Adaptable Block-Based Background Modeling and Real-Time Image Object Detection Algorithm

• Main Authors: Jian-Hui Chen, 陳建輝
• Other Authors: Ming-Hwa Sheu
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/53846358515843925731

碩士 === 國立雲林科技大學 === 電子與光電工程研究所碩士班 === 100 === The intelligent surveillance system makes people have more attention in recent years. At the same time, the advance CMOS processing technologies promote hardware performance in terms of cost, speed and power. In real applications, the demands for home security, shopping malls, and bank protection are growing with each passing day, and making the surveillance system becomes to a popular industry. For intelligent surveillance system, the moving object detection is an indispensable stage. The segmenting accuracy and speed will influence the final results of follow-up tracking and recognition processing. Many presented object detection methods are too complicated, such that they could not achieve real-time detection. To overcome this problem, we propose a fast and simple algorithm to effectively tackle the problem of excessive computational complexity. The traditional object detection methods construct the background model based on pixel-based. Recently, the block-based background is employed. In this thesis, we present adaptable block-based background model that uses major color to determine the color complexity of scene. Then we can divide the image into different size blocks through the color complexity of scene. The experiment results show that we can save 35% memory space than the latest existing algorithms at least. Finally, we can achieve 27.25 frames per second for the benchmark video with image size 768 576.