Satellite Image Classification Method Using ELBP and SVM Classifier

For the classification of satellite images into 24 separate classes, machine learning-based Support vector machine (SVM) and Extended Local Binary Patterns (ELBP) algorithms were applied. This work is capable of classifying 24 different classes in addition to satellite images. However, identifying the features of those other classes, such as the human face, football and rugby is also simple because these other classes have some unique features that can be easily distinguished, allowing for easy classification.

The fundamental challenge in the case of satellite photos is that distinct satellite images may have different attributes, making satellite image classification difficult. Another difficulty is that most satellite photos are noise-corrupted. The SVM Classifier is used to estimate the noise patterns in the wireless image, and the estimated noise patterns are then removed using the SVM signal classification technique.

Using the proposed ELBP approach, this research discovers local binary patterns. Because the patterns of distinct satellite photos and other class images cannot be distinguished using simply LBP, the Extended LBP is required. SVM identifies the test image's class based on the expanded features gathered. The ELBP-SVM approach was applied in this study, and the satellite picture correct recognition rate was 95 percent. The results produced using MATLAB 2020a are superior than those obtained using other methods to classify satellite photos.

Keywords: Local binary patterns, Support vector machine, Satellite imagery, Machine learning, Feature extraction.