Research On Sorting Method Of Non-ferrous Metal Scraps From End-of-life Vehicles Based On Machine Vision

End of life vehicles(ELVs)contain large amount of steel scraps and non-ferrous metal scraps,which are of high recycling value.In the existing metal scraps sorting process,after the steel scraps are sorted out by means of magnetic field and eddy current,the remaining mixed non-ferrous metal scraps with high added value(mainly including copper and aluminum metal scraps)cannot be further sorted.Non-ferrous metal scraps features are confusing and complex in shape,and traditional sorting methods cannot meet the requirements of accurate identification and efficient stacking.In this paper,machine vision recognition is combined with air knife stacking method to establish a machine vision-based sorting method for non-ferrous metal scraps.Aiming at the difficult problems in the process of visual recognition and air knife stacking,the research of metal scrap image deblurring,metal scrap image recognition,metal scrap motion characteristics under the action of air jet and the research of parameter coupling relationship in the process of metal scrap stacking were carried out to realize the accurate recognition and fast stacking of non-ferrous metal scraps.The main work is as follows.(1)Blurred image restoration algorithm for moving metal scraps.Aiming at the image blur caused by the relative movement of the camera and metal scraps during the recognition process,a statistical analysis method of the local gradient feature of the image is proposed to study the image blur characteristics,reveal the gradient change law of the image blurring process,and establish a priori model of directional gradient.Combining the prior model with the maximum posterior probability method,through the iterative solution method of clear image estimation and fuzzy kernel estimation,the restoration of the blurred image is realized,and the influence of the image motion blur on the calculation of the pixel center point of metal scraps is reduced.(2)Accurate recognition of easily confused broken material images.The image superpixel method is proposed to aggregate similar pixels in the local area of non-ferrous metal scrap image to reduce complex pixels.An automatic labeling method for the super-pixel unit data set of non-ferrous metal scrap is proposed,which greatly enriches the data set and shortens the data labeling time.The residual network is used to classify the super-pixel data set,and combined with the large-scale image information,the recognition result optimization method is proposed to realize the accurate recognition and region segmentation of the non-ferrous metal metal scraps image in the complex environment.(3)The movement characteristics of in stacking process.In order to study the movement of the metal scraps with uncertain initial speed and posture under the action of high-speed air knife jet,firstly,the theoretical movement trajectory of the metal scraps is obtained by calculation,and the appropriate air knife strike time is determined.Subsequently,experimental methods are used to study the impact time and duration of the air knife jet on the movement trajectory of the metal scraps,analyze the impact of different jet impact time and different jet duration on the distance of the metal scraps stacking distance and the stability of the movement process,and determine the best jet strike time and jet duration during the stacking process.(4)Optimization of the process parameters of meal scraps stacking.In order to accurately obtain the movement of the complex-shaped metal scraps along with the conveyor belt and the attitude and speed when flying out of the conveyor belt,a modeling analysis was carried out on the process of metal scraps stacking through the discrete element-finite element coupling method.A method to characterize the geometric shape of the metal scraps based on the spherical harmonic series is proposed,and a geometric model conforming to the characteristics of the shape of the metal scraps is randomly generated as the input of the simulation.The interaction and movement laws of the metal scraps on the conveyor belt are analyzed by the discrete element method.The calculated movement state information of the metal scraps is used as the initial boundary condition of the finite element simulation,and the material-jet coupling simulation analysis is carried out,which reveals the change rule of the separation distance of the metal scraps during the stacking process.The response surface optimization method was adopted to carry out the multi-factor multi-level simulation experiment design and parameter optimization of the sorting parameters,and the optimal working parameters in the process of metal scraps were obtained.(5)Integration and experimental verification of the sorting method for non-ferrous metal scraps.Based on the theoretical calculation,we analyzed the time constraint relationship between the identification process and the stacking process and the theoretical sorting efficiency,based on which we selected the hardware and built the sorting experimental platform and integrated the identification and stacking methods,carried out the non-ferrous crushed material sorting experiments,and analyzed the movement trajectory of the crushed material in the actual sorting process.The sorting comparison experiments were carried out with multiple sets of parameters,and the optimal sorting parameters were verified,and the sorting accuracy and purity indexes were obtained under the parameters.The paper establishes a machine vision-based sorting method for non-ferrous metal scraps from ELVs,solves the difficult problems in the process of metal scraps identification and stacking,realizes the improvement of sorting accuracy,and provides technical method support for the efficient recycling of non-ferrous metal scraps,which has important academic value and engineering guidance significance.