Abstract: A video detection system of moving objects based on FPGA is designed，of which the foundation is background difference algorithm.Two novel techniques are introduced：Gray level shift and information fusion of multiple color space.Gray level shift is a newly proposed algorithm by the group recently when studying Gaussian mixture model，which means the shift of pixel gray level to adjacent pixels and often exhibits in the form of noise.Gray level shift can eliminate most of noise resulting from micro seism and air disturbance，et al in background difference algorithm.Information fusion of multiple color space is an effective approach to determine and track the moving objects by making use of some complementary components of several color spaces.These two techniques are implemented by FPGA，and satisfactory results are achieved in real time detection and tracking，which will play important role in enhancing the practical embedded video surveillance system.

Key words: Field Programmable Gate Array（FPGA）, video surveillance system, moving object detection, background differencing algorithm, gray level shift, information fusion

摘要： 利用FPGA实现了一个视频移动目标检测系统，算法的基础是背景差分法。为了准确检测和定位移动目标，利用了灰度漂移算法和多颜色空间信息融合算法。灰度漂移算法是课题组最近在研究高斯混合模型的过程中提出的一种消除背景差分法检测的移动目标存在过多噪声杂点的算法，可以非常有效地消除由于微震动、空气扰动等原因导致的像素灰度在邻近区域发生漂移的现象；多颜色空间信息融合就是利用多个颜色空间中的一些互补分量共同确定移动目标。利用FPGA实现了上述两种算法，在移动目标的实时检测和跟踪中取得了很好的效果，对于丰富和增强监控系统功能有重要意义。

关键词: 现场可编程门阵列, 视频监控, 移动目标检测, 背景差分法, 灰度漂移, 信息融合