【摘要】 海岸带作为海洋、陆地和大气共同作用的地带,其地物混杂度大,变化频繁,单纯利用光谱特征分类难以取得理想的精度。但海岸带地物滨临水体,而水体与地物存在巨大光谱差异,易于识别。据此,本文尝试提出一种简单的加入空间关系的遥感图像分类方法,即,先准确识别出海水类,然后统计非海水类的每一个像元到最近海水的空间距离。由于不同的海岸带地类距离海水的空间距离有其自身的特点,因此,利用这一距离信息辅助分类能提高分类精度,尤其是针对那些光谱特征相近而距离相差较大的地类。更多还原

【Abstract】 Influenced jointly by such factors as ocean,land and atmosphere,the coastal zone is characterized by the mixing of various land types with high extent of variation.Therefore,traditional spectrum-based classification cannot meet the demand of precision.The coastal zone is close to ocean water which can be recognized easily by its significant spectral difference from land objects in the remotely sensed images.Taking advantage of this feature,this paper proposes a new classification method for coastal zone remotely sensed images by adding space information.Ocean water is recognized first,and then the distance from every non-ocean-water pixel to its nearest sea water is calculated.Different objects on the coastal zone have their respective characteristic distances to ocean water.So helped with the space information,we can improve the precision of classification,especially for the objects with similar spectral features but different distances to sea water.Using this method,the authors studied the multi-band QuickBird image of Huangdao island in Qingdao,and the result proves the validity of this method in coastal areas relative to the pure spectrum method.更多还原