Satellite Image Classification Using a Lightweight CNN-SVM Hybrid Model

Satellite image classification holds significant importance across various applications, including agriculture, urban planning, and disaster management, due to the necessity for effective data analysis. Conventional CNN models require learning a large number of parameters, which can be time-consuming and often too complex. This study addresses the issues by introducing an effective feature selection technique that combines a lightweight convolutional neural network (CNN) with a Support Vector Machine (SVM) classifier. The aim of this proposed lightweight CNN is to attain maximum accuracy using a smaller number of parameters. The proposed method consists of three distinct sections: pre-processing, feature extraction, and classification. Firstly, images are preprocessed through resizing, histogram equalization, and a sharpening filter to increase the quality of the image. Then the lightweight CNN is used for effective feature extraction, and finally SVM is used for classifying the satellite images. The proposed lightweight CNN-SVM model achieves a high accuracy of 98.0% on the publicly available RSI-CB256 dataset.