An energy-efficient classification system for peach ripeness using YOLOv4 and flexible piezoelectric sensor

Abstract

Peach is a kind of popular fruit with significant economic value, but it is commonly sorted through manual labor on sorting lines, which can negatively affect the economic efficiency. To address this issue, this paper proposes a system for classifying different ripeness of peaches based on their visual and tactile characteristics. This system consists of two stages. The first stage utilizes YOLOv4 as the core model for preliminary visual-based classification of fruit ripeness. The second stage proposes a method centered on flexible piezoelectric sensors to more deeply classify the fruit ripeness based on tactile characteristics. The system classifies peaches into five categories (A1, A2, B1, B2, and R), which can be applied to the peach sorting line. The mAP and meanIoU are used as evaluation parameters for the visual part and TPA test is used to validate the accuracy of the tactile part. The results of the experiments show that mAP value reaches 0.9304, meanIoU reaches 0.9454, and the accuracy of the tactile part reaches 92.22%. The results indicate that visual and tactile characteristics can accurately classify peach ripeness using YOLOv4 and flexible piezoelectric sensor. The proposed system completes the entire classification with lower cost and less power consumption, providing potential ways to increase productivity and automate the sorting line.