加权融合特征耦合Top-Hat变换的红外目标检测算法

Infrared target detection algorithm based on Top-Hat transform and multi-feature weighted fusion feature

当前红外目标检测算法在复杂背景干扰下通过单帧技术来实现目标的检测,使其难以过滤固定噪声等虚警,为此提出一种基于多尺度Top-Hat选择变换与加权融合特征的红外目标检测算法。通过对经典Top-Hat变换的结构元素进行分割,形成多尺度结构元素,考虑真实目标与背景的灰度差异,构建多尺度Top-Hat选择变换,从不同尺度提取图像的感兴趣信息,将真实目标从背景中分离出来;考虑序列图像的帧间差异,提取红外目标的运动特征,联合灰度特征与梯度特征,建立加权融合特征模型,精确定位候选目标区域;引入二维Ostu分割机制,对提取的候选目标区域进行分割,完成弱小目标检测。实验结果表明,与当前红外目标检测技术相比,在复杂背景下,所提算法可更好地抑制背景与过滤虚警,具有更高的检测精度与效率。

To solve the location ambiguity defects under the complex background interference in current infrared target detection and tracking algorithm,the infrared target detection based on Top-Hat transform and multi-feature weighted fusion was proposed.The sequence detection mechanism was constructed using the gray differences between the real target and background and the continuity of the dim moving target for separating the target from the background.The motion features of infrared target were extracted by considering the difference between the image frames,and the multi-feature weighting function was defined by combining with the characteristics of gray scale and spatial domain.The weighted correlation filter was designed to accurately locate the real target.The small target was detected based on 2D Ostu segmentation mechanism.Experimental results show that the proposed algorithm has higher detection accuracy in the complex background compared with the current infrared target detection technology.