基于面向对象的濒海湿地脆弱区土地利用/覆被信息精细提取

Precise Extraction of Land Use/Cover Information in the Fragile Coastal Wetland Based on Object-oriented Method

土地利用/覆被信息的精细提取,对于维护濒海湿地脆弱区的生态稳定和保障当地经济可持续发展具有重要意义。本文以黄河口为研究区,以Landsat 8影像为数据源,利用面向对象分析技术进行土地利用/覆被类型信息的精细提取研究。在建立土地利用/覆被详细分类体系的基础上,采用基于样本的面向对象分类方法对研究区土地利用/覆被类型进行初步提取,分类结果显示该分类模型只能较准确地提取除藕池、稻田和苜蓿地之外的其他地类。针对该问题,论文采用基于规则和样本相结合的面向对象分类方法进行土地利用/覆被类型的精细提取,通过设置面积和形状属性规则从初步分类结果中进一步精确提取藕池和稻田,同时也改善了其他地类的精度,得到总精度94.93%、Kappa系数0.94的较为理想的分类精度,比最大似然法的总体分类精度提高了38.56%。研究结果表明,基于样本与规则相结合的面向对象分类方法能实现濒海湿地脆弱区土地利用/覆被信息的精细提取。

Precise extraction of the land use/cover information is of great significance for the ecological stability and the economic sustainable development in the fragile coastal wetland region. This paper took the Yellow River Entrance as the study area to extract the land use/cover information accurately by Object-oriented technology based on the Landsat 8 image.Based on the establishment of the detailed land use/cover classification system, we got the preliminary land use/coverage classification results with the object-oriented method which was based on sample. However, the classification results showed this classification model could accurately extract all land types except paddy fields, lotus root pools and Medicago. To solve this problem, combining the rule-based and example-based object-oriented method was used to complete high accuracy extraction of land use/coverage information. Results indicated that the paddy fields and lotus root pools got more accurate by setting the area and rectangular attributes rules, at the same time, other land types precision also have been improved, this classification model produced an overall accuracy of 94.93% and a Kappa value of 0.94 which was 38.56% more than the maximum likelihood method. Therefore, the research demonstrated that this method combining the rule-based and example-based object-oriented methods could achieve high accuracy classification of the land use/coverage information in the wetland.