摘要
提出一种基于分层自适应部分模型(HAPM,hierarchical adaptive part-based model)的目标检测方法,用于遥感图像的飞机目标检测。针对目前目标检测方法在子类型数目确定、模型多分辨率结构关系方面存在的不足,构造如下算法:首先构建一个扩展的Part-based Model模型;其次,分别从子类型的自适应选择、子类型的多层次建模和检测阶段加权距离变换的形变优化等方面对其进行改进;最后,HAPM算法充分考虑了模型的形变,同时结合多层次的建模思想使得目标的检测精度和算法适用性大大增强。用收集的10大国际机场的真实遥感图像数据进行实验验证,证明了算法的有效性。
A remote sensing image object detecting of aeroplane based on hierarchical adaptive part-based model is proposed. Because the present approaches take less of consideration on subtype choice and multi-resolution structure relationships, a new algorithm was put for- ward. First, an extern part-based model is built up. Then, we optimize its effects from sub- type adaptive choice, subtype hierarchical modeling and weighted distance transform. Final- ly, HAPM takes model deformation and together with hierarchical structure, which greatly improve the detection result and its application. By collecting the remote sensing images from ten airports, the effects of the new algorithm also has been tested. In a consequence, the results show that the new approach is worthwhile.
出处
《武汉大学学报(信息科学版)》
EI
CSCD
北大核心
2013年第6期656-660,共5页
Geomatics and Information Science of Wuhan University
基金
国家973计划资助项目(2013CB733404)
国家自然科学基金资助项目(60702041
41021061
41174120)
中国博士后科学基金资助项目
湖北省自然科学基金资助项目
关键词
层次自适应
多层结构
LSVM
目标检测
HAPM (hierarchical adaptive part-based model)
latent hierarchical structurelatent support vector machine(LSVM)
object detection
作者简介
何楚,博士,副教授,现主要从事图像分析与理解、SAR图像解译等方面的研究。E—mail:chuhe@whu.edu.cn