A Novel Detection Approach for Quasi-Convex Quadrilateral Aircraft Covers

To address the challenge of rapid and efficient detection of aircraft cover plate target images resembling quadrilaterals, this paper introduces the concept of quasi-convex quadrilaterals and presents a corresponding target image processing and detection approach. Specifically, the definitions, properties, and determination methods of quasi-convex quadrilaterals are meticulously studied and expounded. First, Fourier transform-based noise reduction is applied to contours that meet the definition of quasi-convex quadrilaterals. Then, curvature is used to identify segmentation points and four quasi-straight edges. Subsequently, the least-squares method is employed to fit these four quasi-straight line segments, generating the adjoint quadrilateral of the quasi-convex quadrilateral. The vertices of the adjoint quadrilateral act as control points for distortion correction of the quasi-convex quadrilateral, and template matching is adopted to detect the contour of the target aircraft cover plate. Experimental validation is carried out on sample photos of three subtypes of quasi-convex quadrilateral aircraft cover plate contours. The proposed approach achieves a 98.81% True Positive Rate (TPR), outperforming both the standard Hough Transform (93.25% TPR) and improved variants (95.63% TPR). It maintains a low False Positive Rate (4.37% FPR) while reducing computational time by approximately 26.5% compared with standard HT. Additionally, compared with the lightweight CNN model MobileNet Single Shot MultiBox Detector (MobileNet-SSD), the proposed method exhibits advantages in low-resource scenarios (no GPU, no training dependency) in terms of detection accuracy and inference efficiency. The algorithm's linear time complexity (O(N)) ensures scalability for high-resolution industrial imaging applications, making it particularly suitable for real-time aircraft maintenance systems.