[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

CN106920247A - A kind of method for tracking target and device based on comparison network - Google Patents

A kind of method for tracking target and device based on comparison network Download PDF

Info

Publication number
CN106920247A
CN106920247A CN201710038541.7A CN201710038541A CN106920247A CN 106920247 A CN106920247 A CN 106920247A CN 201710038541 A CN201710038541 A CN 201710038541A CN 106920247 A CN106920247 A CN 106920247A
Authority
CN
China
Prior art keywords
target
object candidate
region
frame image
tracked target
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
Application number
CN201710038541.7A
Other languages
Chinese (zh)
Inventor
谯帅
蒲津
何建伟
张如高
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.)
Bocom Intelligent Information Technology Co Ltd Shanghai Branch
Original Assignee
Bocom Intelligent Information Technology Co Ltd Shanghai Branch
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 Bocom Intelligent Information Technology Co Ltd Shanghai Branch filed Critical Bocom Intelligent Information Technology Co Ltd Shanghai Branch
Priority to CN201710038541.7A priority Critical patent/CN106920247A/en
Publication of CN106920247A publication Critical patent/CN106920247A/en
Pending legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/20Analysis of motion
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10016Video; Image sequence
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20084Artificial neural networks [ANN]

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)
  • Closed-Circuit Television Systems (AREA)

Abstract

The present invention discloses a kind of method for tracking target and device based on comparison network, wherein, based on the method for tracking target for comparing network, comprise the following steps:It is determined that the current frame image comprising tracked target;Tracked target region is obtained in current frame image;Obtain next two field picture of present frame;Multiple object candidate areas are obtained in next two field picture;Obtain the boundary rectangle comprising each object candidate area;Tracked target region and boundary rectangle input are compared into neural network model to compare;Target following region is determined in multiple object candidate areas according to comparison result;Determine tracked target in target following region.The present invention carries out Integrated comparative by comparing neural network model to tracked target region and object candidate area, and tracked target object can be recognized accurately, and is rapidly completed tracing task.

