基于局部显著度的机动弱小目标检测算法

Maneuvering dim small target detection algorithm based on local saliency

摘要:

由于待检测的红外远距离目标具有尺寸小、辐射低、背景复杂等检测难点,以高检测率、低虚警率、高实时性进行红外小目标检测一直是一个具有挑战性的课题；文章提出一种基于局部显著性的变速运动目标累积检测算法,利用辐射能量积累方法提高目标在背景中的信噪比；首先,建立矢量空间及一阶导数空间,对序列图像的每帧进行基于块显著度的局部对比度增强处理,增强目标辐射能量并抑制背景,同时显著减少了计算量；然后,进行变速运动空间及导数矢量空间的辐射能量叠加,在空间矢量及导数矢量空间中检测序列图像中的目标在矢量空间及导数矢量空间运动特征的存在概率；最后,通过恒虚警检测得到目标的位置向量、速度、加速度向量,完成目标检测；实验结果验证了提出方法的有效性,其检测率及虚警率均优于其他方法,其中信杂比增益提高了22.30,背景抑制因子提高了3775.68,处理时间开销降低了0.65秒；

Abstract:

Because the infrared long-range target to be detected has detection difficulties such as small size, low radiation and complex background, infrared small target detection with high detection rate, low false alarm rate and high real-time performance has always been a challenging subject. In this paper, a variable speed moving target cumulative detection algorithm based on local saliency is proposed, which uses the radiant energy accumulation method to improve the signal-to-noise ratio of the target in the background. Firstly, the vector space and first derivative space are established to enhance the local contrast of each frame of the sequence image based on block saliency, enhance the target radiation energy and suppress the background, and significantly reduce the amount of calculation. Then, the radiation energy of variable speed motion space and derivative vector space is superimposed, and the existence probability of the motion features of the target in the sequence image in the vector space and derivative vector space is detected in the space vector and derivative vector space. Finally, through CFAR detection, the position vector, velocity and acceleration vector of the target are obtained to complete the target detection. The experimental results verify the effectiveness of the proposed method, and its detection rate and false alarm rate are better than other methods. The signal to clutter ratio gain is increased by 22.30, the background suppression factor is increased by 3775.68, and the processing time overhead is reduced by 0.65 seconds.