CN106780563A - A kind of image characteristic point tracing for taking back light-metering stream - Google Patents
A kind of image characteristic point tracing for taking back light-metering stream Download PDFInfo
- Publication number
- CN106780563A CN106780563A CN201710047159.2A CN201710047159A CN106780563A CN 106780563 A CN106780563 A CN 106780563A CN 201710047159 A CN201710047159 A CN 201710047159A CN 106780563 A CN106780563 A CN 106780563A
- Authority
- CN
- China
- Prior art keywords
- tracking
- point
- optical flow
- result
- traceback
- 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
Landscapes
- Image Analysis (AREA)
Abstract
A kind of characteristics of image tracing for taking back light-metering stream, is that feedback is introduced in traditional optical flow tracking method, excludes the big tracking result of error, is retained in the tracking result in certain accuracy rating.Feedback is introduced by returning survey, will primitive character tracking result, by reverse optical flow tracking, obtain traceback result on the original image, the position of primitive character point and the position of traceback result are compared, point of the error in the range of requiring in advance is taken.
Description
Technical field
The invention belongs to the field of computer vision and image procossing, can be applied to preceding scenery motion in sequence image with
The tracking that track or camera are moved in itself.
Background technology
Light stream concept is that American Psychologist Gibson is proposed the forties in last century year, for describing motion to animal
The visual stimulus of generation this physical phenomenon.After the eighties in last century, machine vision researchers utilize this physical phenomenon
The theory and technology that estimation is carried out based on orderly image sequence are developed.Optical flow method estimation is based on 3 points of basic vacations
If that is, so-called brightness constancy, Time Continuous are consistent with space.Briefly obtained in very of short duration adjacent image frame
In time, change cancellation of the brightness for very little image-region in consecutive frame.Thus can be with before on estimated sequence image
The position of point on one two field picture on latter two field picture.This method is just called the optical flow tracking method put on sequence image.It is right
Each point is tracked referred to as dense optical flow tracing on sequence image, and it has the disadvantage that amount of calculation is too big;Sparse optical flow with
Track method is only tracked to some characteristic points on image, substantially increases the arithmetic speed of optical flow method, therefore optical flow tracking method
It is widely used.
It is low precision that optical flow tracking method is used for its fatal shortcoming during visual odometry, and this is because the actual image for obtaining is present
There is the factors such as error in noise, calculating.Therefore the method that the researchers of visual odometry mainly use Feature Points Matching, i.e.,
Every piece image in image sequence carries out independent feature detection, and characteristic matching is then carried out again, finds characteristics of image
Point has the relevant position in front and rear two images.But feature matching method is computationally intensive, and speed is low, as its practical road
Very big obstacle on road.
The fireballing feature of this research and utilization optical flow tracking method, proposes a kind of optical flow tracking method for taking back survey, it is therefore an objective to improve light
Flow the accuracy of tracking.
The content of the invention
The optical flow tracking algorithm of proposed by the present invention time survey can be described as follows:Assuming that current reference frame is the i-th frame(Or be
Primitive frame), the characteristic point location of pixels detected on the i-th frame is point xi(Or be primitive character point), next frame is jth frame,
There is 1 tracking result on jth frame, be set to xji;With the tracking result x of jth framejiAs characteristic point, then reversely utilize optical flow method
It is found out in the i-th tracking result xij(Or be traceback result).Thus defining back survey tracking error E is, E=| xi-xij|。
This to carry out traceback using optical flow method again in the tracking result of a later frame, then the position with primitive character point is entered
The method that row compares is exactly proposed by the present invention time survey optical flow method.Error E is surveyed to first time forward trace further with returning
Result is accepted or rejected, it is ensured that the accuracy of tracking.Return survey introducing, it substantially introduces feedback, by feedback error come
Judge whether the tracking result of primitive character is reliable.
Claims (2)
1. a kind of image characteristic point tracing for taking back light-metering stream, is characterized in using light stream to the primitive character point of sequence image
Method obtains tracing positional in the next frame, then utilizes optical flow method traceback using the tracing positional point as characteristic point, obtains
Traceback position on original image, the error according to primitive character position and traceback position is to the tracking in next frame
Result is accepted or rejected.
