关于摄像运动目标图像优化检测仿真研究

The Simulation Research on Optimizing Detection of Moving Object Image

背景差分法在摄像头运动的情况下,无法有效进行目标检测。通过补偿全局运动更新背景模板,可以有效抵消摄像头运动带来的影响。但由于背景图像中物体景深差异较大时很难获取精确的背景模板,提出一种分层背景模型的摄像运动目标图像优化检测方法。首先,采用SURF算法获取相邻图像特征点运动矢量。接着采用图像腐蚀技术自动判断运动矢量的类别数并完成聚类;随后根据聚类结果对图像进行分层,并在不同层上分别进行仿射变换;最后叠加生成变换后的新背景图像。引入PSNR值对背景精确程度进行衡量。仿真结果表明,分层背景模型能够有效消除物体景深差异带来的干扰,产生更加精准的背景模板。说明上述优化检测方法能够有效完成摄像运动目标的检测任务。

Background subtraction algorithm may fail in object detection when camera moves. In order to remove the interference caused by the moving camera, the background model is updated after compensating the global mo- tion. In order to acquire an accurate background, an optimizing algorithm of moving object detection is proposed. Firstly, the SURF algorithm is applied to compute the motion vectors between contiguous images. Then the vectors are classified into several categories automatically based on image erosion. The background image is segmented into mul- tiple layers according to the result of classification. Affine transformation is performed in different layers separately. Finally, the transformed images of whole updated background are obtained. PSNR is introduced to measure the accu- racy of background model. Simulation result shows that, the multi-layer background model can accurately remove the interference caused by difference of objects＇ depth and generated background, and the proposed algorithm can effec- tively detect moving objects when the camera moves.