Artificial Intelligence Research Analysis

Target Detection Method Based on Average Absolute Gray Difference-Point Spread Function with Experimental Validation

In maritime search and rescue operations as well as anti-frogman surveillance, sonar image target detection faces significant challenges due to the small scale of targets and the presence of environmental noise. Targets typically occupy only a few pixels in sonar images, resulting in low signal-to-noise ratios (SNR) and making the detection of near-surface small targets from single-frame sonar images particularly difficult. Reducing noise interference, improving detection probability, and lowering false alarm rates remain pressing challenges. To address the limitations of existing small-target detection algorithms, this paper proposes an efficient sonar image target detection method based on diffusion theory. First, a multi-directional gradient composite window combined with an improved average absolute gray difference (AAGD) algorithm is designed to enhance detection capability under complex backgrounds. Then, a novel detection model is developed by fitting the point spread function (PSF) with the hyperbolic secant (sech) function, thereby achieving more accurate and efficient target detection. Finally, validation using sea trial data demonstrates that, compared with five benchmark methods, the proposed approach achieves superior performance in terms of signal-to-background gain and background suppression factor. Moreover, it significantly improves detection probability and reduces false alarm rates, highlighting its strong potential for sonar small-target detection applications.

Enterprise Tags: Sonar Imaging, Target Detection, AAGD, PSF Modeling, Noise Reduction