基于卷积运算的纸币红外图像斑马线特征提取

Convolution approach for zebra stripe feature extraction of infrared paper currency image

提出利用卷积运算方法对纸币红外图像特征之——斑马线结构特征进行了提取,与数学形态学、canny边缘检测等特征提取结果进行了比较,并在MATLAB环境下编程实现。实验结果表明,在红外防伪点较难清晰成像的情况下,分析特征固有频率,设计了合适的卷积核,利用图像水平投影和信号卷积运算能够克服光源不均匀及纸币新旧不同带来的识别困难,将红外图像的弱目标特征提取出来,算法简单,应用性强,鉴伪效果好。

Feature extraction is a fundamental step in image recognition. An infrared feature extraction algorithm based on convolution about paper currency is proposed in this paper. This method is implemented by MATLAB software and compared with morpbogical and canny detection. The experimental results show that when there is much background noise, better target identification effect is obtained by using horizontal projection and selecting appropriate convolution kernels. This algorithm is simple and efficient.