Ship Classification Based On Co-Occurrence Matrix and Support Vector Machines

Abstract Synthetic aperture radar (SAR) is an important and efficient imaging technology. This system provides robust information for various applications such as ship detection, climate change, and agricultural land modeling. Ship detection and classification problem is an important object detection problem that involves difficulties. There are deep-learning-based studies to solve this problem. However, mathematical and statistical methods should be developed for ship classification applications. In this paper, gray-level co-occurrence matrix-based method is proposed. The gradient of the input SAR image was calculated using Gaussian derivative filters. The gradient magnitude was calculated with horizontal and vertical gradient information. Gray-level co-occurrence matrix was obtained using gradient magnitude. The meaningful features of the images were calculated by performing 4 different statistical calculations. Results on our SAR database reveal the proposed model's superior classification performance. Cite this article as: K. Hanbay and T. B. Özdemir, "Ship classification based on co-occurrence matrix and support vector machines," Electrica, 24(3), 812-817, 2024.