摘要
针对焦平面红外图像中运动弱小点目标的检测问题,基于形态学滤波器和模糊决策融合构建了一种新的弱小目标检测算法。针对单帧检测,基于目标在实测红外图像上所呈现的凸包结构特点,设计了圆形形态学滤波器结构,并引入神经网络进行圆形形态学滤波器结构元素优化设计。同时在多帧关联检测的基础上,引入决策融合概念,基于贝叶斯最小风险准则建立了基于模糊决策融合的序列关联检测方法。实测数据的处理结果表明:针对低信噪比图像(SNR≈2),在虚警概率≤1%情况下,新算法对复杂红外弱小目标图像检测概率≥98%,有效地提高了检测算法的性能。
For the detection of moving dim small point targets in infrared focal plane image,a new algorithm of target detection based on mathematical morphology filter and fuzzy decision is proposed in this paper.For the single frame detection,the circular mathematical morphology filter is designed according to target characteristic in the measured infrared image,and its structure is optimized with neural networks.Besides,based on the multi-frame detection,the decision fusion is introduced and the sequential detection method,which is based on fuzzy decision fusion is build according to minimum Bayesian risk criterion.Experimental results of practical data show that the detection probability of images(SNR ≈2)can reach more than 98% with 1% false alarm with the optimized circular morphological filter,which could enhance detection probability efficiently.
出处
《火力与指挥控制》
CSCD
北大核心
2010年第9期184-188,共5页
Fire Control & Command Control
基金
航空科学基金资助项目(20075157007)
关键词
圆形形态学滤波器
神经网络
贝叶斯准则
模糊融合
circular mathematical morphology filter
neural networks bayesian criterion
fuzzy fusion