提高森林覆盖区积雪提取精度的方法研究——以玛纳斯河流域为例

森林覆盖区积雪的提取精度很低,由于植被冠层的遮挡,冠层下的积雪很难被提取出来。基于Landsat 8OLI数据,针对玛纳斯河流域下游有大面积森林覆盖的特点,通过传统的积雪指数法,结合NDVI数据的积雪指数法和面向对象图像特征法分别提取积雪面积。结果表明:1传统的NDSI和S3积雪指数法无法较好地提取出森林覆盖下的积雪,提取精度分别为85.23%和87.54%。这两种方法适用于空间尺度较大、植被覆盖面积较大的区域,并不适合所选研究区;2结合NDVI数据后的NDSI、S3积雪指数模型能大大提高森林覆盖下的积雪面积,提取精度分别达到91.47%和90.60%。在影像空间分辨率较高,流域尺度较小,林区覆盖较多的情况下可采用此方法提取积雪;3随着海拔的升高,地形阴影影响逐渐增大,NDVI辅助积雪指数方法提取林区覆盖下积雪面积逐渐减小。因此采用光谱、纹理和空间信息结合的面向对象图像特征方法提取积雪,能够较好地识别出受地形影响下的雪像元,精度达到89.75%,可以满足实际应用的需求。

Methods Research to Improve the Extraction Accuracy of Snow under Forest Cover—A Case Study of Manas River Basin

Extraction accuracy of snow-cover under forest is low,due to occlusion of vegetation canopy,the snow under the canopy is difficult to be extracted.In this study,the data is based on Landsat 8 OLI Sensor.Due to a large area of forest distribution in the north central of Manas river basin.This study extracted snow cover information through three methods respectively,including NDSI and S3,NDSI and S3 aid of NDVI data as well as the object-oriented image feature extraction.The results showed that,for the first method,the snow pixels in forest were difficult to identify.Extraction accuracies were 85.23% and 87.54%.For the second method,with the aid of NDVI data,the snow cover can be well extracted by two types of normalized snow index.Meanwhile,the results were quite similar to each other and of high accuracy.Extraction accuracies were 91.47% and 90.60%.So,when the spatial resolution of the image is high,and basin scale is small,and forest covers more cases,this method can be used to extract the snow cover.For the third method,with the increase of altitude,affected by terrain increased gradually,and the vegetation coverage decreased gradually,the number of snow pixels were extracted by NDVI assisted extraction method decreased gradually.So we used the object-oriented image feature extraction method with spectrum,texture and space information to extract snow cover.The method can identify the snow pixels affected by terrain.The extraction accuracy was slightly lower than the second methods mentioned above.Extraction accuracy was 89.75% which met the needs of practical application.