Line Detection And Image Stitching Algorithms For Quantity Measurement Of Stacked-Paper

In the packaging and printing industry, an accurate measurement of paper quantity is of great significance in high-end printing paper products, which has a direct impact on economic efficiency of related companies. The traditional height and weight measurement of physical parameters have precise defects. With its noncontact, no physical damage and real-time measurement merits, machine vision method has been widely applied to quality control for high-end printing products.The development of stacked-paper detection technology based on machine vision faces two big problems:(1) We must meet the measurement demands of stacked-paper under high resolution conditions. (2) Complex imaging conditions caused by the detection problems such as different colors, uneven illumination, malalignment, slant etc. Aiming at high stacked paper measurement, we designed a dual camera array model; in order to improve the identification precision of stacked-paper, we also adopted and improved a level line guided line-segment growing algorithm to solve complex imaging fringe detection problem. Finally, we implemented the line features matching method in the overlapping area and completed the statistical process of frequency. The main contents of this paper are as follows:(1) To meet the actual demand of real-time measurement in high stacked paper, we successfully build a stable and reliable system based on camera array imaging platform. The core component of visual imaging system such as lens, industrial cameras, lights, support platform etc. are introduced in detail, and a complete dual camera array imaging system is established.(2) Aiming at the complex imaging conditions of detection problem, an improved LSD line detection algorithm was introduced in detail in this paper. We circumvent the fringe detection problem in stacked-sheet images by introducing a level line guided line-segment growing algorithm to improve the precision and completeness on fringe identification. A unidirectional gradient operator is adopted to eliminate multiple responses on a single fringe. The gradient magnitude and level-line direction are combined to improve the growth of line support regions in noisy environment, regional growth and probability of error control steps to completely identify each sheet fringe, a connected component analysis of Bresenham algorithm is integrated to remedy the local gap in line detection.(3) According to the characteristics of the stacked-sheet lines, we designed an image stitching method for line features and proposed a weighted voting statistics method, we also used the automatic correction of alignment pretreatment method dealing with the mismatch problem in dual camera array. Then we introduced overlap area positioning, linear feature extraction and stitching methods to verify the performance of the proposed algorithm. Besides, the weighted voting frequency histogram statistics method was used to get the final count results.The instrument and algorithms in this paper has been successfully applied into the factory workshop applications, through a large number of different types of printing paper tests, the long-term measurement error is less than 0.75%o and various kinds of different colors and thickness of stacked papers have been tested sucessfully, it is sufficient to meet the requirement of factory applications, what could greatly reduce the investment of manpower material resources, and improve the production efficiency. The proposed method is also helpful to the improvement of relevant industry paper quantity statistics.