Research on Defect Detection Model for Green Coffee Beans Based on Machine Vision
Ren Jie
Shao Kaiqing
Hu Xin
Zhang Jingna
Liu Xueyuan
Zhu Daigen
Abstract：In order to improve the defect detection accuracy of green coffee bean during sorting process,a coffee bean defect detection model based on improved YOLOv5s was proposed.Firstly,the collected images were iterated and enhanced to create a data set containing different coffee bean defects.Secondly,a small target detection layer was added to the original network structure of YOLOv5s to improve the performance of the algo-rithm in dense scenes;the coordinate attention(CA)mechanism was introduced into the network,and the fea-ture expression ability of the model was enhanced;the activation function was replaced by Mish function to promote better transmission of information to the neural network,thus improved the accuracy and generalization performance of the model.The experimental results showed that compared with the original model,the im-proved model increased the accuracy and average accuracy by 6.8 and 1.1 percentage points,respectively.The average confidence could also be improved under complex lighting conditions and reached to more than 0.90.The improved model was suitable for detecting unsorted green coffee beans and could provide a theoretical ba-sis for future mechanized sorting of coffee beans.
Keywords：Green coffee beansDefect detectionMachine visionDeep learningCA mechanismMi-sh function
Publication Date：2025-07-30
Online Publishing Date：2026-05-22(First online date of this platform, not the publication date of the document)
Pages：7( 152-158 )