基于CS-LBP与自适应神经网络的虹膜识别

Iris recognition based on CS-LBP and adaptive neural network

针对判断两个虹膜是否为同一类别的问题,提出了基于分块中位排序规则的中心对称局部二值模式(CS-LBP)提取虹膜纹理特征,并使用BP神经网络进行判断.BP神经网络的连接权重采用混沌系统-选择变异算子-粒子群(C-SM-PSO)算法进行自适应优化,提高设置连接权重的全局搜索能力,使神经网络具有自我跳出局部最优的能力,提高算法通用性.用多种算法在不同的虹膜库中进行了实验.实验结果表明,所提出的算法正确率较高,等错率较低,ROC曲线更接近横、纵坐标轴,具有良好的稳定性与鲁棒性.

Aiming at the problem of determine whether two irises are the same category or not,This paper propose center symmetric local binary patterns（CS-LBP）base on sub-block median ordering rules to extract iris texture feature,and use BP neural network to judge whether two irises are the same category or not.Use chaos system-select ＆ mutation operator-particle swarm optimization（C-SM-PSO）algorithm to adaptive optimize connection weights of BP neural network.Improve the global search ability to set connection weights so that neural network can jump out of local optimum by itself and improve generality of algorithm.Experimental results show that this paper algorithm has higher accuracy and lower error rate,ROC curve is closer to horizontal and vertical axis and have well stability and robustness.