基于改进YOLOv8的船板焊缝质量检测方法

Improved YOLOv8-based quality inspection method for ship plate welds

针对目前船舶制造工业之中船板焊缝质量检测精度不高、模型计算量大的问题.提出基于改进YOLOv8模型的高精度、轻量化自动检测方法(REF_YOLOv8).该方法引入了由Res2Net和EMA注意机制整合而成的REM模块,增强了感受野的同时减少了计算开销.同时,利用SIoU融合Focal机制形成Focal-SIoU机制,取代原有的CIoU,优化目标重要性评估与预测框位置判断.此外,使用GhostConv取代了传统的Conv操作,在不影响性能的情况下降低了计算成本.实验结果表明,改进后模型相比原始YOLOv8,mAP@0.5提高3.3%,模型的计算量减少31.5%,参数量减少了36.7%,证明了REF_YOLOv8模型在船板焊缝识别领域的高精度和高效性.

Aiming at the current problems of low accuracy and large modeling computation for the quality inspection of ship plate welds in the shipbuilding industry.A high-precision and lightweight automatic inspection method based on the improved YOLOv8 model(REF_YOLOv8)is proposed.The method introduces a REM module integrated by Res2Net and EMA attention mechanism,which enhances the sensory field while reducing the computational overhead.At the same time,SIoU is integrated with Focal mechanism to form Focal-SIoU mechanism,which replaces the original CIoU to optimize the target importance assessment and prediction frame location judgment.In addition,GhostConv is used to replace the traditional Conv operation,which reduces the computational cost without affecting the performance.The experimental results show that the improved model improves the mAP@0.5 by 3.3%,reduces the computational amount of the model by 31.5%,and reduces the amount of parameters by 36.7%compared with the original YOLOv8,which proves the high accuracy and efficiency of the REF_YOLOv8 model in the field of ship plate weld recognition.