Structure and Depth-Based Blade Detection Algorithm for Multistage Aeroengine Fan Blade Detection

Fast and efficient in-situ inspection of aircraft engine blades is crucial for maintaining the safety and reliability of aircraft engines. The automation and intelligence of aero-engine blade inspection through robotic systems, replacing manual labor, is a key trend for future development, with blade perception technology being one of the crucial components. This paper addresses the challenges of multi-stage blade detection within the complex air intake environment and analyzes the inherent characteristics of blade-like targets. Building on the geometric structure and leveraging prior knowledge, we propose the Structure and Depth-Based Blade Detection (SDBD) algorithm for precise engine blade detection. The method begins with an enhanced Yolo model to locate the center hub of the blades, followed by the Baseline-Blade Guided Coarse Detection Algorithm for initial blade identification. Finally, the Descriptor-Free Line Segment Matching Algorithm is applied to accurately localize the blade edges. Experimental validation in a simulated engine propulsion system confirms the effectiveness of the proposed approach. The SDBD algorithm reduces the dependence on texture information by focusing on the inherent features of blade-like objects. It exhibits strong robustness and generalization, performing effectively even in scenarios where blades are partially obscured or the image quality is compromised. The results demonstrate that under different shooting angles and lighting conditions, all the center hub positioning deviation errors are within 1 pixel, the parallelism errors of blades positioning are below 0.7° and the distance errors remain under 4 pixels.