基于目标检测技术的主船体结构外部裂纹检测方法

Detection Method for External Cracks in the Main Hull Structure Based on Object Detection Technology

设计一种基于目标检测技术的主船体结构外部裂纹自动检测模型,该模型以YOLOv5为框架,首先基于路面开裂等公开数据集中已有的各类表面裂纹样本对模型进行预训练,再利用主船体结构外部裂纹数据集进行迁移训练。最终模型在验证集上的测试情况:当交并比大于0.5时的平均精度均值为0.92,证明了使用目标检测技术检测船体结构裂纹的可行性。利用该模型对无人机、机器人所采集的实时图像开展主船体结构外部裂纹检测,可提高验船师检验效率并降低检验成本。

An automatic detection model was designed for external cracks in the main hull structure based on object detection technology.The model used YOLOv5 as the framework and was pre-trained using various surface crack samples from publicly available datasets such as road cracks,and then fine-tuned using a dataset of external cracks in the main hull structure.The model's performance on the validation set showed that the mean average precision(AP) was 0.92 when the intersection over union(IoU) was greater than 0.5,demonstrating the feasibility of using object detection technology to detect cracks in ship structures.Using this model to detect external cracks in the main hull structure from real-time images captured by drones or robots can improve the efficiency of ship inspectors and reduce inspection costs.