【摘要】 目的 提升柔性印刷电路（FPC）软排线缺陷图像的分割与检测精度、效率，解决传统方法在低对比度、强干扰及细微缺陷图像中的分割模糊和检测误差等问题，提出一种高鲁棒性、高效率的包装缺陷处理方法。方法 构建基于并行动态角色记忆灰狼优化算法（PDM-GWO）的图像分割和缺陷检测方法。通过动态角色分配和历史位置记忆提升优化能力，引入主从并行架构，提高计算效率；分割阶段采用PDM-GWO优化多阈值策略提取清晰边缘；在检测阶段，基于边缘检测获取排线坐标，融合RANSAC拟合提取几何特征，结合Z-score统计分析，实现多类缺陷的识别。结果 多组图像实验证明，该方法在PSNR、SSIM、IoU等3项指标上的平均值为22.42 dB、0.964、0.933，均优于标准GWO和典型改进型算法。在缺陷检测方面，平均检测精度达到0.990 6，处理速度为9.63帧/s，优于YOLOv9、Faster-RCNN等主流方法。结论 所提方法在图像分割质量、检测准确率、运行效率等方面均展现出显著优势，适用于包装自动线复杂工况下的微小缺陷检测，具备良好的工程实用性和推广价值。更多还原

【Abstract】 The work aims to propose a high-robustness and high-efficiency packaging defect processing method to address the challenges of segmentation blur and detection errors in traditional methods when dealing with low-contrast, highly interfered, and subtle defect images of flexible printed circuit(FPC) ribbon cables, and to enhance the accuracy and efficiency of FPC defect image segmentation and detection. An image segmentation and defect detection method was constructed based on a Parallel Dynamic Role Memory Grey Wolf Optimization(PDM-GWO) algorithm. The optimization capabilities were improved through dynamic role assignment and historical position memory, while introducing a master-slave parallel architecture to enhance computational efficiency. In the segmentation phase, a PDM-GWO optimized multi-threshold strategy was adopted to extract clear edges. In the detection stage, the coordinates of the wiring were obtained based on edge detection, the geometric features were extracted by integrating RANSAC fitting, and the Z-score statistical analysis was combined to realize the recognition of multiple types of defects. Extensive image experiments demonstrated that the proposed method achieved average values of 22.42 dB for PSNR, 0.964 for SSIM, and 0.933 for IoU, all surpassing standard GWO and typical improved algorithms. For defect detection, the average mean average precision reached 0.990 6, with a processing speed of 9.63 frames per second, outperforming mainstream methods such as YOLOv9 and Faster-RCNN. The proposed method exhibits significant advantages in image segmentation quality, detection accuracy, and operational efficiency. It is well-suited for micro-defect detection in complex industrial conditions on automated packaging lines, demonstrating excellent engineering practicality and promising generalization value.更多还原