A 3D vision-based method for turbine blade crown cladding area recognition and automatic path planning

Turbine blades are critical aero-engine components whose crowns frequently require wear-resistant cladding. However, manual operations are hindered by small target areas and stringent precision requirements, while traditional robotic teach-playback methods suffer from inherently low efficiency. To eliminate the reliance on manual programming, this study proposes a novel, automated 3D vision-based framework for precise cladding area extraction and path planning on turbine blade crowns. The methodology introduces an adaptive Genetic Algorithm-Particle Swarm Optimization (GA-APSO) equipped with a geometry-tailored fitness function to mitigate local optima during coarse point cloud registration. To ensure extreme boundary fidelity, a Robust Edge-Constrained Iterative Closest Point (RE-ICP) algorithm is developed for fine registration, enabling accurate area extraction via the Convex Hull algorithm. Subsequently, a comprehensive path planning pipeline is established. By integrating Principal Component Analysis (PCA) for spatial segmentation and posture determination with Singular Value Decomposition (SVD) for path segment fitting, the system autonomously synthesizes continuous zigzag trajectories. Experimental evaluations demonstrate that the proposed framework achieves high-precision spatial alignment and robust, teaching-free path planning. Specifically, the algorithm yields an average Root Mean Square Error (RMSE) of 0.1960 mm and a total execution time of 5789 ms, significantly outperforming alternative methods. This performance satisfies the practical requirements for precision part cladding.