基于遥感影像的雪线高程提取方法对比分析——以天山北坡为例

利用2010-2014年6~8月MOD10A1、MOD10A2及HJ-1/CCD数据,以天山北坡为研究区,对比分析了积雪持续时间比率法及最大值合成法在雪线高程提取中的精度和适用性。结果表明:当MOD10A1以旬合成,积雪持续时间比率阈值为40%时,提取的雪线高程与"真值"最为接近;基于混淆矩阵法,以目视解译得到的HJ-1/CCD积雪面积为"真值",分析得出,积雪持续时间比率法提取积雪面积的总体精度较高,为98.59%,khat系数为0.90,最大值合成法的总体精度仅为85.29%,khat系数为0.70;基于统计学的方法,计算得出,积雪持续时间比率法提取的雪线高程与"真值"的相关系数较高,为0.77,其平均误差、正、负误差、绝对平均误差以及均方根误差相对较低;最大值合成法提取的雪线高程与"真值"的相关系数较低,仅为0.54,其平均误差、正、负误差、绝对平均误差以及均方根误差相对较高;积雪持续时间比率法提取的雪线高程结果优于最大值合成法。

Comparison of snow line elevation extraction method based on remote sensing image:a case of north slope of Tianshan Mountains

On the basis of the MOD10A1, MOD10A2 and HJ-1/CCD data from June to August in 2010 to 2014, with the north slope of Tianshan mountains as the study area, this paper compares and analyzes the snow duration ratio method and the maximum synthesis method in the snowline elevation extraction accuracy and applicability. The results show that snow line height' s error is minimum while snow duration ratio method taking the 40% threshold extraction. Based on confusion matrix method, snow area of HJ-1/CCD is "true value" by the visual interpretation. The results demonstrate that snow duration ratio method to extract snow area's overall accuracy is higher, the accuracy is 98.59% Khat coefficient is 0.90,the accuracy of the maximum synthesis is only 85.29%, Khat coefficient is 0.70. Based on the statistical method, the paper calculated that snow duration ratio method to extract the snowline elevation has higher correlation coefficient of 0.77 with "true value". The average error, error of positive and negative, average absolute error and root mean square error is relatively lower. The maximum synthesis method extracting line height and the same area "true value" has lower correlation coefficient of only 0.54. The average error, error of positive and negative, average absolute error and root mean square error is relatively higher. Using snow duration ratio method to extract the snowline elevation results is better than that of the maximum synthesis method.