A novel approach to segment and classify remote sensing pictures more accurately than random forest by employing FFNN

Abstract

Aim: Image classification includes analyzing and categorizing specific objects over the pictures. Our main purpose is to assess the performance of two types of classifiers used in image segmentation. The image of every location is randomly selected and changed into the bitmap formation to create the test type dataset consisting of 5 locations from the 400 different satellite images. Total 10 number of iterations computed with400 images which were collected from 5 various locations. The images were collected at different views. Materials and Methods: The classifiers are making use to segment the remote sensor images using the FFNN framework and Random Forest. In this research work, the ORL(Olivetti Research Laboratory) kind image database is used for analysis, and the implementation is processed with the assistance of Python programming. Results and Discussion:  The outcome of novel FFNN compared with Random Forest, which has an accuracy of 97.15%, is compared with the previous classifier accuracy of 93.75%. The two algorithms are statistically satisfied with the independent sample t-Test two tailed is 0.008 (