Dissertation > Excellent graduate degree dissertation topics show
Multi-feature fusion of visual tracking algorithm
Author: ZhaoLing
Tutor: FengZuo
School: Huazhong University of Science and Technology
Course: Communication and Information System
Keywords: Visual Tracking Block-Based Tracking SVM classifier based tracking Gaussian mixture model Multiple Features
CLC: TP391.41
Type: Master's thesis
Year: 2011
Downloads: 52
Quote: 0
Read: Download Dissertation
Abstract
|
With the computer and network technology in the world within the scope of universal application , a large number of video surveillance resources have emerged , and how efficient use of these resources has become a priority. Visual tracking technology is the basis of related applications , the field of computer vision has become the most important direction of development. This paper introduces the visual tracking application , discusses the focus of visual tracking , pointing out the research background , purpose and content. Tracking algorithm then introduced the classification , as well as representatives of the classic algorithm introduces belong Bayesian filter Kalman filter and particle filter, as well as another classic mean shift algorithm . More concern in recent years while a new tracking algorithm based on block -based classifier tracking algorithm, based on level set tracking algorithm as well as a combination of multi- tracker tracking algorithms have also been introduced . On this basis , and pointed out that the current algorithm analyzes some existing problems and shortcomings , presented our research directions and contributions. This paper presents the appearance of the model based on adaptive block tracking algorithm , the introduction of Gaussian mixture model to model the target and template updates. The algorithm can be changed according to the appearance of the moving object , position and occlusion phenomena adaptive template to implement robust tracking. Through a comprehensive qualitative and quantitative experimental results demonstrate that the proposed algorithm is superior to existing tracking effect similar algorithms . This paper also proposed a multi- feature fusion classifier based tracking algorithm, which will be tracked objects as positive examples , the background and other non- target areas as counter-example , if we can use the classifier ( eg SVM) to two class effectively separate, robust tracking can be achieved . The effect depends largely on the characteristics of the object used to express whether an object can reflect the uniqueness and difference. This paper presents a method using a combination of structure, texture and color features to express the tracking target, where structural component with edge direction histogram , texture component based on texture primitive way of expressing the coefficient vector . Experimental results show that the proposed algorithm is better than existing similar tracking algorithms. Finally, the main research and work have been summarized , and the need to further study and work.
|
Related Dissertations
- Detection and Tracking of Moving Object in Complex Background,TP391.41
- Study on Visual Target Detection Based on SVM,TP391.41
- Research of Moving Object Detection and Tracking Technology,TP391.41
- Video surveillance system moving target detection algorithm,TP391.41
- Research of Hybrid EM Algorithm Based on Gaussian Mixture Models and Its Clustering Application,TN911.7
- Design and Research for a Kind of Hemispheric Differedtial Motion Spherical Robot,TP242
- The Research of Unstructured Road Detection Algorithm Based on Color Features and B-spline and Its Application,TP391.41
- Wideband audio technology nonlinear band broadening,TN912.3
- Content-based image retrieval key technology research,TP391.41
- Research on the Key Technologies of Intelligent Transportation Video Surveillance System,TP391.41
- Distributed sound source localization and tracking algorithm,TN912.3
- Embedded target detection and tracking system design and algorithm implementation,TP391.41
- Image Spam Detecting Based on Combinatorial and Statistical Classifier,TP391.41
- Key Algorithm in High Quality Voice Conversion System,TN912.3
- Research on Speaker Recognition and Its Robustness in Noise Environment,TN912.34
- Detection instead of classification of brain-computer interface research,TN911.7
- Conference room environment based on audio and video information fusion speaker recognition,TN912.34
- Research on Language IDE NT Ification,TN912.34
- GMM - UBM model - based language recognition,TN912.34
- Research on Moving Object Detection Technology for Intelligent Visual Surveillance,TP391.41
- Research of Feature Extraction for Voice Based on Wavelet Transform,TN912.34
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
|