DGAPNet-SIE: A New Enhanced Hybrid Model for Classification of Remote Sensing Satellite Images Integrating Image Enhancement

Remote sensing pictures are widely employed in different industries, including environmental monitoring, agriculture, urban planning, and disaster management. Image classification is one of the prominent tasks influencing the rapid advancement in remote sensing applications. However, literature concerning well established large-scale or application-dependent datasets and models still requires improvement. Satellite image classification has many associated challenges. These challenges include the quality of data, the quantity of data, and data distribution. These challenges make the analysis of satellite images more challenging, in addition to that Handling scale and resolution discrepancies presents issues in image analysis. Moreover, the size of remote sensing datasets has increased dramatically as a result of the expansion of satellite missions and the development of sensor technology. This paper aims to introduce high-capacity convolutional neural network architectures and their variations with large-scale datasets. Various convolutional neural network architectures with scaling method are proposed for the classification of satellite images. The proposed methodology uses spatial information from satellite images to categorize into four categories. A new modified model for classification of remote sensing satellite images, Deep Global Average Pooling-based network with image enhancement (DGAPN et-SIE) was proposed to improve the spatial resolution of remote sensing images. First, DGAPN et-SIE was proposed to learn the spatial features of images. Satellite Image Enhancement plays an integrated role to improve the image resolution, used BM3D which is image denoising technique to reduce noise and enhance the image resolution. And used various methods like MobileNetV2, VGG19, ResNet50, and Xception. With new integrated dense and dropout layers, which are designed in the classification phase to preserve more gradient details in the remote sensing images. The results show that Xception model has reported the best accuracy 98.82%.