最优分辨率配置下典型湿地生境场景溯源方法

摘要

This study aims to enhance the performance of derivation algorithm of the tpyical wetland ecological habitat scene ( WEHS) in China by applying the optimum⁃resolution configuration of spatial envelope ( SE) algorithm. The SE algo⁃rithm was designed as optimal multi resolution in the spatial⁃frequency domain in order to eliminate the effects of redundant features generated by the overlapping frequency bands of nonorthogonal multi⁃scale filters. The normalized algorithm was used to remove the influence of dimensions. Then the nomalized feature variables were centralized in order to eliminate the negative effects of mean drifts. The principal component analysis algorithm was employed to project the preprocessed features to the low⁃dimension subspace to extract the principal components. The large⁃margin nearest neighbor algorithm was used to project the principal components of the WEHS into high⁃dimension feature space and derive the WEHSs in the space. The results showed that the forecast accuracies of modeling and derivation set were boosted by one percentage point and two percentage points compared to the classical algorithm respectively.%为利用最优分辨率配置下的空间网罗特征提升中国典型湿地生态系统生境场景溯源算法的性能，提出了正交配置溯源方法，即采用空频域中最优分辨率配置下空间网罗法，消除非正交多尺度滤波器频带交叠造成的冗余特征输出的影响，使用归一化处理方法去除空间网罗特征量纲影响，使用数据中心化处理方法消除空间网罗特征均值漂移的影响，使用主成分分析算法将高维空间网罗特征数据投影到低维特征子空间，使用大间隔最邻近法算法在高维特征空间中对场景主成分特征进行溯源。研究结果显示正交配置溯源方法对湿地生态系统生境场景的建模集和溯源集预测精度分别比现有非正交配置溯源方法提升了1个百分点和2个百分点。