Optimized Feature Selection for SVM based Crop Classification using Multi-spectral Remote Sensing Images

Abstract

Feature extraction directs the classification process for any sort of data, may be textual database, simple image or complex geo-tiff  images procured using remote sensing technology. In spite of, numerous features that identified for usual textual data types used by researchers across the globe, more complex, but specific features are needed for classifying satellite images with special emphasis upon agricultural crops because of the challenge to discriminate it from customary surrounding vegetation. Identification of such more relevant features pertaining to remote sensing data in order to estimate its crop acreage will be the great for computer professionals and researchers. Once the features are collected, finding the best amongst them, will the second priority task towards classifying crops. Scientist came up with the concept of feature dimension reductionality in order to decide upon the section of best features for problems in hand. Evolutionary algorithms have proven to be the best for such problems, even in remote sensing domains. Further, its application may be extended towards optimization of best features for crop classification based on SVM using multi-spectral remote sensing images.