Image Classification Using SVM Algorithm with Feature Extraction Techniques

In this work, we employ Support Vector Machine (SVM) algorithm for image classification tasks. Specifically, we developed an automated classification system to differentiate between basketball courts and tennis courts. The core implementation utilizes color moment-based feature extraction algorithm for processing color images, which calculates statistical moments (mean, standard deviation, and skewness) from color channels to capture color distribution characteristics. To achieve more comprehensive and accurate classification results, we integrated additional feature extraction methods including Histogram of Oriented Gradients (HOG) algorithm for shape and edge information capture, and Local Binary Pattern (LBP) algorithm for texture feature extraction. The HOG implementation involves computing gradient orientations in localized image patches, while LBP encodes local texture patterns through neighborhood pixel comparisons. Prior to feature extraction, we implemented preprocessing stages comprising noise reduction techniques (using Gaussian filtering) and image enhancement methods (such as contrast adjustment) to ensure consistent input quality and reliability. The complete pipeline processes images through preprocessing, multi-algorithm feature extraction, and SVM classification with parameter optimization using techniques like cross-validation. Our experimental results demonstrate that the implemented algorithm achieves robust classification performance across multiple test datasets, validating its effectiveness and practical applicability in image classification scenarios. The SVM model was trained using kernel functions (particularly RBF kernel) to handle non-linear feature separability, with feature vectors normalized to ensure scale invariance.