Research On Automatic Counting Image Processing System Of Steel Rods

Automatic count of the steel rods in assembly lines is a tough problem which has been puzzling producers for long. At present, the count of the steel rods online is implemented by manual, This method, however, is labor-consuming, inefficient and limited in systematic producing capability while it is difficult to ensure the counting accuracy. Thus, developing a automatic counting system, which is able to automatically detect the number of the steel rods, will enhance productivity effectively, improve the traditional machining pattern that relies on manual calculating, free workers from heavy physical labor and horrible working environment, as well as make the stage of count achieve mechanization and automation in operation.In order to achieve this goal of steel bars'automatic counting, diagnostic engineering center on equipment in Northeastern University and FuShun steel Co. Ltd cooperated on the automatic counting system online for steel bars, while my thesis is the system's image processing.This article serves the counting purpose through the recognition of the steel bars'image, its main research content including image pretreatment, image division and pattern recognition. In the part of image pretreatment, this paper focuses on the binary image's transform algorithm, proposing waveform analysis, the largest category of variance, the smallest error, the maximum entropy method, the smallest skew ness maintain law and the method of moments of several global thresholding segmentation algorithm, its purpose is to get more ideal binary image. In the part of image segmentation, this article proposes the segmentation algorithm based on the limit corrosion and the conditions expansion. This algorithm can make each region contract to a spot or a non-inner point small region, this spot or the small region are called the core region, uses the conditions expansion again, then you can get the non-adhesion region which is restored from the core region. In the image recognition counting part, firstly,doing distance transformation to the processed binary images, then you can get the cores of all regions, search on these core spots again and achieve the bar count at last.In addition, this paper carries on the research of this system through a large number of experiments, and which includes all kinds of lighting equipment, as well as the expansion of the steel bar. In a word, experiments show that the system will be able to handle all kinds of steel bars'image which is in different circumstances.