Research On Position Recognition Method Of Gantry Hoisting Based On Machine Vision

At present,in the process of steel smelting in the steel plant,whether the gantry hook meets the lifting conditions of the iron water tank is mainly identified by the workers on the scene.The environment surrounding is extremely harsh.Harsh environment can cause both chemical and physical injuries to workers,such as chemical skin burns,poisoning of lead and its compounds,high temperatures,electromagnetic radiation,etc.Prolonged close contact can cause workers to become chronically poisoned with the compounds and develop silicosis.In order to ensure the worker's health and realize the identification task by machine instead of human,this paper completed the research on the gantry hoisting position identification method based on machine vision technology.The research object of this paper is the relative position of the hook and the lug in the gantry hoisting system.The convolutional neural network algorithm and S-Canny edge detection algorithm are used to identify the target image and extract the parameters,and the recognition results are verified according to the extracted characteristic parameters.Among them,according to the field investigation to determine the identification content and the parameters,through two groups of vision sensors in the three-dimensional space acquisition image of the hook and lug for the full range.According to the characteristics of the collected images,the AlexNet convolutional neural network,SVM support vector machine,KNN and BP neural network algorithms are used to train the model of the image set in the two cases of six classification and binary classification.Then,calculate the recognition accuracy of the verification.In order to evaluate and analyze the four algorithms in a more comprehensive way,this paper use four parameters,Sensitivity,Specificity,Positive Predictive Value and Negative Predictive Value to compare the performance of the algorithms,and select the best algorithm apply to the gantry hoisting position identification method.In order to solve the problem that the edge of image extracted by Canny edge detection algorithm is not clear,this paper improves the Canny algorithm,and the improved S-Canny edge detection algorithm can extract a more complete edge from the target image.Conduct 500 field experiments.The experimental data show that the average accuracy of the gantry hoisting position recognition method is 96.6%,and the recognition accuracy of the image which does not meet the hoisting conditions is 100%.