低对比度叠层薄片的机器视觉计数测量方法

为解决印刷等行业在薄片计数上的难题,提出了一种基于机器视觉的计数测量方案.采用最新的高分辨率的CMOS相机和百万像素、低畸变的定焦镜头搭建图像采集系统.针对低对比度叠层薄片检测难题,提出了一种局部直方图均衡化和基于双高斯二阶导的线状增强相结合的算法.算法主要分为以下几部分:首先,获取采集平台采集的薄片图像,为提高算法速度,只选取采集的部分图像作为处理区域,并通过小波去噪对其进行预处理;接着,利用局部直方图均衡化对预处理后的图像进行对比度增强;然后,在水平和竖直方向分别用双高斯二阶导数和均值滤波器处理,得到线状增强后的薄片检测结果;最后,对薄片检测结果进行峰值检测计数,并将检测结果实时显示出来.实验证明,此计数测量方法对多种类型的薄片(包括纯白的0.05mm的薄片)的计数误差都小于0.01%,而且可以节省大量人力和物力,值得被推广和运用.

A Count Measurement Method for Low Contrast Stacked Sheets in Machine Vision

To solve the problem of sheets counting in the printing industry and so on, the count measurement method based on machine vision is proposed. Image acquisition system is built by using the latest high-resolution CMOS camera and megapixel, low distortion of the fixed focus lens. Aiming at the problem of detecting low contrast stacked sheets, a local histogram equalization algorithm and a linear enhancement based on bi-Gaussian second order are proposed. The algorithm is divided into the following parts. First, to improve the speed of the algorithm, only part of the image, which is obtained from image acquisition platform, is selected as processing area, and wavelet de-noising is performed. Second, the preprocessed image is enhanced by the local histogram equalization. Then, the detecting of the stacked sheets is acquired by using second-order derivative of bi-Gaussian and mean filter processing in the horizontal and vertical directions, respectively. Finally, the counting result can be acquired by the peak detection, and displayed in time. Experimental results show that the error rate of the count measurement method is less than 0.01% for the various types of sheets (including the 0.05 mm pure white sheets). It can also save a lot of manpower and material sources. Therefore, it deserves being promoted and used.