利用改进的背景差法进行运动目标检测

Moving object detection with improved background subtraction

为了更准确地划分运动目标和背景区域,提出了基于背景差法的改进算法。该算法将选择更新法与中值滤波法相结合,对背景建模与阈值选取进行了优化,解决了复杂场景背景建模不理想问题,并能快速有效地进行背景更新。该算法还提出了双阈值思想,根据场景信息自适应地在图像的不同区域采用不同阈值,将当前帧图像与背景图像的差分作为反馈来调整阈值。实验结果表明,该算法能够较快速准确地提取运动目标,并对光照变化、场景变化、运动干扰有较好的鲁棒性。

In order to mark the moving targets and their background regions,an improved algorithm based on the background subtraction is proposed.The improved algorithm combined the selection update method with the median filtering,optimized the background modeling and the threshold selection,and solved the problem that the background modeling is unsatisfactory in the complex situation.It can quickly and efficiently update the background.A thought of dual-threshold method is proposed in the improved algorithm.It can adaptively adopt the different threshold values in different regions of an image according to the field information,and adjust the threshold values by taking the difference of the current frame image and background image as the feedback.The experimental results indicate that the improved algorithm can rapidly and accurately pick up moving objects,and has a very good robustness against illumination variance,scene change and movement interference.