Development Of Automatic Sorting Device For Dried Jujube Appearance Quality Based On Machine Vision

Dried jujube is originated in the Yellow River and other areas in China,which has high edible value,nutritional value,medicinal value and economic value.As a unique fruit in China,its output exceeds 90% of the world's total output.In recent years,the demand for raw jujube and deep processing industry has continued to grow rapidly.Appearance quality grading technology of the dried jujube is one of the key technologies to improve its quality and economic value.Traditional grading devices can only achieve on-line classification of size,but can not achieve on-line classification of surface diseases and cracks.Therefore,the design of an automatic sorting device for jujube appearance quality(disease,crack,size)based on machine vision with high classification efficiency,classification accuracy,stability and reliability is of great significance to the development of dried jujube industry.This thesis takes “Hetian Junzao” in Xinjiang as the research object,and the main research contents and conclusions are as follows:(1)Overall design of automatic sorting device for jujube appearance quality based on machine vision.According to the basic technical parameters(quality,vertical diameter,transverse diameter)of jujube and the corresponding national standards,the classification basis of jujube grades was determined.In addition,according to the different functions,the device was divided into feeding system,detection system,grading execution system and software system.Based on the overall design principle and design environment,the overall structure was designed.(2)Design of mechanical part and control part of low-position single-row automatic feeding system.Firstly,the overall structure was designed,and the special structure and size design of its core components(wavy conveyor belt and hopper)were also carried out through experiments.Then,according to the designed mechanical structure,the speed of DC gear reducer was controlled by raspberry pie and L298 N dual-channel DC motor driver with PWM speed control module.Therefore,the feeding speed of the automatic feeding system was controlled and the device realized the three-speed operation(60/min,90/min,120/min).(3)Design of mechanical part,control part and detection part of appearance quality automatic detection system based on machine vision.Firstly,the overall structure was designed,and the core unit(light source unit and friction plate unit)was specially designed through experiments.Then,according to the designed mechanical structure,the raspberry pie and laser sensor were used to control two industrial cameras to collect the front and back images of dried jujube.Finally,the collected jujube images were processed by Python and OpenCV,this thesis presented the detection method of jujube disease and crack according to the length-width relationship of the minimum circumscribed revolving rectangle of the defect contour of jujube surface,and the detection method of jujube size according to the number of pixels in the minimal circumscribed polygon of jube contour.(4)Design of mechanical part and control part and optimum design of classified channelof push-pull automatic grading execution system.Firstly,the overall structure was designed,and the layout design of its core components(push-pull electromagnet)was carried out through experiments.Then,according to the designed mechanical structure,the raspberry pie was used to send signals to the electromagnetic relay to control the opening and closing of the circuit through a delay program,so as to control the push and pull action of the push-pull electromagnet and complete the unloading action of the dried jujube.Finally,the structure of grading passage was optimized.When the installation angle of grading baffle was 60°and its installation distance was 10 cm,the mechanical damage of jujube reached the minimum in the unloading process.(5)Design of software system.According to the design requirements of feeding system,detection system,and automatic grading execution system,software operating interface was designed,including the working speed module,the camera module,the image collection module,the image processing module,the image detection results display and statistics module and the grading execution module.It realized automatic feeding,automatic detection and automatic unloading of jujube,and finally could complete the automatic sorting based on machine vision according to jube appearance quality.(6)After the design of mechanical system and software system,the working performance of automatic sorting device for jujube appearance quality based on machine vision was tested online.The test results showed that the grading accuracy of the device for jujube size grade was higher,while the grading accuracy for surface disease and crack was lower.When the grading speed is 60/min,the overall classification accuracy is 86.6%,the grading accuracy of jujube is the highest,the size grading accuracy reaches 88.2%,and the grading accuracies of disease and crack are 80.0% and 77.5% respectively.However,the grading accuracy would decrease with the increase of the grading speed,which needed further optimization and improvement.