| 基于YOLOv7改进模型的轮胎损伤检测算法的研究 |
| Research on Tire Damage Detection Algorithm Based on Improved YOLOv7 Model |
| 投稿时间：2023-07-03 修订日期：2023-07-03 |
| DOI：10.12136/j.issn.1000-890X.2025.03.0226 |
| 中文关键词: 轮胎 YOLOv7模型 损伤检测 注意力机制 损失函数 深度学习技术 |
| 英文关键词: tire YOLOv7 model damage detection attention mechanism loss function deep learning technology |
| 基金项目:山东省重点研发计划（公益类专项）（2019GGX101020）；青岛理工大学研究生优秀教材建设项目（Y052022-005） |
|
| 摘要点击次数: 161 |
| 全文下载次数: 66 |
| 中文摘要: |
| 提出基于YOLOv7改进模型的轮胎损伤检测模型（简称TD-YOLO模型）。TD-YOLO模型融合了无参数的卷积神经网络注意力机制（SimAM），增强了模型的特征学习能力；改进了损失函数，增强了对鼓包、裂纹、嵌入异物检测的准确性和敏感性；在网络结构输出端引入空间到深度（SPD）模块层构建了新的卷积神经网络（CNN）模块，可以提升小目标损伤检测的准确率；TD-YOLO模型的平均检测精度为0.916，比YOLOv7模型增大了0.069。TD-YOLO模型的综合性能好，其具有较高的推广应用价值。 |
| 英文摘要: |
| The tire damage detection model based on YOLOv7 improved model（referred to as TD-YOLO model） was proposed.The TD-YOLO model integrated a parameter free convolutional neural network attention mechanism（SimAM） to enhance the model’s feature learning ability，improved the loss function to enhance the accuracy and sensitivity of detecting bulges，cracks and embedded foreign objects，and introduced space to deep（SPD） module layers at the output of the network structure to construct new convolutional neural network（CNN） module to improve the accuracy of small target damage detection.The average detection accuracy of the TD-YOLO model was 0.916，which was 0.069 larger than the YOLOv7 model.The comprehensive performance of the TD-YOLO model was good，it had good promotion and application values. |
|
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
