Spatial–Spectral Image Classification with Edge Preserving Method

Abstract

Classification of remotely sensed imagery with the integration of spatial context should be a constructive way of improved accuracy in image classification. An efficient spatial–spectral classification approach with colorimetric edge preservation of spatial–spectral modeling is processed. This research paper consists of three phases first, and the multispectral imagery [Captured by unmanned aerial vehicles (UAV)] is classified using a per-pixel classification method, i.e., support vector machine (SVM). Then, the classified image is constituted as various probability thematic maps and an edge preservation using chromaticity mapping is conducted on each probability map, with the principal component analysis of the multispectral imagery (Captured by UAVs) using as the color or gray-guidance imagery. Finally, from the edge preservation probability images, the classes of all the pixels are specified with the maximum probabilities. Significantly, the accuracy assessment of image classification improved in a short computational time.