摘要
提出了一种基于高分辨率卫星遥感图像检测汽车的新方法——背景迭代搜索(Background Iterative Search,BIS)算法。该算法首先利用背景与目标的局部差异,用距离作为判别准则逐步迭代搜索并去除背景,根据汽车的物质特性初步检测汽车;然后采用动态双峰阈值分割方法,利用全局信息把道路和非道路分开,并根据形状特征粗略提取道路;最后利用道路信息约束初步检测的汽车,得到最终的汽车检测结果。通过使用IKONOS和QuickBird卫星遥感数据进行实验,验证了BIS算法的有效性。
In the traditional Space-to-Earth car detection system,thermal infrared,radar or aerial image data are often used,while high-resolution satellite remote sensing data have rarely been employed.To solve this problem,this paper proposes a new method for car detection based on high resolution satellite images,which is named BIS（Background Iterative Search）.Firstly,according to the local differences between the object and the background,the background is searched and removed,and the preliminary car detection is achieved based on the material properties of cars.Secondly,the dynamic twin peak threshold method is used to separate roads from non-roads,and roads are roughly extracted based on shape features.Lastly,the correct car objects are obtained by constraining the elementary ones with the derived road information.The BIS method was applied with IKONOS and QuikBird data and proved to be effective.
出处
《国土资源遥感》
CSCD
2011年第4期46-51,共6页
Remote Sensing for Land & Resources
基金
国家863项目(编号:2008AA121504)
国家科技支撑计划项目(编号:0914131)共同资助
关键词
高分辨率卫星遥感
汽车检测
背景迭代搜索
High resolution satellite remote sensing
Car detection
Background iterative search
