基于TM影像的水体信息提取方法研究(英文)

Study on information extraction method of water body based on TM imagery

以高邮湖为研究区,TM(Thematic Mapper)影像为数据源,结合常规调查数据,分别运用单波段阈值法、谱间关系法、比值法、归一化差异水体指数(NDWI)法、改进的归一化差值水体指数(MNDWI)法和混合水体指数(CIWI)法进行水体信息提取研究。并通过实验比较6种方法的优缺点,确定出最佳水体提取方法。结果表明,谱间关系法易于消除由于建筑物阴影所造成的水体误提;CIWI易于消除建筑用地所造成的水体误提;基于谱间关系法和CIWI模型,可取得较高的水体提取精度,有效地减少水体误提现象,提高水体识别的精度。研究表明,基于谱间关系法和CIWI模型相结合所提取的水体信息精度可达到95.2%。

Taken Gaoyou Lake as the study area, and Thematic Mapper （TM） imagery as data sources, field investiga- tion data, single - band threshold method, spectral relationship method, and water indices of simple ratio, Normalized Difference Water Index （NDWI） , Modified Normalized Difference Water Index （MNDWI） and Combined Index of ND- VI and MIR for Water Body Identification （CIWI） were applied respectively to extract information of water body. The results of information extraction by use of the six methods were compared to determine the optimal method for water body extraction. The results showed that spectral relationship method was easy to eliminate classification errors caused by shadows from the built-up while CIWI water index was easy to eliminate classification errors caused by built-up. There- fore, the combination of spectral relationship method and CIWI water index could achieve a higher accuracy of water body extraction effectively. The combination method presented in this study could reduce the misclassification and in- crease the accuracy of extraction of water body. The overall extraction accuracy of water bodies was 95.2 % and the commission rate was 4 %, which suggested that the combination method was a simple and effective method with relative- ly high precision.