Machine Vision-Based Rivet Detection Method in Civil Aircraft Assembly Scenes

To improve the precision and efficiency of rivet quality inspection in civil aircraft assembly scenes, and overcome limitations of manual inspection, a machine vision-based rivet inspection method is proposed. This method employs an ellipse fitting technique based on the arc-supported least squares method to generate sets of candidate ellipses. It then introduces a fitting degree-based method for matching the edges of the rivet ellipses, thereby achieving rivet positioning and edge extraction. A parallel projection model is established to measure the perpendicularity, diameter, and spacing of the rivets, with pixel equivalent calibration employed to convert pixel measurement values into actual size. The experiment results have demonstrated that the proposed method yields a measurement error of less than 10% for rivet perpendicularity, less than 3% for rivet diameter, and less than 2% for rivet spacing. This offers a novel method and approach for quality control in civil aircraft assembly.