A measurement method with dynamic feedforward compensation for 3D profile of aero-engine pipelines

The 3D contour measurement of aero-engine pipelines is challenged by complex spatial structures and severe occlusion. Traditional dynamic visual measurement methods are prone to image blurring and positioning inaccuracies when the scanner undergoes simultaneous translation and rotation. To address the problem, this paper proposes a measurement method with dynamic feedforward compensation, which mitigates observation quality degradation caused by motion coupling through the integration of motion prediction and dynamic feedforward compensation. The method constructs a compound motion model employing a multidimensional state vector to predict the target's multi-degree-of-freedom motion within the Position-Attitude Kalman Filter (PAKF) framework in real time. The prediction results guide the modeling of image degradation based on the Point Spread Function (PSF), and adaptive deconvolution is applied to improve feature points localization accuracy. In the observation update stage, prediction uncertainty is incorporated into a spatially correlated observation noise model, with a covariance matrix characterizing the error correlations among different targets. This enables optimal fusion of motion prediction and visual observations within a probabilistic framework. By combining predictive modeling with image-level compensation, the method achieves robust and high-precision attitude estimation and 3D point cloud localization under dynamic conditions. Experiments demonstrate that the proposed method significantly enhances measurement accuracy: the average position error is reduced by approximately 23.42% in simulation compared to the motion prediction without attitude, and deviations in actual dynamic measurements remain within the range of 0.0078 mm-0.0082 mm.