HJ-1光学遥感影像船只检测——以鄱阳湖为例
Detection of Ships in Homogeneous from HJ-1 Optical Remote Sensing Images: A Case Study at the Poyang Lake
-
摘要: 以具有高时空分辨率、长时间序列观测优势的国产HJ1CCD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Contrastbox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HJ1CCD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。Abstract: Takingadvantageofhighspatial temporalresolutionofHJ 1CCDremotesensingdata,wedevelopedanEnhancedContrastboxmethodtodetectshipsforinlandorcoastalareas,whereshipde tectionisusuallyimpairedbythesevereheterogeneitiesofhighlyturbidwaters.First,thewholeareawaspartitionedintomanyuniformregionsbasedonobject orientedsegmentationalgorithm.Then,differentEnhancedContrastboxeswereemployedoneachpartition,asdefinedbyvariousparametercombinations.TheproposedmethodwastestedandvalidatedonPoyangLake,thelargestfreshwaterlakeinChina.Ascomparedtothresholdsegmentation,theproposedEnhancedContrastboxmethodshowedhigherperformanceintermsofPD(PercentofDetection)andFoM(FigureofMerit)atarateof85%and0.82respectively.ThisstudyprovidesanefficientwaytodetectshipsatlargescaleandhighfrequencybasedonHJ 1CCDopticalremotesensingdatafromthePoyangLakeaswellasotherinlandlakesorcoastalregions.
下载:
