Grading Research Of Prickly Pear Fruit Based On Deep Learning

Rosa roxburghii Tratt fruit is rich in nutrients and has many functions.It is a typical food-medicine fruit.The planting area of Rosa roxburghii Tratt fruit trees in my country and the annual output of Rosa roxburghii Tratt fruit are in the forefront of the world.Most of the Rosa roxburghii Tratt fruit in my country is mainly used for fresh food,and a small part of the Rosa roxburghii Tratt fruit is used to make dried fruit products.This consumption mode is conducive to promoting the development of the grading of Rosa roxburghii Tratt fruit in the direction of mechanical automation.Nowadays,most of the grading methods of Rosa roxburghii Tratt fruits in my country mainly rely on manual grading.In order to realize agricultural mechanization and agricultural modernization,the grading method of Rosa roxburghii Tratt fruit should be gradually changed from manual grading to automatic mechanical grading to realize its automatic and mechanized rapid classification.Therefore,it is necessary to carry out deep learning-based grading research on Rosa roxburghii Tratt fruit.This research combines deep learning technology,mechanical system,and control system to lay the foundation for the realization of deep learning-based Rosa roxburghii Tratt fruit grading.The main contents of this research are as follows:(1)Rosa roxburghii Tratt fruit grading is based on formulation and measurement of related parameters.Considering the color and the quality of the Rosa roxburghii Tratt,the Rosa roxburghii Tratt fruit is divided into six grades.The size,weight,sliding friction force with rubber material,sliding friction factor with rubber material and other related parameters of Rosa roxburghii Tratt fruit are measured to provide basic data for the design of intelligent grading device for Rosa roxburghii Tratt fruit.Through measurement,the size of Rosa roxburghii Tratt fruit: the range of longitudinal diameter is 28.91 mm～46.83 mm,the range of transverse diameter is 29.90 mm～43.72 mm;the weight range is 10.04 g～26.80 g;the sliding friction force range with rubber material is 0.07 N～0.14 N;The sliding friction factor with rubber material is 0.65.(2)Improved image processing algorithm of Rosa roxburghii Tratt pear fruit.In order to achieve real-time grading of Rosa roxburghii Tratt fruits,this paper improves the original Retina Net target detection algorithm.Improve the original Retina Net target detection algorithm by improving the bias formula of the Focal loss loss function in the Retina Net framework and using the dimensional clustering algorithm to find the better size of the Anchor.Through improvement,compared with the original Retina Net target detection algorithm,this paper improves the Retina Net target detection algorithm:(1)The average recognition accuracy rate is 94.86%;an increase of 1.80%.(2)The number recognition rate is 64.80%;when the recognition accuracy rate is more than 90%,an increase of 11.23%.(3)The detection time of a single Rosa roxburghii Tratt fruit is 57.91 ms;it is reduced by 3.08 ms,and it is reduced by 5.05%.(3)Design of mechanical system for grading research of Rosa roxburghii Tratt fruit.(1)Design of Rosa roxburghii Tratt fruit feeding mechanism: In order to realize single fruit feeding of Rosa roxburghii Tratt fruit,this device is mainly composed of three parts: fruit aging table,inclined cutting surface and fruit sorting machine.(2)Design of Rosa roxburghii Tratt fruit conveying mechanism: In order to Rosa roxburghii Tratttate the acquisition of surface information of Rosa roxburghii Tratt fruit,the shape of the double cone roller is designed with reference to the parameters of Rosa roxburghii Tratt fruit shape.(3)Design of Rosa roxburghii Tratt fruit grading mechanism: The device is mainly composed of three parts,such as a stepping motor,a shift lever,and a rubber sleeve,in order to realize the stable and rapid grading of Rosa roxburghii Tratt fruit.This design enables two levels of classification to be completed when only one classification mechanism is installed,which can shorten the number of classification mechanisms and realize multi-level classification processing.(4)Design of Rosa roxburghii Tratt fruit image acquisition device:This device is composed of industrial camera,lens,sensor,light source,light source box,etc.in order to obtain information on the surface of Rosa roxburghii Tratt fruit on the conveying line.(4)The design of the control system for the classification of Rosa roxburghii Tratt fruit.Design the control system plan,use the control system to integrate the image processing system and the mechanical system to realize the grading and unloading process of the Rosa roxburghii Tratt fruit.