Automatic Target Detection and Analyses in Side-scan Sonar Imagery

The purpose of the current investigation was dedicated to the development of an automatic side-scan sonar imagery analysis program for the detection and identification of stationary targets, such as meter-sized concrete artificial reefs, on the sea floor. The major components of the program include: image acquisition; feature extraction; feature classification; target identification; and target properties analyses. Program verification and optimal parameters determination were conducted with a high quality sonograph of 650 times 650 pixels. Optimal parameters, i.e., a region size of 38 x 38 pixels, a region sliding distance of 4 pixels and 16 quantized grey levels for this specific type of image were determined. A realistic evaluation of this program was conducted and proved successful. The automatic image analysis program based on grey level co-occurrence matrix, unsupervised Bayesian classifier and hierarchical cluster analysis can improve the detection and identification of stationary targets in side-scan sonar imagery. It offered a potential for the development of a completely automatic system for the detection of stationary target in real time at sea.