Description

A kind of method for tracking target and device based on comparison network
Technical field
The present invention relates to image processing field, and in particular to a kind of based on the method for tracking target and device that compare network.
Background technology
The purpose of target following is the movement locus for obtaining specific objective in video sequence, recently as computer network The fast propagation of video, the research of target following is always the heat subject of computer vision field, also in many practicality visions Key player is play in system, and target following refers to the original state for providing target in the frame of video first is tracked, prediction Target exact position in subsequent frames, at the same time, visual target tracking also serves as the basis of artificial intelligence, can simulate The behavior of human vision.
Method for tracking target in currently available technology is mainly by the detection mode extraction target of learning classification task Track target object after the characteristic information of object, but image information species in video streaming is a lot, during tracking, because For Integrated comparative, institute cannot be carried out to the target area in the previous frame image region of interest within and next two field picture in video flowing It is complex with the target that the characteristic information with tracked target is found in all of video image, while cannot be quickly accurate Really complete tracing task.
The content of the invention
Therefore, the embodiment of the present invention technical problem to be solved is method for tracking target of the prior art using study Target object is tracked after the characteristic information of the detection mode extraction target object of classification task, it is impossible to the previous frame in video flowing Target area in image region of interest within and next two field picture carries out Integrated comparative, causes accurately obtain tracked target Object.
Therefore, the embodiment of the invention provides following technical scheme:
The embodiment of the present invention provides a kind of based on the method for tracking target for comparing network, comprises the following steps:
It is determined that the current frame image comprising tracked target;
Tracked target region is obtained in the current frame image;
Obtain next two field picture of present frame;
Multiple object candidate areas are obtained in next two field picture;
Obtain the boundary rectangle comprising each object candidate area;
The tracked target region and boundary rectangle input are compared into neural network model to compare;
Target following region is determined in multiple object candidate areas according to comparison result.
Alternatively, also include:Determine tracked target in the target following region.
Alternatively, the multiple object candidate areas of the acquisition in next two field picture, including:
Obtain the positive sample and negative sample of the current frame image;
Detect the next frame image edge information;
The edge that the tracked target region overlaps with described image marginal information is detected in the current frame image Information.
Alternatively, it is described to carry out the tracked target region and boundary rectangle input comparison neural network model Compare, comprise the following steps:
The boundary rectangle of the object candidate area is mapped to the convolution feature of correspondence position in convolutional layer;
The dimension of the unification convolution feature in the layer of pond;
The convolution feature is comprehensively input into full articulamentum.
Alternatively, it is described to determine target following region in multiple object candidate areas according to comparison result, including:
Calculate the similarity in the tracked target region and the object candidate area;
Obtain maximum object candidate area score.
The embodiment of the present invention provides a kind of based on the target tracker for comparing network, including such as lower unit:
First determining unit, for determining the current frame image comprising tracked target;
First acquisition unit, for obtaining tracked target region in the current frame image;
Second acquisition unit, the next two field picture for obtaining present frame;
3rd acquiring unit, for obtaining multiple object candidate areas in next two field picture;
4th acquiring unit, for obtaining the boundary rectangle comprising each object candidate area;
Comparing unit, enters for the tracked target region and boundary rectangle input to be compared into neural network model Row is compared;
Second determining unit, for determining target following region in multiple object candidate areas according to comparison result.
Alternatively, also include:3rd determining unit, for determining tracked target in the target following region.
Alternatively, the 3rd acquiring unit, including:
First acquisition module, positive sample and negative sample for obtaining the current frame image;
First detection module, for detecting the next frame image edge information;
Second detection module, for detecting the tracked target region and described image side in the current frame image The marginal information that edge information overlaps.
Alternatively, the comparing unit, including:
Mapping block, the volume for the boundary rectangle of the object candidate area to be mapped to correspondence position in convolutional layer Product feature;
Unified modules, for the dimension of the unification convolution feature in the layer of pond;
Input module, for being comprehensively input into the convolution feature in full articulamentum.
Alternatively, second determining unit, including:
Computing module, the similarity for calculating the tracked target region and the object candidate area;
Second acquisition module, the object candidate area score for obtaining maximum.
Embodiment of the present invention technical scheme, has the following advantages that:
The present invention provides a kind of based on the method for tracking target and device that compare network, wherein, based on the mesh for comparing network Mark tracking, comprises the following steps:It is determined that the current frame image comprising tracked target;Obtained in current frame image by with Track target area;Obtain next two field picture of present frame;Multiple object candidate areas are obtained in next two field picture;Acquisition is included The boundary rectangle of each object candidate area;Tracked target region and boundary rectangle input are compared into neural network model is carried out Compare;Target following region is determined in multiple object candidate areas according to comparison result;Determine quilt in target following region Tracking target.The present invention carries out comprehensive ratio by comparing neural network model to tracked target region and object candidate area Compared with, tracked target object can be recognized accurately, it is rapidly completed tracing task.
