一种新颖的强海杂波背景下弱小目标鲁棒检测算法

A Novel Infrared Small-dim Object Detection under Complex Sea-clutter Background

红外弱小目标的自动检测是光电火控系统、红外导引头等武器装备中的关键技术之一。本文针对复杂海杂波背景下低信噪比目标检测问题,结合SURF特征描述符,提出了一种基于自动聚类分割的弱小目标检测算法。该方法首先利用SURF算法计算出红外图像中的兴趣点,并将多帧的兴趣点投影到一帧中形成累积图。由于目标在红外序列中的连续性,目标点会在累积图中进行聚集;然后使用改进的快速聚类分割算法自适应的检测出疑似目标;最后根据目标类具有的大小约束与线性约束先验信息,从海杂波中区分出所需的弱小目标。大量仿真实验及实物验证表明,相比于其他现有算法,本文提出方法在处理具有较低信噪比/杂波比的视频下的弱小目标时可以获得较好的检测性能,同时该方法的实时性强,可在多项式时间复杂度下得到最优检测目标,适合工程应用。

Infrared small-dim object detection under a complex background is one of the key technologies in the photoelectric fire-control system and infrared seeker. As for the problem of low-SNR target detection in a complex sea-clutter background, combining the SURF feature descriptor, a small-dim object detection algorithm based on automatic clustering segmentation is proposed. Firstly, the SURF algorithm is used to calculate the points of interest in the infrared image, and these points are projected ontoa frame to form a cumulative image. The object points are aggregated in the cumulative imagedue to the motion continuity of the object in the infrared frames. Then the improved fast-clustering algorithm is adopted to detect the suspect target adaptively. Finally, according to the prior information onthe size constraint and linear constraint of the object cluster, the small-dim object is distinguished from the sea-clutter. A large number of simulation experiments and fly-by-flight verification show that compared with other existing algorithms, theproposed algorithm can obtain better detection performance when dealing with small-dim objects with low SNR, and it can detect the optimal detection target in the polynomial time complexity, which is suitable for engineering applications.