基于背景自适应生成的监控图像运动目标分割

MOTION OBJECT SEGMENTATION OF SURVEILLANCE IMAGE BASED ON SELF-ADAPTIVE BACKGROUND GENERATION

通过插值填充运动目标区域,生成首帧参考背景。背景更新时,结合当前帧和前一帧图像及其参考背景,根据亮度补偿动态生成当前帧参考背景。然后采用减背景方法对目标进行粗提取,再经目标阴影去除和形态学处理后分割出精确的运动目标。实验结果表明,该方法是有效可行的,能达到实时处理的要求,同时对场景光照变化具有较好的自适应性。

The first reference background frame is built by interpolating and filling the motion object regions. When updating the backgrounds, and according to the brightness compensation, the current frame reference background is dynamically generated with the combination of the current frame, the former frame and its reference background. By the way of deleting background, the objects are crudely picked up, and the precise motion objects are segmented after subtracting the object shadows and processing the morphology. Experiment results show that the method is feasible and effective, and it can achieve the real-time processing request. Meanwhile, it has better adaptability to the variety of scene illumination.