Brief description of the drawings
In order to illustrate more clearly of the specific embodiment of the invention or technical scheme of the prior art, below will be to specific The accompanying drawing to be used needed for implementation method or description of the prior art is briefly described, it should be apparent that, in describing below Accompanying drawing is some embodiments of the present invention, for those of ordinary skill in the art, before creative work is not paid Put, other accompanying drawings can also be obtained according to these accompanying drawings.
Fig. 1 is the flow chart based on the goal approach for comparing network in the embodiment of the present invention 1;
Fig. 2 is the flow chart based on acquisition edge peaks in the goal approach for comparing network in the embodiment of the present invention 1;
Fig. 3 be the embodiment of the present invention 1 in based on compare network goal approach in compare flow chart;
Fig. 4 is the flow chart based on determination target following region in the goal approach for comparing network in the embodiment of the present invention 1;
Fig. 5 is the structured flowchart based on the destination apparatus for comparing network in the embodiment of the present invention 2;
Fig. 6 is the structured flowchart based on the acquiring unit of destination apparatus the 3rd for comparing network in the embodiment of the present invention 2;
Fig. 7 is the structured flowchart based on the destination apparatus comparing unit for comparing network in the embodiment of the present invention 2;
Fig. 8 is the structured flowchart based on the determining unit of destination apparatus second for comparing network in the embodiment of the present invention 2.
Specific embodiment
The technical scheme of the embodiment of the present invention is clearly and completely described below in conjunction with accompanying drawing, it is clear that described Embodiment be a part of embodiment of the invention, rather than whole embodiments.Based on the embodiment in the present invention, this area is general The every other embodiment that logical technical staff is obtained under the premise of creative work is not made, belongs to present invention protection Scope.
, it is necessary to explanation in the description of the embodiment of the present invention, term " " center ", " on ", D score, "left", "right", The orientation or position relationship of the instruction such as " vertical ", " level ", " interior ", " outward " be based on orientation shown in the drawings or position relationship, It is for only for ease of the description embodiment of the present invention and simplifies description, must has rather than the device or element for indicating or imply meaning Have specific orientation, with specific azimuth configuration and operation, therefore be not considered as limiting the invention.Additionally, term " the One ", " second ", " the 3rd " are only used for describing purpose, and it is not intended that indicating or implying relative importance.
, it is necessary to explanation, unless otherwise clearly defined and limited, term " is pacified in the description of the embodiment of the present invention Dress ", " connected ", " connection " should be interpreted broadly, for example, it may be fixedly connected, or be detachably connected, or integratedly Connection;Can mechanically connect, or electrically connect;Can be joined directly together, it is also possible to be indirectly connected to by intermediary, Two connections of element internal are can also be, can be wireless connection, or wired connection.For the common skill of this area For art personnel, above-mentioned term concrete meaning in the present invention can be understood with concrete condition.
As long as additionally, technical characteristic involved in invention described below different embodiments non-structure each other Can just be combined with each other into conflict.
Embodiment 1
The present embodiment provides a kind of based on the method for tracking target for comparing network, as shown in figure 1, comprising the following steps:
The current frame image of S1, determination comprising tracked target;Being input into several continuous images in video streaming could structure Into a complete video, and image is continuous sequential correlation data, thus only by obtain with present frame by with Track target image could complete specific tracking, and the current frame image position comprising tracked target is only determined in video streaming Tracked target can just be found.
Specifically, target following is often referred to provide original state of the target in the frame of video first is tracked, and mesh is estimated automatically Mark object state in subsequent frames.Human eye can compare easily within a period of time with living certain specific objective, but to machine For device, this task is simultaneously remarkable, occurs that target occurs drastic mechanical deformation during tracking, by other target occlusions or There are the various complicated situations of similar object interference etc..Above-mentioned present frame is comprising the initial frame being input into from video flowing or upper The image information of frame or next frame at current time, the information includes the positions and dimensions of present frame.
S2, the acquisition tracked target region in current frame image;Specifically quilt is as detected in current frame image Tracking mesh target area (current patch), this region is the image block being made up of multiple pixels.
S3, the next two field picture for obtaining present frame;The purpose of tracking is exactly in order to catch up with object above, only by obtaining Taking a frame frame image information could finally track this specific object, so need exist in next frame, using present frame The tracked target region of middle acquisition completes further tracking.
S4, the multiple object candidate areas of acquisition in next two field picture;Using proposal generating modes, generation target is built The purpose for discussing position is exactly to generate a relatively small choice box Candidate Set of quantity, as multiple object candidate areas.
As a kind of implementation, the method for tracking target based on target candidate in the present embodiment, as shown in Fig. 2 step S4, obtains multiple object candidate areas in next two field picture, including:
S41, the positive sample and negative sample that obtain current frame image;Positive sample refers in picture the only a certain mesh of searching in need Mark, that is, be often referred to the target sample information related to tracking, and negative sample refers to wherein not comprising the target for needing to find, that is, is often referred to Unnecessary, incoherent sample information, it is also possible to using other candidate regions outside correct target area as negative sample.
S42, detection next frame image edge information, target possible position is obtained by the image edge information of next frame Some candidate regions.