2. according to claim 1, it is characterized in introducing feedback in optical flow tracking method, excludes the big feature of tracking error
Point, is retained in the tracking result in default accuracy rating.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710047159.2A CN106780563A (en) | 2017-01-22 | 2017-01-22 | A kind of image characteristic point tracing for taking back light-metering stream |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710047159.2A CN106780563A (en) | 2017-01-22 | 2017-01-22 | A kind of image characteristic point tracing for taking back light-metering stream |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106780563A true CN106780563A (en) | 2017-05-31 |
Family
ID=58944003
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710047159.2A Pending CN106780563A (en) | 2017-01-22 | 2017-01-22 | A kind of image characteristic point tracing for taking back light-metering stream |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106780563A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103324932A (en) * | 2013-06-07 | 2013-09-25 | 东软集团股份有限公司 | Video-based vehicle detecting and tracking method and system |
CN103871079A (en) * | 2014-03-18 | 2014-06-18 | 南京金智视讯技术有限公司 | Vehicle tracking method based on machine learning and optical flow |
CN103985137A (en) * | 2014-04-25 | 2014-08-13 | 北京大学深圳研究院 | Moving object tracking method and system applied to human-computer interaction |
CN104463859A (en) * | 2014-11-28 | 2015-03-25 | 中国航天时代电子公司 | Real-time video stitching method based on specified tracking points |
-
2017
- 2017-01-22 CN CN201710047159.2A patent/CN106780563A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103324932A (en) * | 2013-06-07 | 2013-09-25 | 东软集团股份有限公司 | Video-based vehicle detecting and tracking method and system |
CN103871079A (en) * | 2014-03-18 | 2014-06-18 | 南京金智视讯技术有限公司 | Vehicle tracking method based on machine learning and optical flow |
CN103985137A (en) * | 2014-04-25 | 2014-08-13 | 北京大学深圳研究院 | Moving object tracking method and system applied to human-computer interaction |
CN104463859A (en) * | 2014-11-28 | 2015-03-25 | 中国航天时代电子公司 | Real-time video stitching method based on specified tracking points |
Non-Patent Citations (2)
Title |
---|
云红全等: ""基于超像素时空显著性的运动目标检测算法"", 《红外技术》 * |
曹宇: ""基于特征点反向跟踪和光流聚类算法的渐进汽车检测算法"", 《智慧工厂》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Mahjourian et al. | Unsupervised learning of depth and ego-motion from monocular video using 3d geometric constraints | |
Zhan et al. | Visual odometry revisited: What should be learnt? | |
Stone et al. | Smurf: Self-teaching multi-frame unsupervised raft with full-image warping | |
Wedel et al. | Efficient dense scene flow from sparse or dense stereo data | |
Brebion et al. | Real-time optical flow for vehicular perception with low-and high-resolution event cameras | |
Lu et al. | Deep unsupervised learning for simultaneous visual odometry and depth estimation | |
Vankadari et al. | When the sun goes down: Repairing photometric losses for all-day depth estimation | |
Wang et al. | Self-supervised monocular depth estimation with direct methods | |
Li et al. | Scale-aware monocular SLAM based on convolutional neural network | |
Zhao et al. | Transformer-based self-supervised monocular depth and visual odometry | |
CN106780563A (en) | A kind of image characteristic point tracing for taking back light-metering stream | |
Kuschk et al. | Real-time variational stereo reconstruction with applications to large-scale dense SLAM | |
Patra et al. | Ego-slam: A robust monocular slam for egocentric videos | |
Bello et al. | A novel monocular disparity estimation network with domain transformation and ambiguity learning | |
Harisankar et al. | Unsupervised depth estimation from monocular images for autonomous vehicles | |
Xie et al. | Hierarchical quadtree feature optical flow tracking based sparse pose-graph visual-inertial SLAM | |
Fu et al. | Distractor-aware event-based tracking | |
Gan et al. | A dynamic detection method to improve SLAM performance | |
Li et al. | Unsupervised joint learning of depth, optical flow, ego-motion from video | |
Wu et al. | Interactive multi-camera soccer video analysis system | |
Babu V et al. | A deeper insight into the undemon: Unsupervised deep network for depth and ego-motion estimation | |
Kowalczuk et al. | Real-time temporal stereo matching using iterative adaptive support weights | |
Wang et al. | Physical Priors Augmented Event-Based 3D Reconstruction | |
Xu et al. | DOS-SLAM: A real-time dynamic object segmentation visual SLAM system | |
Rzeszutek et al. | Efficient automatic depth estimation for video |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20170531 |
|
WD01 | Invention patent application deemed withdrawn after publication |