Hyper Spectral Satellite Image Classification Based Land And Water Bodies Area Detection Using Machine Learning And Opencv Techniques

Remote sensing is one of the important applications for monitoring the ecological conditions in the world. Over the years of climate change and urbanization produced a greater impact on land utilization as compared to its original form. With the help of satellite images like LIDAR, Senitel and Hyperspectral images, the remote sensing process were performed. Among these images, the Hyper-Spectral image have a greater advantage and impact in analyzing the land utilization. Because this HS images has both spectral and spatial properties of the region and it helps to improve the land classification. Several research works also stated that both spatial and spectral properties of HS image is required for analysis and these features were collected using multi-models or individual network. Hence, in this, the land classification using HS images were performed using Attention based modified VGN network. To perform this, the images undergo two main processes namely channel transformation from RGB to LAB and LBP image formation for an efficient feature extraction and classification in A-M-VGN. This proposed method was tested on multi-class like agricultural, barren, green and water bodies region in HS images and evaluated using Accuracy, computation time, Kurtosis, Standard Deviation and mean. The Kurtosis, mean and SD helps to analyze the pixel variations in the image. The results showed that the proposed approach is efficient for land classification using HS images.