角点与边缘信息相结合的遥感图像飞机检测新方法

A New Approach to Detect Aircrafts in Remote Sensing Images Based on Corner and Edge Information Fusion

利用遥感图像中飞机目标的边缘特征和角点特征,提出边缘与角点信息相结合的遥感图像飞机检测新方法.首先,进行Canny边缘检测,提取遥感图像中飞机目标边缘信息,并利用OTSU算法二值化处理图像;然后,进行Harris角点检测,找出满足飞机角点分布特征的区域,去除伪目标;最后,进行区域生长式聚类,通过求取类心最终确定飞机位置.对60幅高分辨率遥感图像进行飞机检测测试,正确检测出238架飞机中的220架,漏检18架,48个虚警.实验表明,该方法可以有效解决复杂背景下飞机检测问题,具有良好的检测性能.

This paper presents a new method to detect aircrafts in remote-sensing image,in which the edge and corner features are used.Firstly,the whole image is operated by canny edge detector and then the OTSU binarization is employed to segment it.By these preprocessing procedures,the main information in the image is enhanced.Then Harris corner detection is conducted on the image,finding the areas where the corner quantity meets the standard.Finally the false alarms are removed by a region-growing clustering method.Experiments on 60 remote-sensing images are performed,including 238 airplanes.Totally 220 airplanes are detected,18 airplanes are missed and 48 false alarms are reported.The experimental results demonstrate that the newly proposed approach can effectively solve aircraft detection problem in complex background,and have a good detection performance.