针对无人飞艇地面目标检测中细节信息缺失的问题,提出一种静态目标和运动目标的检测方法.利用Lucas-Kanade方法跟踪目标区域内特征点,从而实现静态目标的连续检测.通过图像特征点的跟踪估计相邻帧图像间的全局运动,进而对图像进行运动补偿,利用补偿后的帧差图实现运动目标的检测.采用上海交通大学“致远一号”无人飞艇采集的实际视频数据进行实验与分析,结果验证了该方法的有效性.%Aiming at the problem of lack of information in ground targets detection of unmanned airship, this paper proposes a novel method for both moving and nonmoving ground objects detection. The nonmoving object is detected successively in the image sequence by tracking the feature points selected in the initial static object region with the Lucas-Kanade technique. For the moving object, the overall image motion between the adjacent frames is estimated by matching the feature points, and the overall frame motion is compensated. Appling the frame difference approach to the compensated frame and the current frame, the moving object is detected. The proposed method is evaluated and experiments are carried out with the real data provided by the unmanned airship, experimental results show that the effectiveness of the proposed methods.
展开▼
