Research And Application Of Fruits Object Detection Based On Deep Learning

In China,fruit has been an indispensable food in people’s daily life,and the fruit production will increase greatly every year.In recent years,in the offline large and medium-sized supermarkets,fruit checkout is carried out with the help of the cashier,which is time-consuming and laborious.With the development of society and the progress of science and technology,self-checkout system is becoming more and more common.Self-checkout system can not only reduce the pressure of cashiers,improve work efficiency,but also save manpower cost for businesses.In supermarkets,there are many kinds of fruits with different shapes.The self-checkout system needs to identify the fruits quickly and accurately,which is convenient for customers to pay quickly and reduce the waiting time,so as to improve customers’ consumption experience.This paper studies the fruit object detection algorithm,and designs and implements an android fruit self-checkout system.The main work of this paper is as follows:1.Aiming at the problem that the image features extracted by the traditional object detection algorithms are difficult to accurately express the appearance and texture of fruit,this paper uses the deep learning algorithm YOLOv3 to enrich the fruit feature information extracted by the model and improve the accuracy of fruit target detection.2.The object detection algorithm of YOLOv3 is improved.Aiming at the problem that the loss function of the original YOLOv3 network only works when the prediction frame and the target frame overlap,this paper adopts the DIoU loss function,which can directly minimize the distance between the prediction frame and the target frame,making the model perform better.Aiming at the problem of complex calculation and large parameter of feature extraction network of original YOLOv3 network,this paper replaces the feature extraction network with lightweight Mobilenetv3 network to realize the lightweight of network model.3.Based on the above method,an Android fruit self-checkout system is implemented,and the system is tested.The test results show that the system can effectively detect and identify the fruit object,and calculate the total price of fruit to realize self-checkout.