Performance improvement in satellite image classification using adaptive supervised multi-resolution approach

摘要

Multispectral satellite images have a few straps, which have a lower determination and tight data. The purpose is to increase the external and spectral data of satellite images by utilizing high demand intelligence in a mix of force shadow immersion using Multi-resolution 3D modeling classification approach tracked for this work acceptance. The proposed 3D-based Adaptive Supervised Multi-Resolution (ASMR) method that automatically classifies different regions of space-time detection of remote images. Firstly, a 3D central multispectral and multi-volatile remote sensor is designed to compile data structure. Secondly, 3D multi-resolution framework with the acceptance under purified parameters tracking aims at creating 3D region models and learning spatiotemporal attribute representations. Connected parameters are being monitored to evaluate the performance of a 3D compatible multi-resolution project: Root stands for quadratic error, correlation coefficient, structural similarity index measure, and spectral mean relative error. The generated information uses more arranged knowledge to reduce the number of satellite images used to be used by independent band component evolutionary bias. The performance of the proposed method has been validated by simulation using Matlab software. As compared to the traditional 3D layout of the 3D image classification, the proposed solution achieves 97.72% accuracy, 98.25% sensitivity and 94.02% specificity.