LAUSR

Infrared small maritime target detection under strong ocean waves, a challenging task, plays a key role in maritime distress target search and rescue applications. Many methods based on directionality or gradient properties have proven to perform well for infrared images with heterogeneous scenarios. However, they tend to perform poorly when facing strong ocean wave background, mainly due to the following: 1) infrared images have low signal-to-clutter ratio with low intensity for small targets; 2) some waves have high local contrast that may be similar to or higher than targets. To solve these issues, a new method based on gradient vector field characterization (GVFC) of infrared images is proposed. First, we construct the gradient vector field and coarsely extract suspected targets. Then, gradient vector distribution measure (GVDM) is presented, which comprehensively integrates a synergistic homogeneity test based on Kolmogorov–Smirnov test with absolute difference standard deviation for gradient direction angle and regression analysis for gradient modulus. The proposed GVDM takes advantage of pixel-level gradient distribution property to further filtrate refined suspected targets. Moreover, gradient modulus horizontal local dissimilarity is proposed to measure the diversity of gradient modulus in horizontal direction between targets and waves, so as to enhance target saliency and suppress residual clutter simultaneously, which achieves preferable performance. Finally, a simple adaptive threshold is applied to confirm targets. Extensive experiments implemented on infrared maritime images with strong ocean waves demonstrate that the proposed method is superior to the state-of-the-art methods with respect to robustness and detection accuracy.