| Summary: | 碩士 === 國立雲林科技大學 === 電子與資訊工程研究所 === 99 === There is a growing emphasis on research in intelligent surveillance system recently. On the one hand, with the development of scientific and technological progress over time; on the other hand, the home, shopping malls, banks, airports and other security protection requirements are ever increasing. And because of the falling cost of hardware equipment and related surveillance systems industry to flourish, we can get better and faster results.
Moving object detection is significant process for surveillance system and makes use of stabilization in functional areas, such as tracking, classification, recognition, etc.There is higher computation complexity in other approaches and they can not achieve real-time result. To solve this problem, we proposed a novel object segmentation algorithm which can be fast and simple to identify moving objects in the video sequences. The traditional approachs are pixel-based to deal with; this article is block-based to do it. The key element of our algorithm is major color, and with the use of texture of the element to detect the moving objects from the video sequences. The fast moving object detection method we proposed can effectively solve the problems caused by the complex background. The frame rate can achieve 15 fps in a resolution of 768x576 pixels、68.45 fps in a resolution of 320x240 pixels、217.76fps in a resolution of 160x120 pixels. Compare with others, we use 44% memory at most.
|
|---|
