Neural Network Based Terrain Classification Using Remote Sensing
Abstract
The grouping of the image pixel values within the substantial groups is involved by the method of Satellite image classification. The various procedures as well as approaches of satellite image classification are present. Automatic, manual as well as hybrid are the 3 kinds of the Satellite image classification techniques and the individual benefits as well as drawbacks are possessed by them. Most of the techniques of satellite image classification are classified within the 1st group. Selecting the suitable classification technique depending upon the necessities are required by the classification of the satellite image. Studying the classification techniques as well as procedures of satellite image is the main aim of this research. The relative outcomes of the numerous investigators upon the classification techniques of satellite image is compared in this work. A plain graphical indicator which is implemented for analysing the measurements of remote sensing frequently out of a space platform that evaluates the target that is detected with live green vegetation is referred as normalized difference vegetation index (NDVI). This paper proposes a new method of NDVI image segmentation using SLICE superpixel segmentation and classification using 1D CNN. The SLICE Algorithm segments the NDVI image (input image) into many small regions based on colour and texture these smaller segments are then classified using 1D CNN in to different regions namely, water, agriculture, hills, and forest area. The proposed method has higher performance when compare to existing algorithms.
Keywords: NDVI (normalized difference vegetation index), superpixel segmentation, classification, 1D CNN.
