基于SuBSENSE的内河船舶检测波纹干扰抑制算法

摘要

SuBSENSE is a universal detection algorithm for moving objects which combines color and texture features.The background model can adapt to a variety of inland environments and achieve parameter optimization through its adaptive feedback mechanism.However, the ripples cannot be removed when directly detecting inland ships by SubSENSE.Aiming at solving this problem, a novel algorithm combining SuBSENSE and a method for detecting significant regions based on global contrast is proposed.There is a fact that the saliency values of ships and ripples are typically different, ships and ripples are separated in binary saliency image.Logical bitwise AND is performed between the binary saliency image and the SuBSENSE detection image to get final results.The method has shown excellent results in simulations, with a 14.6% margin over the original SuBSENSE performance.%SuBSENSE是一种融合颜色特征和纹理特征的通用运动目标检测算法,同时算法中的参数自适应反馈机制使得背景模型能够良好地适应内河环境的多样性,在多种检测环境下达到参数最优化设置.针对一般运动目标检测算法用于内河船舶检测时,难以克服水波纹干扰这一问题,提出将SuBSENSE与基于全局对比度的显著性区域检测方法结合进行波纹抑制.利用水面显著值较低这一特性,通过设置适当阈值对显著图进行二值化,从而分离船舶与水面区域.将显著图与SuBSENSE检测结果进行与运算滤除背景干扰,即可得到船舶区域.实验证明,该方法能有效抑制内河环境中的波纹干扰,相比原SuBSENSE算法将综合表现提高了14.6%.