Dissertation > Excellent graduate degree dissertation topics show

Video object tracking based on mean shift detection algorithm and system implementation

Author: LiGuang
Tutor: ZhangHuaXiong；JinXueBo
School: Zhejiang University of Technology
Course: Signal and Information Processing
Keywords: Video target tracking Mean shift algorithm Kalman filter AdaBoost target detection OpenCV
CLC: TP391.41
Type: Master's thesis
Year: 2010
Downloads: 168
Quote: 1
Read: Download Dissertation

Abstract


Video target tracking is an important part of the research areas of computer vision, it combines the knowledge of pattern recognition, artificial intelligence, as well as automation and control, and many other related fields. The video tracking target position information extracted from the video track the target technology, can be used in intelligent video surveillance, video-based human-computer interaction, autopilot, in areas such as agriculture automation and medical image. Mean shift in recent years, the application of the algorithm in video object tracking aroused great concern, it has a good real-time and is not very sensitive to the target deformation and robustness. But it also has some shortcomings, such as window width can not be adaptive, and can not effectively track fast and large area obscured targets, the need to manually initialize the tracking. The mean shift algorithm has two versions: the standard Mean Shift algorithm and CamShift algorithm. The main line of the video object tracking based on mean shift algorithm, these defects do better improvement. Thesis content and innovation as follows: (1) a detailed study of the Mean Shift algorithm in the field of video object tracking demand parameter affine transformation solution algorithm adaptive window width. (2) based on tensor gradient histogram Mean Shift algorithm. Traditional Mean Shift algorithm is modeled in the color feature space, easy tracking fails when the target and background similar target space texture information, but it may be a good distinction between the target and background, this paper, the tensor gradient The histogram method not only the use of texture information but also the texture of the goal from the three-dimensional space are mapped directly to the one-dimensional space to reduce the amount of computation, tracking the effect of the target and background color similar case. (3) In order to solve the problems fast target tracking and target is blocked, the Kalman filter to predict the state of the target in the next frame, and predictive value as the starting point of the search target mean shift algorithm. In this paper, the use of similarity coefficient or the the Kalman residuals and threshold contrast similarity coefficients or the Kalman residuals are less than a certain threshold that the target is obscured Kalman prediction algorithm using only the track until the goal is to re-appear. (4) the use of the the AdaBoost target detection method the target offline learning in advance and obtain the characteristics of the target in a video sequence according to the target feature automatically detect target tracking to solve the problem of automatic target initialization. (5) In the OpenCV frame, designed and completed an experimental video tracking software systems, modular, by the target detection, briquettes detection, the briquettes tracking trajectory processing, and the trajectory generation modules, carried out experimental verification, and achieved good results. This provides a convenient system for the experimental tests of the future research work. Were studied under stationary and moving camera scene video target tracking, video object tracking single target tracking and partial occlusion treatment, the experimental results show that the algorithm of this paper is good, strong anti-jamming capability can the more complex context of work.

Related Dissertations

  1. Algorithm Research on SINS/GPS Integrated Navigation Syste,V249.328
  2. Satellite Attitude Determination Based on Gyro and Star Sensor,V448.2
  3. Research on Transfer Alignment of the Missile Onboard the Aircraft,V249.322
  4. Maneuver Detection and Tracking with the Radar Rate Measurement,TN953
  5. Research on Visual Detection and Tracking of Mobile Robots,TP242.62
  6. Design of Gyro Stabilized Pod Control System,TP273
  7. Direct Torque Control of Winding Induction Motor,TM343
  8. Time Delay Estimation in Distributed Optical Fiber Sensing and Positioning System,TN95
  9. Research of Moving Object Detecting and Robust Tracking Methods in Unstable Background,TP391.41
  10. Research and Application on Detection and Tracking of Object Based on Video Stream,TP391.41
  11. Design Control Algorithm of A Dynamic Positioning System for the Large Trailing Suction Hopper Dredger,U674.31
  12. Vehicle Detection and Tracking under Occlusion,TP391.41
  13. Research and Implementation of Non-rigid Medical Image Registration Method Based on Improved Optical Flow Model,TP391.41
  14. The Study of Fatigue Detection of Drivers Based on the Infrared Conditions,TP391.41
  15. Research of Channel Estimation Based on Pilot for MIMO-OFDM System,TN919.3
  16. Research on Face Detection System,TP391.41
  17. Research on Moving Object Detection and Tracking Algorithm in Video Surveillance,TP391.41
  18. On Application of GPS to Navigation and Positioning for Mini-UAV,V249.3
  19. The Study of Several Key Technologies of Micro-robot System for Cell Operation,TP242
  20. Researchand Design of Localization System in Wireless Sensor Networks,TN929.5
  21. The Study of Hedging Models Based on Shanghai and Shenzhen 300 Stock Index Futures,F224

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net  Mobile