An Apparatus For Stacked Packing Paper Counting Based On Machine Vision

With the rapid development of the tobacco industry, packing paper grows fast in quantity and quality. Due to the diversification of the printing process and application of anti-counterfeiting technique, the price of packing paper rises a lot, thus a large number of accurate sheet-counting is required. The conventional ways for sheet-counting include manual counting and using some measurement analysis, which both generate inaccuracy and harm the packing paper because of the physical contact. As microelectronic technology and computer science develops, machine vision methods are widely adopted in industrial field due to their nondestructive and real-time merits.In domestic tobacco company, a standard stack usually contains hundreds or even thousands of packing papers during the process of printing and packaging. A single-camera scanning system suffers from insufficient measuring range or resolution while handling these ultrahigh stacks; if using a motor to move the camera to image in several positions, this system will be time-consuming. To solve the problem, we design a new real-time counting system using camera array. The research contents are as follows.①The background and the significance of this paper are introduced, the concept and application of machine vision are presented. Overseas and domestic research status of sheeting counter is introduced and relative methods are described; key technology and difficulty of counting is analyzed.②By analyzing the characteristics of stacked packing papers, an available scheme for image system is designed. The core devices for imaging are introduced, including industrial camera, optical lens and light source, device selection is conducted after analysis of parameters and experiments. Lighting methods are described, high-angle direct shooting is selected through experiments. A new imaging system consists of camera array are applied to capture high-resolution images. At last, a steady imaging system is established.③Base on the feature of images, the edge contour of packing paper are selected as pointcut. The principle and general method of edge detecting are presented in detail. Canny detector is modified to only detect the left edge contours of packing paper. Hough transform and a false line filter are applied to extract a single line corresponding to each packing paper. More importantly, to avoid double-counting the overlapping sheets, a marking plate is used to fuse the results from multi-camera imaging to get the number within the particular stack.④The software for counting system is designed, mainly including camera control, counting program, display interface system. How to operate the counter device and data flow diagram are presented. Some key functions are introduced at length. The reliability of the proposed algorithm is verified from two angles. On one hand, several unfavorable packing paper images are processed, the number is compared with the manually discerned results; on the other hand, statistical analysis of experiments are done to obtain precision and recall.Through more than one year’s test in factory production lines, it is demonstrated that our apparatus can satisfy the high-accuracy requirement of real-time sheet-counting task with an error less than6%oo. This counter may be conveniently extended to quality control of high-quality printing and counting of sheets, card etc.