Research On Bar Adaptive Counting Based On Picture Processing

Chinese iron and steel industry has developed rapidly in recent years, bars as all one ofthe necessities of the building are the main products of iron and steel production. Afterinvestigation, the quantitative packing domestic bar level remains low, mainly by artificialcounting. Manual pointing count’s requirements for technologies is not high, but it needpeople eyes’ long time high concentration. This will hurt the human eye greatly, and theaccuracy falls down. Besides, the temperature of Steel rod which just burning well is high, itwill cause great harm to human body. Collectively speaking, manual pointing countingmethod, supports high strength of work, low productivity, high error rate. Because the barproduction workshop environment’s complex, Cutting technology’s instability, barscounting methods detected by electro-optical sensors and weight sensor are unable to meetthe needs of production lines. This article has been presented a solution based on imageprocessing to solve bar’s adaptive counting problems. The program is to number thepackaged bars which have been bundled.This article includes the following main parts:(1) Firstly, briefly explain the bar’s image acquisition methods, analysis thecharacteristic of the original image and common image pre-processing methods and focuson format conversion, color images to grayscale, image filter and other common imagepre-processing methods.(2) An image segmentation algorithm based on LBP and region competition isproposed. Considering the specific goal of bar image segmentation, this method improvesthe original region competition algorithm, proposes a new assumption about regioncompetition, rewrites the energy function based on LBP histograms, and also develop thetwo-stage iterative algorithm to make our energy converge to a local minimum. Experimentsshow that this method simplifies the parameter estimation, reduces the time and effectivelyfilters background part from image, having a great help for the subsequent bar recognition.(3) Research identification and counting of bar targets, and come up with an adaptivealgorithm which can calculate the circle radius of bars. The algorithm detects circles byHough transform, and calculates radius and average radius. Builds bar template according tothe average radius and recognizes bar in the image. Then uses increasing Thresholdcorrosion to find the center of bar and labeling and counting. Experiments show that thealgorithm has a very high accuracy. (4) Design a simple error filter algorithm. The algorithm based on clustering principle.Count numbers according to the morphological characteristics and the distance of two barcenters. This algorithm has good correcting effect for the situation of the bar’s enddisplaying irregular.