快速背景重建的在线运动目标检测

为了能快速地从视频图像序列中创建可靠的背景图像,进而提取运动目标,文中提出了一种基于反馈信息的运动目标检测算法.首先提出了基于相邻帧信息和背景估计信息相融合的背景重建算法,保证了在视频场景改变时仍能迅速捕捉背景;还提出了基于一种在线Otsu法的运动目标检测,将相邻帧运动目标信息反馈到目标提取算法中,弥补传统Otsu法的不足;最后提出了对光线变化具有一定鲁棒性的背景估计算法.实验表明,该方法的重建速度快,准确率高,能满足实时检测的需要.

Online moving object detection based on fast background reconstruction

In order to reconstruct reliable background image in video sequences, and to pick up moving objects, an algorithm of detecting moving objects based on feedback information is presented. Firstly, fast background reconstruction method based on the information of adjacent frame and background estimation is applied to insure that the backgrounds are taken all the same when scene is changing. Then, an improved Otsu method which feed back parameter of object in adjacent frame to segmentation algorithm is introduced, Lastly, adaptive background estimation which is robust to light varying is presented. The experiment results show that this algorithm has the advantages of veracity and real time.