基于机器视觉的木质勺子表面缺陷识别方法  

Surface Defect Recognition Method of Wooden Spoon Based on Machine Vision

作　　者：孙思凡 李春伟[1] SUN Si-fan;LI Chun-wei(College of Home and Art Design,Northeast Forestry University,Harbin Heilongjiang 150040,China)

机构地区：[1]东北林业大学家居与艺术设计学院,黑龙江哈尔滨150040

出　　处：《林业机械与木工设备》2025年第2期18-24,共7页Forestry Machinery & Woodworking Equipment

基　　金：黑龙江省自然科学基金项目(LH2019E001);东北林业大学院级大学生创新训练项目资助(202310225649)。

摘　　要：针对人工挑选中缺陷木质勺子所存在的劳动量大、效率低、准确率不足等问题,为实现木质勺子表面缺陷识别智能化、自动化,提高木质勺子表面缺陷识别效率与准确率,提出了一种基于改进YOLOv8的木质勺子表面缺陷识别方法。以YOLOv8模型为基础,将原有特征金字塔部分改为BiFPN重复双向特征金字塔结构,以融合更多的特征,在Neck中加入全局注意力机制,以更好地捕捉空间高度与宽度和通道三者之间显著的特征,将原有模型中CIoU损失函数更换为SIoU损失函数,以提高检测精度和训练速度。实验结果表明,使用改进后的模型对拍摄的木质勺子图片中表面缺陷进行识别的mAP达到了91.7%,较原有YOLOv8模型提高了3.3%,优于其他目标检测算法,可以为木质勺子表面缺陷识别智能化、自动化提供参考。Aiming at the problems such as large amount of labor,low efficiency and insufficient accuracy in manual selection of defective wooden spoons,an improved YOLOv8 based wooden spoon surface defect recognition method was proposed to realize intelligent and automatic surface defect recognition of wooden spoons and improve the efficiency and accuracy of surface defect recognition of wooden spoons.Based on YOLOv8 model,the original feature pyramid is changed into BiFPN repeated bidirectional feature pyramid structure to integrate more features,and the global attention mechanism is added to Neck to better capture the significant features between the height,width and channel of the space.The CIoU loss function in the original model is replaced with SIoU loss function.To improve the detection accuracy and training speed.The experimental results show that the mAP of surface defect recognition in the picture of wooden spoon taken by the improved model reaches 91.7%,which is 3.3%higher than that of the original YOLOv8 model,superior to other target detection algorithms,and can provide a reference for intelligent and automatic surface defect recognition of wooden spoon.

关 键 词：木质勺子 缺陷识别 YOLOv8 智能化 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]