基于无人机双目图像的线目标测量的研究

Research of artificial object measurement based on UAV binocular image

摘要:

利用无人机双目图像实现线目标的测量对输电线路巡检具有重要的意义。为提高无人机双目图像下线目标的测量精度，改进Census立体匹配算法，在代价聚合过程中，首先对聚合窗口中的初始匹配代价进行异常筛选，然后计算聚合代价值进而生成视差图，实验证明改进立体匹配算法，提高图像立体匹配精度，且平均误匹配率为5.79%；在线目标测量方面，针对线目标视差图存在的缺陷，提出一种基于目标识别的线目标视差图优化算法，该算法依据目标识别获取线目标视差图，然后根据四个原则进行优化处理，最后将优化后的线目标视差图用于测量，实验证明采用优化后的线目标视差图测量得到结果要优于直接采用视差图得到测量结果。

Abstract:

It is of great significance to use UAV binocular image to measure the line target, In order to improve the measurement accuracy of line targets in UAV binocular images, In the aspect of stereo matching algorithm, the Census stereo matching algorithm is improved. In the process of cost aggregation, the algorithm increases the value of the initial value of the initial match in the aggregation window and improves the accuracy of binocular stereo matching. Experiments show that the average false matching rate of improved stereo matching algorithm can reach 5.79%, In the aspect of drone line target measurement, aiming at the defects of the line target disparity map, an optimization algorithm of line target disparity map based on target recognition is proposed. The algorithm acquires line target disparity map according to target recognition. then according to the four principles of optimization processing, and finally to obtain the optimized line target disparity map, the experiment shows that using the optimized line target disparity map measurement results better than the direct measurement results.