An Enhanced Segmentation Algorithm Based on Adaptive Threshold for Object Detection

• Main Authors: Chun-Ting Chen, 陳俊廷
• Other Authors: Chung-Yen Su
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/03659863525897076412

碩士 === 國立臺灣師範大學 === 應用電子科技學系 === 99 === Object detection Algorithm is a fundamental research for intelligent video system. It is widely used in intelligent surveillance system and Intelligent Transportation System (ITS). The purpose is to improve the performance of the system for object tracking or object recognition by detecting object effectively. In general, most of the research is using the difference between the frames to detect objects by background subtraction method. But it needs to set a fixed threshold to classify objects and background in traditional methods. It must redefine the threshold value when the testing environments are changed. Furthermore, shadow component is also a factor to interfere with the correctness of object detection algorithm. It needs to use a shadow removal algorithm to refine the parts of object after classifying objects and background. In this thesis, we present an enhanced segmentation algorithm with a pixel-dependent threshold for locating shadow regions according to the characteristic of shadow to reduce the interference of shadows and segment objects effectively in real-time. With that, the shadow regions are located more accurately and the moving objects are extracted more completely. Experimental results verify the proposed approach and show that it is helpful for object detection in various environments.