Dissertation > Excellent graduate degree dissertation topics show

Research on Moving Object Detection and Shadow Suppression

Author: XuZhu
Tutor: LinQing
School: Jiangsu University
Course: Applied Computer Technology
Keywords: Adaptive Gaussian mixture model Moving target detection HSV Shadow Elimination
CLC: TP391.41
Type: Master's thesis
Year: 2010
Downloads: 368
Quote: 3
Read: Download Dissertation

Abstract

In recent decades, with the increasing development of computer technology and image processing technology, target detection in sports video has been widely applied to many other areas of national defense and national economic construction. With the continuous expansion of its field of application, detection and shadow of the moving object in a video sequence to eliminate a lot of detailed and in-depth research, many effective methods. Moving target detection and shadow elimination due to its potential applications in the field of intelligent monitoring and potential economic value has become one of the hot topics of the computer industry. First of all, This paper reviews the research background and significance of the study a number of domestic and foreign, compared to its performance advantages and disadvantages of the classical theory and methods of moving target detection based on video image. In this paper, the classical algorithm mainly includes inter-frame difference method, optical flow method, background subtraction, Gaussian background modeling method. The basic idea and the realization of the principle of these algorithms and experimental comparison of these methods are discussed in detail. Secondly, the paper focuses on the Gaussian mixture model background modeling method. The initial frame of the traditional method for inaccurate, resulting in slow convergence when the model identification, this paper proposes a combination of the method of the median method, and all the value method, resulting in improved access to the accuracy of the initial frame. This paper also introduces a reflection of the current frame and the model matches the parameters of the effective pixels to improve the background model update algorithm, you can prevent the slow movement of the object gradually blend into the background. In addition, the traditional method of computing a large amount of space overhead, in order to improve the efficiency of modeling on the basis of the traditional Gaussian mixture background modeling, proposed a new Gaussian mixture model Gaussian component The number of adaptive selection strategy. The new method can reduce the overhead in time and space, quickly and accurately to establish the background of the model, in order to achieve the good result of the segmentation of the foreground object. Finally, due to the influence of the light conditions, most of the image there is a shadow, the shadow of the presence of interference target detection. Based on shadow in HSV vector space characteristics, proposed a new threshold shadow elimination method can efficiently detect a moving target with a shadow. And traditional shadow segmentation method, the method in the case of the threshold value is not set, to eliminate the shadow of moving target, the robustness and practicality.

Related Dissertations

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