高分辨率卫星影像车辆检测研究进展  

Progress in Vehicle Detection from High Resolution Satellite Imagery

作　　者：刘珠妹[1,2] 刘亚岚[1] 谭衢霖[3] 任玉环[1] 

机构地区：[1]中国科学院遥感应用研究所,北京100101 [2]中国科学院研究生院,北京100049 [3]北京交通大学土木建筑工程学院,北京100044

出　　处：《遥感技术与应用》2012年第1期8-14,共7页Remote Sensing Technology and Application

基　　金：国家自然科学基金项目"基于小目标探测的高分辨率遥感影像交通参数提取研究"(40801121)

摘　　要：高分辨率卫星遥感技术具有在更小的空间尺度上探测地表目标的能力,利用其影像数据进行车辆检测已成为新的研究热点。在概述遥感影像车辆检测研究现状的基础上,对车辆目标影像特征及车辆检测过程进行了探讨;将车辆检测方法分为利用光谱/几何结构特征的基本检测方法和综合运用多种特征的智能化检测方法,并详细叙述了多种车辆检测方法的原理与适用性以及车辆提取中的关键技术。通过分析发现:结合多特征的机器学习和面向对象的车辆检测方法更适合较复杂环境下的车辆检测。The high resolution satellite technique has an ability to detect the smaller targets on the ground.Vehicle detection using high resolution satellite imagery has become a hotspot in remote sensing research field.This paper firstly analyzed the study situation of vehicle detection from remote sensing imagery,and then discussed the imagery features of vehicles,the process and the primary methods for vehicle detection,including the basic detection methods using spectrum/structure features of vehicles and the intelligent methods combining various of vehicle features.The crucial technologies to detect vehicles were introduced in detail as well.Finally,this paper concluded that methods such as machine learning or object-oriented method using more object features can be more adaptable to complex road environment.

关 键 词：智能交通系统 车辆检测 卫星遥感 机器学习 面向对象 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]