基于面向对象法艾比湖卤虫信息提取  

Information extraction of the Ebinur Lake artemia based on object-oriented method

作　　者：李微 刘伟男 贾越平 刘洪洋 汤勇 LI Wei;LIU Weinan;JIA Yueping;LIU Hongyang;TANG Yong(Institute of Marine Science and Technology and the Environment, Dalian Ocean University, Dalian 116023, China;The Nearshore Environmental Science and Technology Key Laboratory of Liaoning Province Colleges and Universities, Dalian 116023, China)

机构地区：[1]大连海洋大学海洋科技与环境学院,大连116023 [2]辽宁省高校近岸环境科学与技术重点实验室,大连116023

出　　处：《国土资源遥感》2018年第4期176-181,共6页Remote Sensing for Land & Resources

基　　金：国家自然科学基金项目"单体鱼目标散射强度测量的研究"(编号:10774021);辽宁省自然科学基金项目"基于GOCI数据的北黄海叶绿素浓度反演模型研究"(编号:201602100)共同资助

摘　　要：选择艾比湖为研究区域,以ZY-3多光谱影像为数据源,经正射校正、辐射定标和大气校正等预处理,分析湖区内不同水体覆盖的光谱特征,确定并构建各类水体的解译标志。综合运用光谱、纹理和形状信息等解译标志,采用定性与定量结合的规则,建立基于面向对象法的艾比湖卤虫卵虫条带信息提取模型,并对分类结果进行精度评价。研究表明,分类结果的总体分类精度达到91. 74%、Kappa系数为0. 89;与基于像元法进行对比分析显示,针对卤虫卵虫高、中密度及潜在覆盖区域,面向对象法的分类精度分别达到95. 08%,92. 30%和91. 26%,而基于像元法的分类精度分别为90. 16%,87. 18%和86. 40%; RIartemia指数与形状特征结合可以有效提取不同密度分布的卤虫卵虫信息,相比基于像元法,面向对象法更为有效;由分类结果图可知,面向对象法大大避免了"椒盐"现象,卤虫卵虫区分度高。本研究探索了高空间分辨率图像不同密度卤虫卵虫信息提取的技术方法,可为准确掌握和预测卤虫卵虫的分布与强度提供有效手段,对揭示卤虫的繁殖过程和资源评估具有实用意义。With the Ebinur Lake as the research area and the ZY-3multi-spectral image as the data source,the authors preprocessed the data by such means as ortho-rectification,radiometric calibration and atmospheric correction.The authors analyzed the spectral characteristics of different water bodies,found interpretation signs for artemia information extraction,and built the oriented-object artemia information extraction model by the spectral,texture and shape information.The classification results were validated using the confusion matrix,with the overall classification accuracy being91.74%and Kappa coefficient being0.89.In addition,classification accuracy between object-oriented method and pixel based method was analyzed and compared for the artemia water of different densities.The classification accuracies of object-oriented method for high density,medium density and potential regions were95.08%,92.30%and91.26%,respectively,whereas those of pixel based method were90.16%,87.18%and86.40%,respectively.The results show that the object-oriented method is more effective than the pixel based method.The object-oriented method greatly avoids the phenomenon of“salt and pepper”and can distinguish the artemia densities.The study can provide the effective method for monitoring the distribution and intensity of artemia and has great significance for scientific and reasonable artemia fishing.

关 键 词：面向对象法 卤虫卵虫 多尺度分割 解译标志 

分 类 号：X87[环境科学与工程—环境工程] P96[天文地球—自然地理学]