上海市湿地信息遥感提取方法研究

Extracting Method of Remote Sensing on Information of Shanghai Wetlands

上海市地处长江三角洲东部,是中国重要的海滨城市,有丰富的湿地资源。利用TM遥感影像数据,结合非遥感数据和资料,在对各类地物光谱特征进行分析的基础上,提出分区分类的方法,辅助地物形状、斑块面积和纹理特征等,对上海市湿地资源信息进行遥感提取,并评价了分类精度。分区分类方法的总体分类精度达到了84.5%,kappa系数为0.815。与传统的分类方法相比,分区分类方法在一定程度上提高了分类精度,实现了上海市湿地信息的遥感提取。其中,由于利用了斑块形状指数等特征,水体的分类精度较高;然而,由于光谱混淆等原因,植被提取的分类精度较低。研究结果表明,上海市湿地面积为1 695.81 km2,其中,水田面积最大,占湿地总面积的72%,滨海湿地面积、库塘面积、河流面积和湖泊面积分别占湿地总面积的10%、9%、5%和4%。

Shanghai, located in the eastern part of the Yangtze River Delta, is an important coastal city, with rich in wetland resources. Urban development is closely related to the rational use of wetland resources and scientific management. However, the wetland resources have been badly damaged by degradation and pollu- tion. With the development of wetland research, the traditional modes on wetland investigation cannot satisfy the needs of wetland information. Remote sensing technology has gradually become an important date source of information for wetland research. Wetland is a complex ecologist system and the spectrum of remote sens- ing image on study area is one of the mixed various component values. Although many scholars made great ef- forts, there still exists a considerable problem on the accuracy of remote sensing classification. This paper us- ing TM remote sensing image data, combined with auxiliary data material, on the foundation of analyzing all kinds of ground object spectral characteristics, extracted the information of Shanghai wetlands by the method of sub-area classification. At first, the region of study area was divided into 3 types of wetlands： water, vegeta- tion and side beach. Then, do secondary classification in the region of each type of the wetlands. The main purpose of secondary classification was to do some difficult classification categories. Considering the limita- tions of spectral radiometer distinct, the author took full advantage of the characteristics of the surface of shape, texture and other auxiliary interpretation. The subarea classification is a step-by-step process of classifi- cation. Sorting features from easy to difficult, this classification can effectively weaken confused pixel in each region of interest and improve the classification accuracy. By using the method to extract the information of Shanghai wetlands, the desired results could be obtained. Statistical results showed that the total area of Shang- hai wetlands was about 1 695.81 km2. The rice cultivated area was the largest, accounted for 72% of the total area of the wetlands, and the next the area of coastal wetlands, ponds. The areas of rivers and lakes were the smallest. The result indicated that the overall accuracy of subarea classification method reached 84.5%. Com- pared with the traditional classification method, in a certain extent, subarea techniques improve the classifica- tion accuracy. Because of the use of the patch shape information and other characteristics, water had higher classification accuracy. However, due to the situation of spectral confusion, the classification precision of veg- etation was low.