Image Stitching And Line Detection Algorithms For Counting Of Thin-type Stacked Paper

Accurate statistics sheet counting have an urgent need in printing and packaging industry, which has significance for the enterprise cost control, the traditional weighing and mechanical counting measurement have large error and low efficiency. Particularly for the thin-type paper counting. With the rapid development of machine vision technology, the counting method based on machine vision has been applied in the paper counting. But the thin-type stacked paper counting exists a problem that a single camera can’t balance between high resolution and wide measurement range. Therefore, we design a device for thin-type stacked paper counting based on moving camera image capture and combining with the image stitch algorithm to reconstruct the complete end image according to the characteristics of thin-type paper, The detail contents are as follows:Firstly, imaging system is designed according to the characteristics of thin-type paper, the imaging components are selected by user requirement. This paper also discuss the operation flow and control method of the imaging system. A stable and reliable imaging platform has been builded by single camera muti-station mobile acquisition. We obtain sequence images of the stacked paper successfully.Secondly, we design the image stitch algorithm to restructure complete end image. We use double labeled ruler to improve the stitching efficiency in view of the high regularity and high repetition rate of gray scale. The speeded up robust features of stacked paper have extracted through reducing the matching area、feature point optimization and using the new point descriptor, Adjacent image transformation relation is determined by feature point matching and thereby complete the image stitch. And through experiments, we prove the proposed algorithm is improved in compute speed and matching efficiency.Thirdly, a line detection algorithm which use the zero-threshold canny algorithm and combined with Radon transform was proposed in this paper according to the characteristics of thin-type paper. Maximally edge effective information are obtained by zero-threshold canny edge detection, And we limit the gradient direction to obtain a single response. Lines in the image are convert to local extremum points in the radon space through radon transform. The local extremum points satisfying the threshold condition are obtained in the search space, the purpose of paper counting can be achieve by calculate the number of local extremum point. In order to verify results of line detection, we determine the position of each line according to the corresponding local extremum point and mark the line in the image.Finally, we use LabVIEW and C++programming language to develop the paper counting software platform, under the visual studio programming environment, image process modules have developed by opencv computer vision library, communication between devices, process control and interface design are realized under the LabVIEW Programming environment. Instrument test and result evaluation are completed finally. The research and development of stacked thin-type paper counting instrument are also contribute to the improvement of related industry automation and intelligent improvement.