基于YOLOv8的轮毂检测算法研究  

Research on Wheel Hub Detection Algorithm Based on YOLOv8

作　　者：贠甜甜 魏阳 何冰[1,2] 

机构地区：[1]渭南师范学院物理与电气工程学院,陕西 渭南 [2]渭南师范学院X射线成像与检测陕西省高校工程研究中心,陕西 渭南

出　　处：《图像与信号处理》2025年第1期34-44,共11页Journal of Image and Signal Processing

基　　金：陕西省教育厅重点研究计划项目;陕西省大学生创新创业训练计划项目(项目编号:S202310723035);渭南师范学院教学改革项目(项目编号:JG202135)。

摘　　要：鉴于传统轮毂分类检测流程中存在的劳动重复度高、人力成本高企及生产效率低下等挑战,文章提出了一种应用YOLOv8网络的汽车轮毂自动分类系统。该系统的工作流程包括:首先,收集丰富的汽车轮毂X光图像,构建一个囊括多种轮毂类型的综合性数据集;接着,利用YOLOv8算法对该数据集实施训练,以生成一个能够精确分辨轮毂种类的模型。在模型训练阶段,针对轮毂分类的具体特性,对YOLOv8算法进行了改进,引入了Focal Loss作为损失函数,从而有效缓解了正负样本不均衡的问题,进一步提升了轮毂分类的精确度。在实验验证环节,文章在不同噪声干扰和光照条件下对轮毂图像进行了测试。实验结果显示,该系统能迅速且准确地识别出各类轮毂,平均识别准确率高达98.43%,展现出了卓越的分类精度和强大的鲁棒性。In response to the challenges of high labor repetition, escalating labor costs, and low production efficiency in traditional hub classification and inspection processes, this paper proposes an automatic automobile hub classification system using the YOLOv8 network. The workflow of this system includes: first, collecting extensive X-ray images of automobile hubs to construct a comprehensive dataset encompassing various hub types;next, utilizing the YOLOv8 algorithm to train this dataset to generate a model capable of accurately distinguishing hub types. During the model training phase, improvements were made to the YOLOv8 algorithm based on the specific characteristics of hub classification, with Focal Loss introduced as the loss function, effectively mitigating the issue of imbalance between positive and negative samples and further enhancing the accuracy of hub classification. In the experimental validation stage, hub images were tested under different noise interference and lighting conditions. The experimental results demonstrate that the system can swiftly and accurately identify various types of hubs, wit

关 键 词：轮毂分类 X光轮毂图像 卷积神经网络 YOLOV8算法 Focal Loss函数 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]