S43, the marginal information that detection tracked target region overlaps with image edge information in current frame image.
Specifically, the boundary response of each pixel in image is obtained using structuring edge detector, can be thus obtained To a dense boundary response, then perform non-maxima suppression and find edge peaks, so and can obtain one it is sparse Edge graph, then edge is grouped, it is believed that the edge pixel that straight border is connected has correlation higher, and not straight Connect in succession or there is lower correlation by the edge pixel that the too high curve of curvature is connected, next, using sliding window Method detects that the target more with the marginal information of the coincident of multiple object candidate areas is waited in current frame image region Favored area, so as to obtain specific object candidate area.
The boundary rectangle of S5, acquisition comprising each object candidate area;Obtaining the target of the more marginal information that overlaps The maximum boundary rectangle comprising each object candidate area of its composition, (rectangle patch) are obtained in candidate region.
S6, by tracked target region and boundary rectangle input compare neural network model compare;This compares nerve Network model is mainly the similarity one-time calculation of tracked target region and object candidate area out.
As a kind of implementation, based on the method for tracking target for comparing network in the present embodiment, as shown in figure 3, step S6, compares tracked target region and boundary rectangle input neural network model and compares, comprises the following steps:
S61, the convolution feature that the boundary rectangle of object candidate area is mapped in convolutional layer correspondence position;Roll up herein The effect of lamination is mainly used in extracting convolution feature, i.e., some image blocks are extracted from video image, learns some features, then By the use of these features as wave filter go it is inswept it is whole magnify figure, i.e., do convolution, the boundary rectangle of object candidate area line by line Correspondence position is mapped to, while by sharing the target candidate area that convolution is compared in tracked target region and boundary rectangle Domain.
S62, in the layer of pond unified convolution feature dimension;The effect of pond layer is mainly used in unifying the dimension of convolution feature Degree, because after convolutional layer finishes convolution line by line, dimension can change, so need to reduce dimension, so as to reduce volume The feature of lamination output, while improving result.
S63, convolution feature is comprehensively input into full articulamentum, full articulamentum is used to the characteristic synthetic for above extracting rise Come.
S7, target following region is determined in multiple object candidate areas according to comparison result.Such as, a certain car is being tracked , when acquiring all vehicles on highway for positive sample, after pedestrian is negative sample, and obtain the similar of tracked vehicle Degree, the scope as target following region where now needing the tracked all much like vehicle of concern.
As a kind of implementation, based on the method for tracking target for comparing network in the present embodiment, as shown in figure 4, according to Comparison result determines target following region in multiple object candidate areas, including:
S71, the similarity for calculating tracked target region and object candidate area;Calculate current frame image in obtain by with Similarity between track target area and multiple object candidate areas, it is understood that to calculate previous frame patch and next frame Target candidate patch between similarity, choose similarity score highest object candidate area traced into as present frame Target location.
S72, the object candidate area score for obtaining maximum, using the object candidate area of highest scoring as tracking result.
S8, in target following region determine tracked target.Determined according to definite target following region tracked Target object, completes tracing task.
Embodiment 2
The present embodiment provide it is a kind of based on compare network target tracker, as shown in figure 5, with embodiment 1 in base It is corresponding in target tracker, including such as lower unit:
First determining unit 51, for determining the current frame image comprising tracked target;
First acquisition unit 52, for obtaining tracked target region in current frame image;
Second acquisition unit 53, the next two field picture for obtaining present frame;
3rd acquiring unit 54, for obtaining multiple object candidate areas in next two field picture;
4th acquiring unit 55, for obtaining the boundary rectangle comprising each object candidate area;
Comparing unit 56, is compared for tracked target region and boundary rectangle input to be compared into neural network model It is right;
Second determining unit 57, for determining target following region in multiple object candidate areas according to comparison result.
As a kind of implementation, based on the target tracker for comparing network in the present embodiment, also include:3rd determines Unit 58, for determining tracked target in target following region.
As a kind of implementation, based on the target tracker for comparing network in the present embodiment, as shown in fig. 6, the 3rd Acquiring unit 54, including:
First acquisition module 541, positive sample and negative sample for obtaining current frame image;
First detection module 542, for detecting next frame image edge information;
Second detection module 543, for detecting tracked target region and image edge information weight in current frame image The marginal information of conjunction.
As a kind of implementation, based on the target tracker for comparing network in the present embodiment, as shown in fig. 7, comparing Unit 56, including:
Mapping block 561, the volume for the boundary rectangle of object candidate area to be mapped to correspondence position in convolutional layer Product feature;
Unified modules 562, for the dimension of the unified convolution feature in the layer of pond;
Input module 563, for being comprehensively input into convolution feature in full articulamentum.
As a kind of implementation, based on the target tracker for comparing network in the present embodiment, as shown in figure 8, second Determining unit 57, including:
Computing module 571, the similarity for calculating tracked target region and object candidate area;
Second acquisition module 572, the object candidate area score for obtaining maximum.
Obviously, above-described embodiment is only intended to clearly illustrate example, and not to the restriction of implementation method.It is right For those of ordinary skill in the art, can also make on the basis of the above description other multi-forms change or Change.There is no need and unable to be exhaustive to all of implementation method.And the obvious change thus extended out or Among changing still in the protection domain of the invention.

