基于多特征融合图像分割的焊缝检测焊接系统

Welding System Based on Seam Detection Using Image Segmentation with Multi-Feature Fusion

摘要:

针对工业生产中焊缝检测系统存在易受光照影响、依赖人工和碰撞事故风险高等问题，开发了一套基于多特征融合分割算法的自动化焊缝检测系统；系统采用激光深度相机获取焊缝深度图像，结合多特征融合与斜矩包络线拟合的图像分割算法解决了表面不平整导致的分割难题；通过标定算法，确保了焊缝位置坐标和宽度的准确获取；提出了基于定向包围盒(OBB)的避障检测算法，防止焊接过程中的碰撞；利用机械臂建立了测试系统。经实验测试该系统在检测焊缝宽度和定位方面有效，平均误差分别为0.3029 mm和0.3393 mm，达到工业应用标准；该系统优化了焊接路径和碰撞检测，有效提升了焊接质量与生产效率。

Abstract:

In response to the issues prevalent in industrial production's weld seam inspection systems, such as susceptibility to lighting conditions, reliance on manual operation, and high risk of collision accidents, an automated weld seam detection system based on a multi-feature fusion segmentation algorithm has been developed. The system employs a laser depth camera to capture depth images of the weld seams. It combines multi-feature fusion with skew moment envelope fitting in the image segmentation algorithm to overcome the challenges posed by uneven surfaces. A calibration algorithm ensures the accurate acquisition of weld seam coordinates and width. Additionally, a collision avoidance detection algorithm based on Oriented Bounding Boxes (OBB) is proposed to prevent collisions during the welding process. The system was tested using a robotic arm setup. Experimental results demonstrate the system's effectiveness in detecting weld seam width and positioning, with an average error of 0.3029 mm and 0.3393 mm, respectively, meeting industrial application standards. This system optimizes the welding path and collision detection, significantly enhancing welding quality and production efficiency.