基于仿射不变量的运动目标遮挡检测和跟踪

Occlusion Detection and Tracking in Image Sequence Based on the Affine Invariant

图像序列中运动目标的遮挡检测和处理是目前在动态图像处理时经常碰到,且比较难于解决的问题。本文针对复杂背景下图像序列的特性,给出了一种基于仿射不变量的多运动目标的遮挡检测和跟踪方法。该法首先给出了仿射不变量向量的构造,以及基于它的遮挡判决准则;然后,结合被跟踪目标的先验知识,给出丁,轨迹线以及遮挡区域的预测方法;最后,给出了依赖于仿射不变量向量的运动目标遮挡和跟踪的详细算法。实验表明新算法较好的解决了复杂背景下序列图像中的运动目标遮挡检测和跟踪的问题,较大的提高了运动目标跟踪算法的鲁棒性;同时新算法具有较快的响应速度。

In processing the dynamic images, occlusion detection and tracking of moving targets in an image sequence are very significant and difficult problems. The objective of this research was to present a novel method to enhance the efficiency and accuracy of occlusion detection and tracking in the image sequence, and to resolve the detection difficulties brought by targets distortion while those targets moved. Firstly, the construction method of affine invariant vector was proposed, and the principles of occlusion detection based on the affine invariant vector were suggested. Secondly, the trajectory prediction technique and the occlusion area prediction method, which employed some previous information of the moving target, were presented. Finally, the detail theory and algorithm of new occlusion detection and tracking method, which were based on the affine invariant vector, were described. The results indicated that the moving objects occlusion detected by the new technique were more robust than those detected by traditional methods, In addition, the detection speed had been become more rapid, The results seem to suggest that the new method be able to provide efficiency and accuracy advantages in computer version.