An Efficient Feature Selection and Extraction Technique for Agriculture Crop Classification using Hypersectral Data

Abstract

In recent years, various effective research have been considered regarding the agricultural based application; crop
identification through HSI (Hyperspectral imagery information) collected from satellites is one of the rising
mechanism. Moreover huge research has been carried out for crop recognition to design the efficient model. In
general HSI comprises plenty of narrow bands (NB) with high spectral relation and it is continuous in nature;
moreover this induces the space and time complexity and further results in huge computation overhead in processing
these kind of data. Dimension reduction mechanism such as feature extraction and band selection plays eminent role
in improvising the HSI. However existing mechanism of dimension reduction does not justify the performance metric
for crop classification, since it is mostly affected due to the noise presence including the climate condition. Hence to
address the above mentioned problem, proposed dimension reduction technique aims at identifying the semantic and
useful features through given HSIusing fusion method. Further feature of dimension reduction is trained with SVM
(Support Vector machine) and experiments are performed on the standard dataset of HSI. Moreover the outcome
shows that proposed model achieves better classification performance in comparison with the existing HSI based
crop classification model.