非合作区域星载遥感图像的车辆目标检测

针对非合作区域遥感图像中复杂背景下车辆目标检 测困难的问题, 提出一种改进形态学重建的车辆目标检测方法。首先,利用遥感图像的近红外和红色波段的 信息,获取场 景的归一化植被指数(NDVI),去除植被背景的干扰；其次,设计方向模板, 对全色图像进行滤波,使人 造物背景得以完整保留,并生成标记图像,利用标记图像重建得到人造物背景；最后,利用 原图和重建背 景之间的差异,进行车辆检测,有效消除了复杂背景的干扰。将本文提出的方法应用于实际 遥感图像的车 辆目标检测,结果表明,检测效果好,鲁棒性强,无需先验信息,可用于大幅面遥感 图像的车辆目标自动提取。

Vehicle targets detection in non-cooperative areas using satellite remote sensing images

Detecting vehicle targets in non-cooperative areas is of great import ance in both military and civil fields. For the intensive interferences of complex background in satellite images,it is a challenging task to detect vehicle targets automatically from non-cooperative areas using satellite images.In thi s paper,a new vehicle target detection approach is presented,which introduces a modified grayscale morpholog ical reconstruction to reduce the interferences of complex background in non-cooperative areas.Initially,the norm alized difference vegetation indices (NDVIs) are obtained by using the near infrared and the red bands of a satellite image.And the NDVIs are used to remove the vegetation pixels from the satellite image.Next,a marker image is g enerated by filtering the panchromatic image using a group of directional filters that are designed to pre serve man-made objects.And the man-made objects in the satellite image are acquired by a grayscale reconstruct ion from the marker image.Finally,the vehicle targets are detected by using the difference between the original satell ite image and the reconstructed image,which can eliminate the interferences of complex background.The proposed method is applied to real satellite images for vehicle targets detection.The experimental results show th at our proposed approach has high detection rate,good robustness and non-manual assistance.It can be used for detecting vehicle targets automatically in large scale satellite image s.