基于空间约束关系的土地利用/覆被遥感分类方法研究

Land-use/land-cover classification method under spatial constraint

该文提出了基于空间约束关系的遥感分类方法,即利用数字高程模型、土地利用数据库等提供的先验知识,构建地物空间约束关系,实现先验信息和光谱信息的融合,以消除或部分消除单纯使用光谱知识造成的“同物异谱”和“异物同谱”,抑制因空间分辨率提高造成类内噪声较大等现象。试验表明,该方法简单实用,同时也说明了空间知识在遥感分类中的重要性。

A classification method under spatial constraint was proposed. Based on the prior knowledge such as digital elevation model and land use database, spatial constraint relationships were constructed; prior knowledge and spectral information were syncretized. The results showed that the method was resistant to the problem of having apparent same spectrum but different ground covers and was capable of removing salt and pepper noise. The result also proves the importance of spatial information in remote sensing image classification and the effectiveness of classification method under spatial constraint.