Image Analysis And Quantity Detection For Stacked-sheets Based On A Minimum Curvature Method

In the manufacturing field of sheet products,such as paper,cigarette package,packing box,pcb circuit board,solar silicon chip and so on,the detection of laminated sheet is an essential part.The precision of the counting directly affects the economic benefit of the factory and the subsequent production operation.However,the traditional manual and physical counting methods have many problems,such as large error and low efficiency,which can not meet the requirements of real-time and accuracy in industry.With the continuous development of machine vision technology and its advantages of non-contact and strong timeliness,the research on the application of machine vision to the number detection of thin slice is increasing gradually.There are many problems in laminated sheet detection.First,there will inevitably be some abnormal conditions in the course of sheet storage,such as adhesion to sundries,moisture,damage and so on,which will increase the difficulty of sheet detection;The second is that different types of thin sli ces require different exposure value and resolution,and the third is that the feature points of thin slice image are few,so it is difficult to locate and count the splicing position.In order to solve the above problems,this paper designs a image analysis and number detection system based on minimum curvature.The details are as follows:(1)A multi-camera imaging system with convenient adjustment is designed,which can change the working distance of the camera according to different measuring range and resolution requirement,so that the laminated slice images with different visual field range can be collected.(2)Aiming at the difficulty in detecting the thin slices with abnormal conditions,a line detection algorithm based on the extremum of curvature is designed according the texture characteristics of the image.By calculating the curvature value of the grayscale profile of the laminated sheet,the central point of the slice is detected.The picture can not ensure that the clarity is uniform and the slice has no abnormal condition.So we need to correct the center detected point s according to the center distance and the gray value.Finally the result of the line detection of the laminated sheet is acquired.(3)The imaging field of a single camera is limited,so it can not meet the needs of industrial detection of more thin pieces at a time.This paper adopts a method of image stitching by determining the position of stitching.First,the splicing point is estimated and corrected by gray value.Then the camera is calibrated,the internal and external parameters of the camera and the rotation translation matrix are obtained,and the image is corrected by polar line.Then the final stitching position is determined by correlation measurement to complete the image mosaic.Finally,the sheet is counted by peak detection.(4)The application software is designed according to the customer demand s.And three different exposure values are designed to deal with the different luminance image.Many kinds of thin pieces are tested,such as thick smoke label,thin smoke label and pcb circuit board,bicycle tag,clothing trademark and so on.