Moisture Content Detection in Tea Withering Process Based on SG-YOLOv8n Algorithm
PI Minghuan
DONG Xiaojie
YANG Yan
LIU Zhi
CHEN Daiming
HUANG Kun
Abstract：To address the problem of destructiveness and long detection time of traditional moisture content detection methods,a tea withering moisture content detection algorithm based on spatial channel and group shuffle-you only look once version 8 nano(SG-YOLOv8n)was proposed.Firstly,a dataset of tea leaf images with varying moisture content during the withering process was constructed.To enhance the perception capabilities of the algorithm,the convolutional block attention module(CBAM)was integrated into the algorithm,which improved performance without increasing network complexity by focusing on important spatial and channel features.Furthermore,to enhance mean average precision and floating-point speed,the group shuffle convolution(GSConv)was used to replace the standard convolution of the neck network.In the backbone network,the convolution to fully connected-spatial and channel reconstruction convolution(C2f-SCConv)was introduced as a substitute for the original convolution to fully connected(C2f)module.The results showed that,compared to the original YOLOv8n algorithm,the SG-YOLOv8n algorithm improved mean average precision and precision by 4.6％and 5.7％,respectively,and the detection speed reached 156.0 frames/s.The algorithm can improve the detection precision of the moisture content in the process of fresh tea leaf withering and realize real-time detection,which can meet the requirements of edge computing equipment.
Keywords：YOLOv8nmoisture contentreal-time detectionteaCBAMGSConvC2f-SCConv
Publication Date：2024-12-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：6( 458-463 )