飞机蒙皮等航空航天薄壁零件常采用化铣刻型方法完成表面复杂形状图案加工，对其加工图案进行精确检测是保证薄壁零件加工质量的关键环节。但是，受自身重力、夹紧力等因素影响，这类薄壁零件在检测过程中易发生弯曲变形，导致检测结果不稳定。常规方法往往需定制完全贴合曲面形状的夹具对零件进行固定，成本高且效率低，难以满足工业场景的快速检测需求。提出一种基于点云降维的飞机蒙皮化铣刻型轮廓检测方法，创新性地将三维曲面轮廓降维至二维平面，克服零件变形，实现复杂曲面图案的高精度检测。首先，获取蒙皮曲面的CAD标准轮廓作为理想数据，基于结构光技术获取实际零件的图案轮廓点云作为初始数据；然后，基于测量轮廓模型和CAD模型计算测地距离矩阵；最后，基于测地距离矩阵将测量轮廓和标准轮廓同步降维至二维平面，实现加工轮廓图案的误差分析。实验结果表明，所提方法可以实现无夹具的飞机蒙皮化铣刻型图案检测，精度达0.039 mm。

Aerospace thin-walled components， such as aircraft skins， commonly use chemical milling techniques to create complex surface patterns. Therefore， accurately inspecting the processed patterns is crucial in ensuring the machining quality of the thin-walled parts. However， due to factors such as gravity and clamping forces， these parts are susceptible to bending and deformation during the inspection process， leading to inconsistent detection results. Traditional methods often rely on custom fixtures that conform to the curved surface of the parts， which is both costly and inefficient， and may not meet the rapid detection demands of industrial environments. To address these challenges， this paper proposes a novel chemical milling contour detection method for aircraft skins based on point cloud dimensionality reduction. This method innovatively reduces the dimensionality of the 3D surface contour to a 2D plane， thereby mitigating part deformation and enabling high-precision detection of complex surface patterns. Initially， the CAD standard contour of the skin surface is obtained as the reference data， and the pattern contour point cloud of the actual part is captured using structured light technology as the initial data. The geodesic distance matrix is then calculated based on the measured contour model and the CAD model. Finally， both the measured and standard contours are synchronously reduced to a 2D plane using the geodesic distance matrix， facilitating error analysis of the processed contour pattern. Experimental results demonstrate that the proposed method can achieve fixture-free detection of aircraft skin milling patterns with an accuracy of 0.039 mm.