The Research And Design Of Online Bar Counting Based On Fast Center Location Algoirthm For Quasi-circular

The quantificational packing of reinforcing steel bars in steelworks is still staying at a lowlevel. Reinforcing steel bars are counted by labor in most steelworks. On one hand, this work isso monotony that mistakes always take place, so it is inefficient and inaccurate; on the otherhand, as the packing number is not accurate, most plants sell bars by weight, it results ineconomic losses for plants and leaves trouble to the following business of selling, because theinternational way of selling bars is by bundle which has fixed number, that means if the weight isthe same, the plants can product more bars by negative tolerances, So developing an automaticsystem for reinforcing steel bar counting shall meets practical need of steelworks.Among many practical counting methods, mechanical-electrical method and digital imageprocessing method are the most important ways. But when the bars are crossed and moving fast,mechanical-electrical method is very difficult to meet production needs, while its maintaining iscomplexity and high cost. Considering the disadvantages of mechanical-electrical method, thispaper chose digital image processing way to achieve online count, this method is not only easy tomaintain, but also has a higher precision because of counting bar before they are packed.For the character of live product line, present a complete general structure of on-line barcounting system in this paper. It mainly includes hardware and software, the part of hardwareconsists of light, CCD cameras, frame grabbers, industrial machine, CRT display devices andother communication modules, while software portion of the image processing algorithm is theresearch focus in this paper.Image processing process includes four parts: image preprocessing, image segmentation,object center location, motion estimation, matching and counting. The image preprocessingsection involves the selection of interest region, morphological processing. The segmentationpart uses the multi-threshold OTSU segmentation based on improved particle swarmoptimization (IPSO-MOTSU) to obtain more effective steel bar digital images and improve thereal time performance. After getting the clear and effective images of the steel bar, this papercreates a new center location algorithm of high-fitting circle’s chord perpendicular bisector focusfor quasi-circular which is based on Hough Transform, in order to locate all centers in eachframe of the images quickly and precisely. The part of moving center tracking is Muti-Targettracking, making use of the Kalman filter motion prediction and the small fixed windowsearching to math the center one by one accurately and count them effectively between theadjacent frames. By a series of chain optimization and parallel processing, greatly improving theprocess real-time, while images of live bar are tested, the results show that the scheme has a highcount accuracy and good real time performance.