Design Of Weighing System Based On Visual Fruit And Vegetable Recognition

Today,with the rapid development of science and technology,people’s lives are gradually becoming intelligent.The traditional electronic weighing system of fruits and vegetables usually adopts manual weighing method,which has the problems of cumbersome operation and insufficient intelligence.In view of the above problems,this paper combines image recognition technology with sensor technology,and proposes a weighing system based on visual recognition of fruits and vegetables,which realizes the automatic recognition and weighing of fruits and vegetables.In this paper,Faster R-CNN,YOLO and SSD networks are used to compare experiments on self-built fruit and vegetable data sets and public data sets.The accuracy of SSD networks on self-built data sets and public data sets is 96.7 % and 84.3 %,respectively,which is significantly better than the other two networks.In order to further improve the recognition effect,this paper combines the residual network with the dense network,and proposes an improved SSD algorithm based on the dual-module cascade network as the basic feature extraction network.The experimental results show that the accuracy of the improved SSD on the self-built dataset and the public dataset is 98.4 % and 89.8 %,respectively.Compared with the original SSD model,the accuracy is increased by 1.7 % and 5.5 %,respectively.Therefore,the feature extraction ability of the improved SSD algorithm is significantly improved.Finally,the improved SSD algorithm is copied to the Raspberry Pie 3B +,and the hardware circuit of the system is designed with the Raspberry Pie 3B + development board and STM32 microcontroller as the core.The weighing and display functions are completed by Keil software programming.All components are welded and assembled,and a weighing system based on visual fruit and vegetable recognition is designed.The recognition and weighing functions of the system are tested,and the test results show that the average accuracy reaches 97.8 %.The weighing system based on visual fruit and vegetable recognition is designed by improving the classical SSD network.The experimental results show that the improved neural network algorithm improves the recognition ability and enhances the robustness of the system.The whole system meets the expected requirements,runs stably and has certain engineering application value.