HJ-1光学遥感影像船只检测——以鄱阳湖为例  

Detection of Ships in Homogeneous Regions from HJ-1 Optical Remote Sensing Images:A Case Study at the Poyang Lake

作　　者：李建[1] 陈晓玲[1] 刘子潇[2] 田礼乔[1] 

机构地区：[1]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079 [2]天津市测绘院,天津300381

出　　处：《武汉大学学报（信息科学版）》2015年第6期727-731,共5页Geomatics and Information Science of Wuhan University

基　　金：国家863计划资助项目(2012AA12A304);国家973计划资助项目(2011CB707106);国家自然科学基金资助项目(41101415;40906092;41071261);国家自然科学基金委创新研究群体科学基金资助项目(41021061);测绘遥感信息工程国家重点实验室专项科研经费;985;国家重点实验室仪器设备专项经费资助项目~~

摘　　要：以具有高时空分辨率、长时间序列观测优势的国产HJ-1CCD数据,在使用面向对象方法实现水面均质区域分割的基础上,改进传统的Contrast box方法,增强了船只目标与背景像素的对比,实现了水面状况复杂的鄱阳湖船只提取。使用该方法分别对鄱阳湖枯水、丰水期HJ1-CCD数据进行船只检测实验,经精度验证,船只检测率可达85%。对比传统阈值提取方法,该方法在浑浊复杂水体中船只提取的优质因数(FoM)远优于传统方法。由于采砂、运砂、航运等人为水上活动对水生态环境产生的严重影响,亟需加强船只活动的有效监管。该研究为鄱阳湖船只检测、采砂、运砂活动及水环境检测提供了有效途径,也为其他湖泊、沿海等区域的船只检测提供了方法,同时提升了国产HJ卫星在环境检测中的应用能力。Taking advantage of high spatial-temporal resolution of HJ-1 CCD remote sensing data, we developed an Enhanced Contrast box method to detect ships for inland or coastal areas, where ship detection is usually impaired by the severe heterogeneities of highly turbid waters. First, the whole area was partitioned into many uniform regions based on object-oriented segmentation algorithm. Then, different Enhanced Contrast boxes were employed on each partition, as defined by various parameter combinations. The proposed method was tested and validated on Poyang Lake, the largest freshwater lake in China. As compared to threshold segmentation, the proposed Enhanced Contrast box method showed higher performance in terms of PD （Percent of Detection） and FoM （Figure of Merit） at a rate of 85 % and 0.82 respectively. This study provides an efficient way to detect ships at large scale and high frequency based on HJ-1 CCD optical remote sensing data from the Poyang Lake as well as other inland lakes or coastal regions.

关 键 词：光学遥感 HJ-1A/B CCD 船只检测 鄱阳湖 

分 类 号：P231.5[天文地球—摄影测量与遥感]