Claims (10)

1. it is a kind of based on the method for tracking target for comparing network, it is characterised in that to comprise the following steps:
It is determined that the current frame image comprising tracked target;
Tracked target region is obtained in the current frame image;
Obtain next two field picture of present frame;
Multiple object candidate areas are obtained in next two field picture;
Obtain the boundary rectangle comprising each object candidate area;
The tracked target region and boundary rectangle input are compared into neural network model to compare;
Target following region is determined in multiple object candidate areas according to comparison result.
2. method according to claim 1, it is characterised in that also include:Determine in the target following region by with Track target.
3. method according to claim 1, it is characterised in that described multiple targets are obtained in next two field picture to wait Favored area, including:
Obtain the positive sample and negative sample of the current frame image;
Detect the next frame image edge information;
The marginal information that the tracked target region overlaps with described image marginal information is detected in the current frame image.
4. method according to claim 1, it is characterised in that described by the tracked target region and the external square Shape input compares neural network model and compares, and comprises the following steps:
The boundary rectangle of the object candidate area is mapped to the convolution feature of correspondence position in convolutional layer;
The dimension of the unification convolution feature in the layer of pond;
The convolution feature is comprehensively input into full articulamentum.
5. method according to claim 1, it is characterised in that it is described according to comparison result in multiple object candidate areas Determine target following region, including:
Calculate the similarity in the tracked target region and the object candidate area;
Obtain maximum object candidate area score.
6. it is a kind of based on the method for tracking target for comparing network, it is characterised in that including such as lower unit:
First determining unit, for determining the current frame image comprising tracked target;
First acquisition unit, for obtaining tracked target region in the current frame image;
Second acquisition unit, the next two field picture for obtaining present frame;
3rd acquiring unit, for obtaining multiple object candidate areas in next two field picture;
4th acquiring unit, for obtaining the boundary rectangle comprising each object candidate area;
Comparing unit, is compared for the tracked target region and boundary rectangle input to be compared into neural network model It is right;
Second determining unit, for determining target following region in multiple object candidate areas according to comparison result.
7. device according to claim 6, it is characterised in that also include:3rd determining unit, for the target with Tracked target is determined in track region.
8. device according to claim 6, it is characterised in that the 3rd acquiring unit, including:
First acquisition module, positive sample and negative sample for obtaining the current frame image;
First detection module, for detecting the next frame image edge information;
Second detection module, for detecting that the tracked target region is believed with described image edge in the current frame image Cease the marginal information for overlapping.
9. device according to claim 6, it is characterised in that the comparing unit, including:
Mapping block, the convolution for the boundary rectangle of the object candidate area to be mapped into correspondence position in convolutional layer is special Levy;
Unified modules, for the dimension of the unification convolution feature in the layer of pond;
Input module, for being comprehensively input into the convolution feature in full articulamentum.
10. device according to claim 6, it is characterised in that second determining unit, including:
Computing module, the similarity for calculating the tracked target region and the object candidate area;
Second acquisition module, the object candidate area score for obtaining maximum.
CN201710038541.7A 2017-01-19 2017-01-19 A kind of method for tracking target and device based on comparison network Pending CN106920247A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710038541.7A CN106920247A (en) 2017-01-19 2017-01-19 A kind of method for tracking target and device based on comparison network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710038541.7A CN106920247A (en) 2017-01-19 2017-01-19 A kind of method for tracking target and device based on comparison network

