Research On Fruit Recognition And Positioning Technology Based On Machine Vision

Most of the fruit is picked in batches from the orchard,and then divided into different levels by hand according to size and maturity.This method of manual sorting is time-consuming,laborious,and inconsistent in the judgment of different people.In order to solve this problem,it is necessary to have an objective judgment standard,which can automatically classify the size and maturity of fruit products.In the sorting task,the use of machine vision can not only improve work efficiency,but also solve the problem of artificial judgment nonstandard.Therefore,this paper designs a fruit sorting system based on machine vision,which mainly includes visual control module and motion control module.The main research content of this paper is the visual control module,which includes image acquisition,image preprocessing,image color recognition,image segmentation,feature extraction and coordinate positioning.In the sorting system,the image is captured by using a camera placed above the conveyor belt,and the object features are identified by the software execution algorithm.At the same time,the identification information and the position information of the object are provided to the database.The identified projects mainly include image color,edge,area and centroid coordinates.In this paper,the color of the fruit is recognized for maturity classification,the image area is calculated for the size classification,and the centroid coordinates are calculated for location.In this paper,the adaptive neural fuzzy inference system(ANFIS)is applied to the object positioning,using ANFIS to adjust membership function's input parameters and output parameters of the fuzzy logic controller,a hybrid learning algorithm for training.The algorithm adopts the least square algorithm to adjust the linear output parameters of the membership function and adjust the nonlinear input parameters of the membership function by the gradient descent method.In the training phase,the pixel coordinates as sample input,the actual coordinates as expected output,the sample input and the expected output according to certain format combination into training data of ANFIS algorithm,and according to certain error criterion to adjust the corresponding parameter,so that the error is minimized.In this paper,the experiment was carried out with tomato and passion fruit as an example,and the experimental results were analyzed.Finally combined with motion control module,which makes the realization of fruit sorting,and improve the quality and efficiency of sorting.Further verified the feasibility and superiority of fruit sorting system based on machine vision.