基于纹理图像的点张计数算法设计与分析

Design and Reseach of Paper Counting Algorithem Based on Texture Image

针对手工点张耗时长、精度低的问题,提出了一种应用图像纹理分析的纸张计数算法。用高精度图像摄取装置对纸张图像进行采集,对原始纹理图像采用高斯滤波法去除噪点,大大提高了计算机处理速度。引进方向滤波与Gabor小波变换结合的方法对纹理图像进行增强提升了点张准确度;针对图像中难以处理的纸张堆叠问题,提出了计数修正算法。对不同厚度及张数纸堆进行了点数试验。对于0.3 mm的堆叠纸张,点张准确率100%;对于0.2 mm的堆叠纸张,点张准确率达99%以上。试验结果表明该算法具有较好的计数能力,计数准确高。

A paper counting algorithm based on the paper texture analysis is presented. The aim of the optimal design is to replace the manual paper counting with time-consuming and low precision, which collect the paper image with high precision image pickup device. To improve the speed of processing speed, Gauss filtering method was used to remove noise. Then directional filtering method and Gabor wavelet transform are combined to enhance the texture image. A counting correction algorithm is proposed according to the paper stack case, for the 0.3 mm stack of paper, the algorithm points to an accuracy rate is 100%, for the 0.2 mm stack of paper, the point of the accuracy rate is more than 99%. Test results show that the algorithm has good counting ability and high count accuracy.