Publications (1)

Publication Number Publication Date
CN106920247A true CN106920247A (en) 2017-07-04

Family

ID=59454141

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710038541.7A Pending CN106920247A (en) 2017-01-19 2017-01-19 A kind of method for tracking target and device based on comparison network

Country Status (1)

Country Link
CN (1) CN106920247A (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107492115A (en) * 2017-08-30 2017-12-19 北京小米移动软件有限公司 The detection method and device of destination object
CN107609485A (en) * 2017-08-16 2018-01-19 中国科学院自动化研究所 The recognition methods of traffic sign, storage medium, processing equipment
CN107992819A (en) * 2017-11-29 2018-05-04 青岛海信网络科技股份有限公司 A kind of definite method and apparatus of vehicle attribute structured features
CN108133197A (en) * 2018-01-05 2018-06-08 百度在线网络技术(北京)有限公司 For generating the method and apparatus of information
CN108596957A (en) * 2018-04-26 2018-09-28 北京小米移动软件有限公司 Object tracking methods and device
WO2019042419A1 (en) * 2017-09-04 2019-03-07 腾讯科技(深圳)有限公司 Image tracking point acquisition method and device, and storage medium
CN111709978A (en) * 2020-05-06 2020-09-25 广东康云科技有限公司 Cross-screen target tracking method, system, device and storage medium
CN112985263A (en) * 2021-02-09 2021-06-18 中国科学院上海微系统与信息技术研究所 Method, device and equipment for detecting geometrical parameters of bow net

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101290681A (en) * 2008-05-26 2008-10-22 华为技术有限公司 Video frequency object tracking method, device and automatic video frequency following system
CN104200210A (en) * 2014-08-12 2014-12-10 合肥工业大学 License plate character segmentation method based on parts
CN104200236A (en) * 2014-08-22 2014-12-10 浙江生辉照明有限公司 Quick target detection method based on DPM (deformable part model)
CN105224947A (en) * 2014-06-06 2016-01-06 株式会社理光 Sorter training method and system
CN105260997A (en) * 2015-09-22 2016-01-20 北京好运到信息科技有限公司 Method for automatically obtaining target image
CN105632186A (en) * 2016-03-11 2016-06-01 博康智能信息技术有限公司 Method and device for detecting vehicle queue jumping behavior
CN105678338A (en) * 2016-01-13 2016-06-15 华南农业大学 Target tracking method based on local feature learning
CN105933678A (en) * 2016-07-01 2016-09-07 湖南源信光电科技有限公司 Multi-focal length lens linkage imaging device based on multi-target intelligent tracking

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101290681A (en) * 2008-05-26 2008-10-22 华为技术有限公司 Video frequency object tracking method, device and automatic video frequency following system
CN105224947A (en) * 2014-06-06 2016-01-06 株式会社理光 Sorter training method and system
CN104200210A (en) * 2014-08-12 2014-12-10 合肥工业大学 License plate character segmentation method based on parts
CN104200236A (en) * 2014-08-22 2014-12-10 浙江生辉照明有限公司 Quick target detection method based on DPM (deformable part model)
CN105260997A (en) * 2015-09-22 2016-01-20 北京好运到信息科技有限公司 Method for automatically obtaining target image
CN105678338A (en) * 2016-01-13 2016-06-15 华南农业大学 Target tracking method based on local feature learning
CN105632186A (en) * 2016-03-11 2016-06-01 博康智能信息技术有限公司 Method and device for detecting vehicle queue jumping behavior
CN105933678A (en) * 2016-07-01 2016-09-07 湖南源信光电科技有限公司 Multi-focal length lens linkage imaging device based on multi-target intelligent tracking

