Application Research Of Packaging Detection And Sorting Technology Based On Machine Vision

With the wide application of intelligent equipment in the industrial field,the combination of machine vision and industrial robots is increasingly close.As an important part of quality control,detection plays an important role in the production of products.Machine vision detection technology has attracted the attention of various industries due to its advantages of non-contact,high speed and high precision.In this paper,the key technologies of high-speed and high-precision visual detection and dynamic target tracking algorithm are studied,which are closely combined with the requirements of detection and precise sorting of light and small materials in industrial automation production line.The main research contents are as follows:In the aspect of image preprocessing algorithm,firstly,a histogram equalization algorithm is studied to compensate nonlinear illumination,aiming at the problem that the image is too bright or too dark due to the change of illumination intensity and nonuniformity.Then,in order to suppress the image noise,an improved adaptive median filtering algorithm is studied.The dynamic filtering window is used to selectively filter whether the pixel points are noise points and retain the original edge details of the image.Finally,to highlight the edge,an improved unsharp mask(USM)image sharpening algorithm is proposed.The threshold value and scale factor of edge judgment are introduced to determine whether pixels are used for sharpening and sharpening.In the aspect of machine vision detection and recognition algorithm,firstly,Sobel operator is used to extract image edge features,and false edge and weak edge are eliminated by non maximum suppression and lag threshold processing.Then,a template matching algorithm based on edge features is studied.The similarity measure is the sum of the normalized point product of the edge gradient vector of the template and the image to be tested.The defect is judged by comparing the result of similarity measure with the set threshold value.On this basis,in order to improve the detection speed,an early termination search strategy is proposed,which can terminate the calculation of the template in the current position in advance by setting up the part of similarity measurement and limiting conditions.Then,the image pyramid is built to match layer by layer according to the matching rules,which reduces the complexity of the algorithm to further speed up.Simulation results show that the optimized speed of detection algorithm is about 126 ms.Finally,a character recognition algorithm is studied.Otsu is used to segment the character region,and mathematical morphology close operation is used to connect the single character discontinuity to form a separate connected region.In the aspect of dynamic target tracking algorithm,firstly,aiming at the problem that there are duplicate targets in the adjacent frame image caused by continuous camera taking,a kind of image de duplication algorithm based on generalized coordinates is studied to avoid the interference of multi-target position points.Then a dynamic target tracking algorithm is proposed to calibrate the delta robot and conveyor belt,vision system and conveyor belt respectively.The algorithm combining the Zhang Zhengyou calibration method and conveyor belt calibration is used to track the position of the moving target in real time,and the robot grabbing strategy is analyzed.Finally,combined with the actual production process of mask automatic production line of a factory,the application of mask packing detection and sorting is studied.The application results show that the system can effectively distinguish whether the mask code is qualified,the recognition rate and the missing detection rate can reach 99.83%and 0.2%,the character recognition rate is higher than 99.50%,and the tracking accuracy is less than 0.81 mm,so as to meet the actual engineering requirements.