Hybrid Wavelet–GLCM Feature Fusion Framework for Automatic Underwater Object Detection in Side-Scan Sonar Images

Automatic detection of underwater objects in side-scan sonar imagery is essential for maritime surveillance, seabed mapping, shipwreck localization, and mine-like target
identification. However, reliable object detection remains
challenging due to multiplicative speckle noise, acoustic shadow distortion, heterogeneous seabed textures, and limited contrast variations in sonar images. Although deep learning–based detection frameworks have demonstrated improved performance, they often require large annotated datasets and high computational resources. To address these challenges, this paper proposes a hybrid Wavelet–GLCM feature fusion framework for automatic underwater object detection in sidescan sonar images. The proposed approach integrates medianfilter-based speckle noise suppression, multi-resolution discrete
wavelet decomposition, and gray-level co-occurrence matrix
(GLCM) texture feature extraction to enhance highlight–
shadow separability and seabed texture discrimination. A
feature fusion strategy is introduced to combine multi-scale
edge-shadow representations with statistical texture descriptors
for improved localization reliability. The fused features are used
for threshold-based object detection followed by bounding-box
localization of candidate targets. Experimental evaluation
performed on a subset of the SWDD side-scan sonar dataset
demonstrates improved detection accuracy compared with
conventional handcrafted feature-based approaches.
Quantitative results show that the proposed framework achieves
higher precision, recall, and F1-score values while maintaining
computational efficiency. The proposed method provides an
effective and lightweight solution for automatic target
recognition in complex underwater acoustic imaging
environments.