Research And Implementation Of Vegetables And Fruits Object Detection Algorithm On Mobile Devices

Vegetables and fruits are an essential and important source of nutrition for human daily life.The application of an automatic vegetables and fruits checkout system in self-service supermarkets will greatly reduce the pressure on cashiers and will also improve the consumers' shopping experience.In order to achieve an intelligent vegetables and fruits checkout system,one of the core technologies is to perform automatic object detection on vegetables and fruits.Because the variety of fruits and vegetables in supermarkets is so diverse and different in shape,it is very challenging to detect them automatically.At the same time,in order to make it easier for people to detect vegetables and fruits using mobile devices in self-service supermarkets,research on mobile-oriented vegetables and fruits object detection algorithms has become very urgent.In order to solve this difficult problem,this paper proposes a mobile-oriented lightweight object detection algorithm based on deep learning.The main research content is divided into the following three aspects:(1)For the current object detection field,there is no special vegetables and fruits dataset.We first spend a lot of energy to build a large-scale vegetables and fruits dataset,and for each object,the positions are carefully labeled,which lays the data foundation for the research of deep learning algorithm;(2)A deep neural network is proposed which combines high-level semantic information and low-level detail information of vegetables and fruits.(3)In order to overcome the problems caused by the imbalance of training samples between object detection algorithm dataset,We used data amplification and transfer learning to reduce the overfitting problem of the model and improve the generalization ability of the model.Based on the algorithm proposed in this paper,we set up a mobile end vegetables and fruits object detection system.We have performed a lot of tests on the function and performance of the system,and transplanted the proposed vegetables and fruits object detection algorithm on multiple different mobile devices.A large number of experimental results show that our proposed algorithm has achieved good results in term of detection accuracy and object speed.