猪舍固定摄像背景去除法*

针对主要的背景去除法在养殖场猪舍视频监控应用中的不足，提出一种背景去除法。通过初始化背景帧、以并行结构进行对称差分和背景减、进行视频状态分析、以状态码为控制条件分别进行背景更新和前景融合，改进的背景更新算法能适应没有前景和有前景、背景光线突变和渐变等情况，前景融合算法仅提取包含阴影在内的前景区域；改进的阴影检测算法以并行结构联立阴影判别条件；最后，获得当前帧的精确前景。实验显示本方法具有自适应性、低计算代价，为后继的猪视频智能分析创造有利条件。

Background subtraction for static camera in pigpen

This paper proposed a new background subtraction for static camera in pigpen after analying the faults of primary related methods. First initialized a background frame, and then executed symmetrical difference and background difference respectively. The state codes resulted from analyzing video was regarded as controlling condition of background updating and foregrounds fusing. Background updating was available for the foregrounds of existence or nonexistence under critical or gradual situations like illumination variation. The foreground fusion only drew the areas that contain both shadows and pigs. The shadows detection in parallel structure grouped criterion conditions. Finally,obtained the precise foregrounds of the current frame. Experiments show that this method is adaptive and computionally inexpensive,and favored the next intelligent video analyzing.