基于神经网络的虹膜图像上眼睑的精确定位

由于眼睫毛等复杂因素的干扰,虹膜图像中上眼睑的精确定位一直是虹膜识别中公认的难题。如果上眼睑定位不准确,将会在识别过程中引入大量噪声,严重影响虹膜识别的准确率。为了获取上眼睑的准确位置,首先构建了用于分类的神经网络,借助于滑动窗口,精确获取了上眼睑的关键点;然后,用二次曲线对关键点进行拟合,得到上眼睑的精确位置。最后,对该算法进行了评估。结果表明,该算法具有高准确率和强泛化能力。

Accurate Location for Upper Eyelid of Iris Image with Neural Networks

Detecting upper eyelid accurately is considered difficult in iris recognition due to the interference of complex factors such as eyelashes. If the upper eyelid positioning is not accurate, a lot of noise is introduced in the process of recognition, which seriously affects the accuracy of iris recognition. First, a neural network is constructed to precisely locate the upper eyelid. Next, the key points of the upper eyelid are obtained by the network with the help of the sliding window. Then, the exact position of the upper eyelid is obtained by fitting the key points with parabolic curves. Finally, under the strict evaluation criteria, it is proved that the algorithm has high accuracy and strong generalization ability.