基于压缩特征的鱼眼视频目标跟踪算法研究

Research on Target Tracking Algorithm from Fisheye Camera Based on Compressive Sensing

该文针对畸变严重的鱼眼图像中的目标跟踪,提出一种能适应尺度变化、姿态变化以及形状畸变的鱼眼视频目标跟踪的方法。该方法首先将灰度特征和相对梯度特征相结合得到目标的高维特征,然后对其平均降维得到目标的压缩特征。并根据鱼眼成像模型得到投影点的运动特性,确定目标的运动范围。为了适应尺度变化,在块匹配运动估计思想的基础上,对目标跟踪框的顶点分别进行由粗到精的定位,并在此过程中根据跟踪框的尺度相应改变压缩特征的尺度。实验结果表明:该算法在目标畸变、尺度变化、姿态变化以及局部遮挡等情况下,判断指标均优于其他对比算法。

For object detection in fisheye images which present serious distortion, an object tracking method is proposed to deal with scale variance, pose change and distortion. Firstly, gray feature and gradient feature are combined to obtain a high dimensional feature of the target, then reduce its dimensionality by averaging to obtain target’s compressive feature. According to fisheye imaging model, motion of object point is modeled, and range of motion of target is “predicted”. In order to adjust to scale variance, corner points are positioned respectively in a coarse to fine manner based on the block matching motion estimation, and the scale of compressed feature is changed along with scale change of object box. Experimental results show that the proposed algorithm is superior to other algorithms in the case of distortion, scale change, pose change and part occlusion.