针对红外序列图像中运动弱小点目标的检测问题,设计了一种基于改进遗传算法优化的修正Top-Hat形态学滤波器算子。其中,优化的修正Top-Hat形态学滤波器可以很好地抑制背景和噪声的影响;改进遗传算法采用新的区间离散化编码和自适应的主次式交叉与变异算子,通过优化搜索全局空间得到的形态学滤波器参量具有较好的滤波性及时效性。并且针对不同信噪比的点目标检测建立了自适应门限。实测数据的处理 结果表明:在虚警概率小于5%情况下,优化的修正Top-Hat形态学滤波器算子对信噪比约为2的复杂图像检测概率大于等于70%,与固定结构元素的Top-Hat形态学滤波器相比检测概率提高了近10%,与用经典遗传算法训练的传统Top-Hat形态学滤波器相比检测概率提高了4%。

Toward detection of feeble moving infrared spot target, improved Top-Hat morphological filtering operator is presented based on improved genetic algorithms. The optimized improved Top-Hat morphological filter restrains background and noise. And the genetic algorithm is improved with new interval discretization code and adaptive master-slave crossover and mutation operator. The optimized morphological filter based on global search has better filtering and time performance. To different signal to noise ratio (SNR) spot targets, the adaptive threshold is adopted for detection. Experimental results show that the detection probability of complicate images (RSN≈2) can reach more than 70% with false alarm no more than 5%. Compared with fixed Top-Hat filter, the detection probability improves nearly by 10%. Also compared with traditional Top-Hat morphological filter optimized by classical genetic algorithms, the detection probability is improved by 4%.