摘要
针对红外序列图像中运动弱小目标的检测问题,提出了一种基于自适应形态学Top-Hat算子和改进的自适应门限的弱小目标检测方法,其中形态学滤波器的结构元素采用两层前馈神经网络通过大量样本训练优化.将Top-Hat运算作为一个整体当作一层,输出层节点定义为作Top-Hat运算后图像矩阵的最大值,并针对所检测的大多数弱小点目标采用自适应门限进行分割,同时对SNR>4左右的点目标用固定门限进行分割.实验结果表明,该方法对SNR较低的复杂图像具有良好的滤波效果.
A novel method for self adapting morphological Top-Hat operator and improved self adapting threshold in spot target detection was presented. The structural elements of opening Top-Hat are trained by utilizing the two-layer feedback neural network from a mass of sample sets. The algorithm utilizes the two-layer feedback neural network to train structural elements by a mass of samples. The opening Top- Hat network is used as the input layer while the maximum of gray-scale image vector after Top-Hat operation is defined as the node of output. And adaptive threshold is adopted for low SNR spot target detection while the fixed threshold for high SNR（〉4） spot target detection. The experimental results show that this method is practical and easy to perform.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2006年第1期90-93,97,共5页
Journal of Shanghai Jiaotong University
基金
国家自然科学基金(60304007)
国防科技重点实验室预研基金(51483020203JW0310
51476010604JW0303)
航空科学基金(03F57003)
上海交通大学青年教师基金(51483020203JW0310)资助
