A non-parametric approach for accurate contextual classification of LIDAR and imagery data fusion

摘要: Light Detection and Ranging (LIDAR) has become a very important tool to many environmental applications. This work proposes use LIDAR image data fusion develop high-resolution thematic maps. A novel methodology is presented which starts building matrix of statistics from spectral spatial information by feature extraction on the available bands (RGB images, intensity height LIDAR). Then, contextual classification applied generate final map using support vector machine (SVM) classify every cell nearest neighbor (NN) rule sequentially reclassify each cell. The results obtained this method, called SVMNNS (SVM NN Stacking), are compared with non-contextual SVMs. It shown that obtains best when real Iberian peninsula.