On storage, search and object identification of video sequence

Pao Yue-kong Library Electronic Theses Database

On storage, search and object identification of video sequence

 

Author: Wu, Hao
Title: On storage, search and object identification of video sequence
Degree: M.Phil.
Year: 2016
Subject: Digital video.
Image processing -- Digital techniques.
Hong Kong Polytechnic University -- Dissertations
Department: Dept. of Electronic and Information Engineering
Pages: xii, 101 pages : color illustrations
Language: English
InnoPac Record: http://library.polyu.edu.hk/record=b2890635
URI: http://theses.lib.polyu.edu.hk/handle/200/8496
Abstract: Digital video is an important information carrier nowadays, as a representation of moving visual images in the form of encoded digital data. Video based applications are more and more popular, better knowledge of video coding, video retrieval, video surveillance, robot vision, etc are always required. In this thesis, we initially introduce some applications with existing approaches, including video recording for surveillance, smart robot vision and driver assistant system; followed by the improvement work we have done correspondingly. In the video coding for surveillance video, the approach is very different from normal video coding because of high temporal redundancy. Conventionally, large Group of Pictures (GOP) and long term reference frame will be used to set up surveillance video encoder. In addition, some research works make use background extraction to help encoding process. Through these approaches, some bit-rate reduction can be achieved to encode surveillance videos. However, conventional methods are trapped into complying encoding standard with just one encoder. In this thesis, we propose a double encoder coding scheme. Foreground and background materials are firstly extracted and processed as masked foreground and condensed background sequences. Then separate encoders are used to encode foreground and background sequences. According to our experimental results, A large bit-rate reduction can be achieved through the proposed scheme. With defined side information, the video can be well reconstructed at decoder side without foreground distortion.
Moving object detection from a moving camera is a fundamental task in many applications, such as smart robot. The fundamental assumption of conventional moving object detection method is that the background is either static or moving as a 2D plane. However, for the moving robot car vision, the background movement is 3D motion structure in nature. In this situation, the conventional moving object detection algorithm cannot handle the 3D background modeling effectively and efficiently. We have proposed a novel scheme by utilizing the motor control signal and depth map obtained from a stereo camera to model the perspective transform matrix between frames under a moving camera. Hence, the relationship between a static background pixel and the moving foreground corresponding to the camera motion can be related by a perspective matrix. The proposed scheme is able to detect moving objects in our moving robot car efficiently. Different from conventional approaches, our method can model the moving background in 3D structure, without online model training. More importantly, the computational complexity and memory requirement are low, making it possible to implement this scheme in real-time, which is even valuable for a robot vision system. Rail extraction is a fundamental and important step in railway Driver Assistant System, which is now an important application of image processing. The task is challenging as the railway is exposed to different environments. In this research, we propose a railway extraction scheme, using a novel connectivity measure method named Angle Alignment Measure. The proposed scheme is robust to luminance and color variation, without explicit edge extraction process. Railways with different lengths and patterns can be extracted under various lighting and weather conditions. More importantly, the computational complexity of the proposed scheme is very low, requiring only on average 26ms to process a frame on a smart phone and 5.5ms on a desktop computer, which are significantly better than other algorithms in the literature.

Files in this item

Files Size Format
b28906354.pdf 3.285Mb PDF
Copyright Undertaking
As a bona fide Library user, I declare that:
  1. I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
  2. I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
  3. I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.

     

Quick Search

Browse

More Information