109816689A (en) | A kind of motion target tracking method that multilayer convolution feature adaptively merges | |
CN103268495B (en) | Human body behavior modeling recognition methods based on priori knowledge cluster in computer system | |
Li et al. | Adaptive deep convolutional neural networks for scene-specific object detection | |
CN110555387B (en) | Behavior identification method based on space-time volume of local joint point track in skeleton sequence | |
CN109919981A (en) | A kind of multi-object tracking method of the multiple features fusion based on Kalman filtering auxiliary | |
CN107330357A (en) | Vision SLAM closed loop detection methods based on deep neural network | |
CN109285179A (en) | A kind of motion target tracking method based on multi-feature fusion | |
CN109902806A (en) | Method is determined based on the noise image object boundary frame of convolutional neural networks | |
CN108665481A (en) | Multilayer depth characteristic fusion it is adaptive resist block infrared object tracking method | |
Cao et al. | Rapid detection of blind roads and crosswalks by using a lightweight semantic segmentation network | |
CN106056628A (en) | Target tracking method and system based on deep convolution nerve network feature fusion | |
CN105678231A (en) | Pedestrian image detection method based on sparse coding and neural network | |
CN105550678A (en) | Human body motion feature extraction method based on global remarkable edge area | |
CN111862145B (en) | Target tracking method based on multi-scale pedestrian detection | |
CN109101938A (en) | A kind of multi-tag age estimation method based on convolutional neural networks | |
CN105976397B (en) | A kind of method for tracking target | |
CN103440471B (en) | The Human bodys' response method represented based on low-rank | |
CN109241995A (en) | A kind of image-recognizing method based on modified ArcFace loss function | |
CN108154113A (en) | Tumble event detecting method based on full convolutional network temperature figure | |
CN109635695A (en) | Pedestrian based on triple convolutional neural networks recognition methods again | |
CN105184229A (en) | Online learning based real-time pedestrian detection method in dynamic scene | |
CN108682022A (en) | Based on the visual tracking method and system to anti-migration network |
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 | ||
AD01 | Patent right deemed abandoned | ||
AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20230317 |