Research On Fruit Sorting System Based On Machine Vision

Fruit sorting is a very necessary step in fruit production and transportation.At present,the fruit sorting in traditional agriculture mainly depends on manual completion.However,this method has disadvantages such as large workload,low efficiency,and slow speed.Moreover,fruit sorting is a long-term repetitive task.If people simply rely on increasing labor to increase their sorting speed,it will enhance the production cost and affect the fruit production and the competitiveness of the fruit in the market.This paper has designed a machine vision system that is specially used for fruit sorting.The main work of system design is as follows:1.In order to obtain high-quality fruit images,the hardware design part of the system has selected the camera,lens,and lighting program of the machine vision system one by one.2.In the image preprocessing stage,the system uses the median filter and binomial filter in the spatial filtering technique to eliminate random noise such as salt and pepper noise and Gaussian noise in the image.At the same time,image enhancement techniques are used to enhance the edges of the image blur.3.In the fruit classification process,the supervised clustering Gaussian mixture model algorithm in machine learning is studied.Through the clustering of fruit color and regional characteristics,the corresponding fruit classifier is trained to classify the unknown fruit.According to statistics,the classification accuracy of the system is maintained at about 98%.4.In the fruit grading process,the results of the horizontal diameter measurement of the fruits and the results of the color classification of the Gaussian mixture model were used as the standards for judging the level.The fruit is divided into large,medium,and small grades by the actual length of the horizontal diameter of the fruit,and the maturity is determined by the information on the surface color of the fruit.5.In the fruit sorting posture calculation process,the pose calculation work of the monocular camera-guided robot is studied.The camera's hand-eye calibration is performed by using the Zhang Zhengyou calibration method,and the internal and external parameters of the camera are obtained,and then used.The calibration data calculates the exact world coordinates of the fruit,and prepares the robot's robot for picking and sorting.