An Unsupervised Method for Detecting and Segmenting Shadow Areas of Sunken Targets in Sonar Images

To address the issue of blurred underwater target features and the challenges in target detection caused by significant speckle noise and severe noise interference in sonar images, a novel shadow area detection method for sonar images, termed Sparse Kullback-Leibler Divergence Fuzzy C-Means (SKLFCM), is proposed. Firstly, the method reduces the noise interference in the image by morphological reconstruction and constructs a residual denoising structure to reduce speckle noise and preserve target information simultaneously. Then, the Kullback-Leibler(KL) divergence regularization term is used to constrain the intensity of the current pixel value to be close to the average intensity of its neighboring pixels, ensuring its similarity. Finally, to accelerate the convergence of the objective function, a sparse membership function is introduced to increase the convergence speed while ensuring segmentation accuracy. Through a series of simulation tests and experimental analyses, it has been verified that the proposed algorithm exhibits robust noise reduction and anti-noise capabilities, achieves higher clustering accuracy, and effectively extracts comprehensive target shadow area features even in complex underwater reverberation environments. These findings offer novel technical support for underwater target detection and recognition based on sonar imaging.