国土资源遥感
國土資源遙感
국토자원요감
REMOTE SENSING FOR LAND & RESOURCES
2014年
2期
99-104
,共6页
盖颖颖%周斌%孙元芳%周燕
蓋穎穎%週斌%孫元芳%週燕
개영영%주빈%손원방%주연
溢油%HJ-1%墨西哥湾%纹理特征%决策树
溢油%HJ-1%墨西哥灣%紋理特徵%決策樹
일유%HJ-1%묵서가만%문리특정%결책수
oil spill%HJ-1%Gulf of Mexico%texture characteristic%decision tree
快速、准确地获取溢油污染信息,对海洋的动态监测、保护和可持续利用具有重要意义。环境与灾害监测预报小卫星星座一号(HJ-1)是我国针对生态环境污染和灾害监测发射的新型卫星平台,但HJ-1 CCD多光谱数据的光谱波段较少,仅依赖光谱信息获取海面溢油范围的精度较低。因此,以墨西哥湾溢油事件为研究对象,在分析不同地物光谱特征的基础上,采用灰度共生矩阵,选择合适的纹理结构因子,提取HJ-1 CCD图像中影响溢油识别的地物纹理特征;建立光谱特征和纹理特征相结合的决策树模型,提取海面溢油信息,并与只考虑光谱信息的传统分类方法进行精度对比。结果表明,与最大似然分类法相比,决策树方法的油膜提取用户精度和制图精度分别提高了11.85%和4.28%。
快速、準確地穫取溢油汙染信息,對海洋的動態鑑測、保護和可持續利用具有重要意義。環境與災害鑑測預報小衛星星座一號(HJ-1)是我國針對生態環境汙染和災害鑑測髮射的新型衛星平檯,但HJ-1 CCD多光譜數據的光譜波段較少,僅依賴光譜信息穫取海麵溢油範圍的精度較低。因此,以墨西哥灣溢油事件為研究對象,在分析不同地物光譜特徵的基礎上,採用灰度共生矩陣,選擇閤適的紋理結構因子,提取HJ-1 CCD圖像中影響溢油識彆的地物紋理特徵;建立光譜特徵和紋理特徵相結閤的決策樹模型,提取海麵溢油信息,併與隻攷慮光譜信息的傳統分類方法進行精度對比。結果錶明,與最大似然分類法相比,決策樹方法的油膜提取用戶精度和製圖精度分彆提高瞭11.85%和4.28%。
쾌속、준학지획취일유오염신식,대해양적동태감측、보호화가지속이용구유중요의의。배경여재해감측예보소위성성좌일호(HJ-1)시아국침대생태배경오염화재해감측발사적신형위성평태,단HJ-1 CCD다광보수거적광보파단교소,부의뢰광보신식획취해면일유범위적정도교저。인차,이묵서가만일유사건위연구대상,재분석불동지물광보특정적기출상,채용회도공생구진,선택합괄적문리결구인자,제취HJ-1 CCD도상중영향일유식별적지물문리특정;건립광보특정화문리특정상결합적결책수모형,제취해면일유신식,병여지고필광보신식적전통분류방법진행정도대비。결과표명,여최대사연분류법상비,결책수방법적유막제취용호정도화제도정도분별제고료11.85%화4.28%。
Rapid and accurate access to the oil spill information is of great significance for dynamic monitoring, conservation and sustainable use of the oceans. HJ - 1 is a new satellite platform designed for ecological environmental pollutions and disasters. However, the multispectral image obtained from HJ-CCD has insufficient spectral bands, and the accuracy of acquiring the oil spill coverage only by spectral information is low. In this paper, the oil spill that occurred in the Gulf of Mexico was selected as the research object. Based on the spectral analysis of different features, the authors chose the right texture structure factors and extracted the texture characteristics which affect oil spill identification by gray co-occurrence matrix. A decision tree model combining spectral characteristics with texture characteristics was established to extract the oil spill on the sea surface. A comparative analysis by using the result of maximum likelihood supervision classification method was performed, and the results show that, in comparison with the maximum likelihood classification method, the decision tree method could improve the user's accuracy and the producer's accuracy of oil spill extraction by 11. 85% and 4. 28%respectively.