论文部分内容阅读
针对红外成像末导引阶段飞行器姿态调整及高速运动导致的目标尺度和姿态迅速变化的问题,提出了一种基于加速鲁棒特征(SURF)的红外成像目标跟踪算法。为了实现在末导引阶段对目标进行精确跟踪,采取了点跟踪的策略。首先根据跟踪点在上一帧的位置,在当前帧选取以相同位置为中心的图像子块并求其SURF特征,通过SURF特征匹配得到当前帧图像子块和模板的匹配点集,采用随机抽样一致性(RANSAC)算法剔除误匹配点对,进一步用最小二乘算法(LSA)精确地估计出对应的单应性矩阵;然后通过单应性矩阵把跟踪点映射到当前帧获取跟踪点在当前帧的位置,从而实现精确跟踪。试验结果表明,本文算法有较高的跟踪精度和较好的实时性。
In order to solve the problem of rapidly changing the target scale and pose caused by the high-speed motion during the final stage of infrared imaging, an infrared imaging target tracking algorithm based on Accelerated Robustness Feature (SURF) is proposed. In order to achieve the goal of precise tracking in the last stage, a point tracking strategy was adopted. Firstly, according to the location of the previous tracking point in the previous frame, the image sub-block centered at the same location is selected and the SURF features are obtained. The matching point sets of the sub-blocks and the template of the current frame are obtained by SURF feature matching. Random sampling (RANSAC) algorithm to remove the mismatched point pairs and further to estimate the corresponding homography matrices by the least-squares algorithm (LSA). Then, the tracking points are mapped to the current frame through the homography matrix to obtain the current tracking points at the current Frame position, in order to achieve accurate tracking. Experimental results show that the proposed algorithm has higher tracking accuracy and better real-time performance.