复杂背景下双邻域局部权重对比度红外小目标检测算法

Dual-Neighborhood Local Weighted Contrast Algorithm for Infrared Small Target Detection within Complex Backgrounds

摘要:

为降低复杂背景中的干扰杂波对红外小目标检测的影响,提出一种双邻域局部权重对比度算法。首先,考虑到不同尺寸的红外小目标的背景特性,采用双邻域窗口策略有效捕捉目标与背景特性。其次,分别计算方向信息特征图和权重系数增强图,前者充分利用目标的弥散方向信息,后者结合区域灰度响应强度及离散程度生成权重信息,两者结合生成目标显著图。最后,采用自适应阈值分割从目标显著图中提取目标。在4组不同背景的公开数据集上与6种算法进行了比较,所提出的算法具备较强的抗干扰能力和准确的检测性能。

Abstract:

To mitigate the impact of interference clutter in the detection of infrared small targets within complex backgrounds, a dual-neighborhood local weighted contrast algorithm is proposed. First, considering the background characteristics of small targets of different sizes, a dual-neighborhood window strategy is employed to effectively capture target and background features. Subsequently, directional information feature and weight coefficient enhancement maps are computed separately. The former fully utilizes the dispersal direction information of the target, while the latter generates weight information by utilizing the intensity and dispersion of grayscale responses in different regions. The combination of these two maps through image fusion results in a target saliency map. Finally, adaptive threshold segmentation is applied to extract targets from the saliency map. Comparative evaluations were conducted on four publicly available datasets with different backgrounds, involving six different algorithms. The proposed algorithm demonstrates robust anti-interference capabilities and accurate detection performance.