Authors

Abstract

To address the challenges of irregular morphology and the difficulty in rapidly and accurately measuring the area of laser cladding zones during the inspection of agricultural rotary tiller blades, this paper proposes a machine vision-based method for rapid area extraction. A comprehensive processing workflow encompassing image preprocessing, contour extraction, region of interest (ROI) extraction, and pixel integration was established. For region segmentation, an improved Alpha Shapes segmentation algorithm was proposed and compared against conventional Convex Hull and Delaunay triangulation methods. Validation was conducted using 100 rotary tiller blade samples, with electron microscopy manual calibration results serving as reference. Results indicate the improved Alpha Shapes algorithm delivers optimal segmentation accuracy, yielding the smallest absolute area error (−7.40×10-⁷ ± 2.69×10-5)m² and lowest relative error (13.48 ± 8.47)×10-3, with high consistency against microscopic measurements. Compared to conventional manual measurement, the area extraction algorithm proposed in this study offers the advantages of automation, non-contact operation, and high efficiency, meeting the engineering application requirements for laser cladding quality inspection.

Abstract in Chinese

针对农用旋耕刀片检测中激光熔覆区域形态不规则、面积难以快速准确测量的问题，本文提出了一种基于机器视觉的面积快速提取方法。构建了包含图像预处理、轮廓提取、ROI提取与像素积分的完整处理流程。在区域分割时，提出了一种改进的Alpha-Shapes区域分割算法，将其与常用Convex Hull与Delaunay剖分合并两种算法进行对比，试验选取100个旋耕刀片样本进行验证，以电子显微镜人工标定结果为参考。结果表明，改进的Alpha-shape算法在区域分割精度上表现最佳，最终计算面积绝对误差最小（−0.000000740±0.000026906）、相对误差最低（0.013484±0.008470），且与显微镜测量结果高度一致。与传统人工测量相比，本研究提出的面积提取算法具有自动化、非接触和高效的优势，满足激光熔覆质量检测的工程应用需求。