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
NAM H 等: "Learning Multi-Domain Convolutional Neural Networks for Visual Tracking", 《COMPUTER SCIENCE》 *
ZITNICK C L等: "Edge Boxes:Locating object proposals from edges", 《COMPUTER VISION–ECCV 2014》 *
孔军 等: "一种面向高斯差分图的压缩感知目标跟踪算法", 《红外与毫米波学报》 *

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107609485A (en) * 2017-08-16 2018-01-19 中国科学院自动化研究所 The recognition methods of traffic sign, storage medium, processing equipment
CN107492115A (en) * 2017-08-30 2017-12-19 北京小米移动软件有限公司 The detection method and device of destination object
CN107492115B (en) * 2017-08-30 2021-01-01 北京小米移动软件有限公司 Target object detection method and device
WO2019042419A1 (en) * 2017-09-04 2019-03-07 腾讯科技(深圳)有限公司 Image tracking point acquisition method and device, and storage medium
US11164323B2 (en) 2017-09-04 2021-11-02 Tencent Technology (Shenzhen) Company Limited Method for obtaining image tracking points and device and storage medium thereof
CN107992819A (en) * 2017-11-29 2018-05-04 青岛海信网络科技股份有限公司 A kind of definite method and apparatus of vehicle attribute structured features
CN107992819B (en) * 2017-11-29 2020-07-10 青岛海信网络科技股份有限公司 Method and device for determining vehicle attribute structural features
CN108133197A (en) * 2018-01-05 2018-06-08 百度在线网络技术(北京)有限公司 For generating the method and apparatus of information
CN108596957A (en) * 2018-04-26 2018-09-28 北京小米移动软件有限公司 Object tracking methods and device
CN108596957B (en) * 2018-04-26 2022-07-22 北京小米移动软件有限公司 Object tracking method and device
CN111709978A (en) * 2020-05-06 2020-09-25 广东康云科技有限公司 Cross-screen target tracking method, system, device and storage medium
CN112985263A (en) * 2021-02-09 2021-06-18 中国科学院上海微系统与信息技术研究所 Method, device and equipment for detecting geometrical parameters of bow net

Similar Documents

Publication Publication Date Title
CN106920247A (en) A kind of method for tracking target and device based on comparison network
JP6842520B2 (en) Object detection methods, devices, equipment, storage media and vehicles
CN106909885A (en) A kind of method for tracking target and device based on target candidate
CN106920248A (en) A kind of method for tracking target and device
CN105260712B (en) A kind of vehicle front pedestrian detection method and system
Cao et al. Rapid detection of blind roads and crosswalks by using a lightweight semantic segmentation network
CN111027481B (en) Behavior analysis method and device based on human body key point detection
KR20200040665A (en) Systems and methods for detecting a point of interest change using a convolutional neural network
CN105260749B (en) Real-time target detection method based on direction gradient binary pattern and soft cascade SVM
CN108447078A (en) The interference of view-based access control model conspicuousness perceives track algorithm
CN107545263B (en) Object detection method and device
CN104590130A (en) Rearview mirror self-adaptive adjustment method based on image identification
CN110765906A (en) Pedestrian detection algorithm based on key points
CN102510734A (en) Pupil detection device and pupil detection method
JP7079358B2 (en) Target detection methods and devices, computer systems and readable storage media
CN104574393A (en) Three-dimensional pavement crack image generation system and method
CN105760846A (en) Object detection and location method and system based on depth data
CN103247038B (en) A kind of global image information synthesis method of visual cognition model-driven
CN109087337B (en) Long-time target tracking method and system based on hierarchical convolution characteristics
CN103455795A (en) Method for determining area where traffic target is located based on traffic video data image
CN110992424A (en) Positioning method and system based on binocular vision
CN104463240A (en) Method and device for controlling list interface
CN104376323A (en) Object distance determining method and device
CN111862511A (en) Target intrusion detection device and method based on binocular stereo vision
CN103006332A (en) Scalpel tracking method and device and digital stereoscopic microscope system

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20170704