基于YOLOX-MSBP机器视觉技术的生菜品质预测  

Research on Lettuce Quality Prediction Based on YOLOX-MSBP Machine Vision Technology

作　　者：李珈慧 姬长英[2] 胡鹏昱[1] 王妍[1] 赵文旻[1] LI Jiahui;JI Changying;HU Pengyu;WANG Yan;ZHAO Wenmin(Shanghai Vocational College of Agriculture and Forestry,Shanghai 201699,China;School of Engineering,Nanjing Agricultural University,Nanjing 210095,China)

机构地区：[1]上海农林职业技术学院,上海201699 [2]南京农业大学工学院,南京210095

出　　处：《机械设计与研究》2024年第5期211-215,222,共6页Machine Design And Research

基　　金：上海市科委“科技创新行动计划”农业科技领域项目(20392001300);国家重点研发计划项目(2021YFD1900800)。

摘　　要：针对传统机器视觉方法如YOLO在农产品质量检测中效率低下及准确性不足的问题,提出一种基于YOLOX-MSBP的机器视觉技术检测方法。YOLOX-MSBP检测方法旨在开发一种改进的机器视觉模型,以提高生菜品质的预测精度和处理速度。YOLOX-MSBP检测方法采用了Micro-branch检测分支、优化的主干网络和特征融合网络,并结合交叉验证技术对模型进行训练和优化。YOLOX-MSBP检测方法首先对YOLOX-MSBP模型进行了详细的网络结构设计,重点优化了特征提取和融合过程。通过对大规模数据集的训练和测试,模型在生菜品质预测任务上达到了超过95%的准确率。实验结果表明,该模型在提高预测速度和准确性方面均优于现有的机器视觉方法。基于YOLOX-MSBP的模型在生菜品质预测中表现出较高的性能,可为农业生产提供更有效的品质评估工具。建议未来研究可以将此模型应用于更多类型的农产品,以推广其在农业自动化检测中的应用。In response to the inefficiencies and lack of accuracy of traditional machine vision methods like YOLO in agricultural product quality inspection,a detection method based on YOLOX-MSBP machine vision technology is proposed to enhance the prediction accuracy and processing speed of lettuce quality.The YOLOX-MSBP incorporates a micro-branch detection branch,an optimized backbone network and a feature fusion network,and combines these with cross-validation techniques to train and optimize the model.The method initially involves a detailed design of the network structure of the YOLOX-MSBP model,with a focus on optimizing the feature extraction and fusion processes.Through training and testing on large datasets,the model achieves over 95%accuracy in lettuce quality prediction tasks.Experimental results show that this model surpasses existing machine vision methods in terms of prediction speed and accuracy.The YOLOX-MSBP model demonstrates high performance in predicting lettuce quality,providing a more effective quality assessment tool for agricultural production.It is recommended that future research should apply this model to more types of agricultural products to expand its application in agricultural automation detection.

关 键 词：YOLOX-MSBP 机器视觉 生菜品质 品质预测 

分 类 号：S636.2[农业科学—蔬菜学] TP391.41[农业科学—园艺学]