Design And Implementation Of Vision Detection System For Steelmaking Sublance-probe Handling

The technological development of each steelmaking manufacturing process is related to the process of industrial modernization and intelligent manufacturing in our country.In the traditional steelmaking temperature measurement process,sublance-probe handling need to be operated by workers periodically,and remove the burned probe after the temperature measurement is completed.Since this work is of high-risk and in harsh working environment,it is urgent to construct a robotic automatic operation system for sublance-probe handling process to replace manual operation.Concerning with the task of temperature measurement and sublance-probe handling,this study focuses on how to effectively detect and locate the probes on the magazine and the sublance fixed on the chain transmission mechanism under complex illumination conditions.A histogram key point constrained homomorphic filtering adaptive image enhancement algorithm is designed,which is combined with a feature detection algorithm based on line segment detector and arc segment detector to finish the object detection and localization task.In the vision detection system,the probe is of a circular tube structure,which will cause degraded image with defective or fake features,this study proposes a histogram key point constrained homomorphic filtering image enhancement method.Firstly,image Lambertian reflectance model is presented and the appropriate key points in the grayscale histogram are selected as the control matrix parameters by calculating the ratio of interval pixels.Afterwards,We constrain the form of filter in the homomorphic filtering by the control matrix and adaptive to illumination conditions.This method simultaneously normalizes the brightness across an image and increases contrast under low illumination conditions,while for high illumination conditions,it will suppress the high-frequency signals and compress contrast with a low-pass filter,which can reduces the adverse effect of fake contours in this circumstances.In the part of feature extraction,detection and localization,since the image enhancement algorithm can only reduce the adverse effect of ambient light to a certain extent,and the limitation of the processing accuracy of the probe,which will cause the traditional feature extraction algorithm of poor efficacy or even lose efficacy,this study proposes a feature extraction algorithm based on line segment detector and arc segment detector,and combines line fitting and ellipse fitting algorithms to detect and locate the object.The method first extracts the fundamental line segment features and arc segment features of the probe and the sublance,and improves effectiveness of the based algorithms,and then designs the constraint optimization ellipse feature fitting algorithm with specific contour extraction strategy.This method performs higher accuracy and robustness than traditional correlative feature extraction algorithms,and provides more accurate information for robot operations.Experiments are performed of the proposed method.Firstly,the effect of image enhancement algorithm is tested on the image of the end-face of the sampler under different illumination intensity and different exposure parameters.Subsequently,a laser tracker is used to perform an accuracy test on the localization results of the sublance,and a photoelectric sensor is used to perform an accuracy test on the detection and localization results of the probe end-face,and compared with the results of other algorithms.Finally,we perform the peg-hole assembling task experiment based on the detected and located results of the vision detection system to verify the effectiveness in actual operation.