Abstract: A novel automatic detection algorithm for infrared ship target under the cluttered background has been proposed,which uses differences in the histogram of image brightness between the FLIR ship target and the background.The initial threshold corresponding to ship is gained through seeking the peak of the polynomial curve robustly fitting the gray-level histogram based on the histogram analysis.The local recursive segmentation is followed if the initial threshold is not reasonable.Finally,the background is further removed according to the priori information such as the waterline and the boundary features of sky and the background.The experimental results with real FLIR images prove that the presented method can reliably detect infrared ship targets under heavy clutter background,and verify the adaptability and validity of the method.

Key words: target detection, infrared ship target, histogram analysis, local recursive segmentation, adaptive

摘要： 针对背景杂乱的红外舰船目标检测问题,提出了一种红外舰船目标的自动检测新算法。该方法利用红外舰船图像中目标与背景在灰度直方图上的差异,通过对拟合直方图的多项式曲线参数鲁棒求解,进而求出舰船目标的分割阈值。然后,根据红外舰船目标亮度与图像平均亮度的关系等,对求得的阈值合理性进行判断。若该阈值不合理,则将其作为阈值初值,对红外舰船图像进行自适应局部递归分割。最后,结合红外舰船目标吃水线、天空与背景的边界特征等先验知识,对分割出的背景进行剔除。实验结果表明,该方法对强杂波干扰的红外舰船目标能实现可靠的检测,具有很好的适应性和鲁棒性。

关键词: 目标检测, 红外舰船目标, 直方图分析, 局部递归分割, 自适应