Detection And Characterization Of Surface Defects Of Aircraft Skin Based On 3D Point Cloud

The fuselage skin is an important shape-maintaining component exposed to the exterior of military and civil aircraft,and many defects such as cracks,corrosion,pits and scratches are often caused by factors such as fatigue damage and external environment.These defects destroy the strength and integrity of the aircraft skin structure,and will bring major safety hazards if they cannot be detected and maintained in time.At present,the detection and marking of aircraft skin defects mainly adopts artificial visual detection and marking technology,which has problems such as low use efficiency,high error rate,and limited data utilization,which cannot meet the timeliness requirements of aircraft maintenance.Therefore,the automatic defect detection method based on visual measurement technology has gradually become the key research content of various countries and researchers.However,there are few researches on the complete feature measurement of 3D defects by visual measurement technology,and the accuracy of the measurement results is also lacking.Based on the visual measurement method,this paper divides defects into 2D defects without depth information and 3D defects with depth information.The main researches are 2D defect detection and identification methods,and 3D defect detection,segmentation and feature measurement methods.The specific work is as follows:（1）In the research work of the skin defect detection system design,the overall architecture of the skin defect detection system is designed.Aiming at the problem of 2D image defect detection,an improved local contrasts saliency defect detection method is proposed.First,the saliency value of the defect is obtained by correcting the contrast to the initial image and calculating the local contrast at different scales;Then the brightness component of the original image and the saliency value is fused to obtain the final defect saliency map;Finally,combined with the Canny edge detection operator of adaptive threshold,the defect edge is extracted,and the effective defect edge information is obtained.The experimental results of defect detection for skin crack to demonstrate the feasibility of the image defect detection algorithm.（2）Aiming at the problem of accurate measurement of defect features of 3D point cloud,a defect detection and quantitative characterization method based on 3D point cloud is proposed.First,the whole measurement process is divided into a defect detection stage and a defect feature measurement stage;Then in the defect detection stage,the initial point cloud with huge data volume is processed to remove noise points and reduce the data volume,and use the Moving Least Squares（MLS）method to smooth out some erroneous small feature data caused by measurement errors,and combine the SVD decomposition method to obtain the normal and curvature information of each point.On this basis,the region growing segmentation algorithm is used to obtain the point cloud data of defective and non-defective areas.The experimental results show that the segmentation algorithm can divide the defect area into smaller point cloud clusters,and the defect point cloud is obtained by subtracting the non-defect area point cloud from the global point cloud,so as to achieve the purpose of accurate defect detection and segmentation.（3）For the problem of defect feature measurement after segmentation,the core idea is to calculate the size information of the largest outer contour of the defective area.Firstly,the rotation matrix is constructed according to the background point cloud,and the best fitting function of the background point cloud is constructed by using the global Weighted Least Squares（WLS）fitting method,and the optimal ideal reference surface of the defect is calculated,and the depth information of the defect is solved;Secondly,the 3D defects are projected to the plane,the 2D OBB bounding box of the defect area is established,and the 2D size information of the defects is solved by the PCA algorithm.The experimental results show that the 3D defect detection and quantitative characterization algorithm can effectively obtain the 2D size and depth information of defects.（4）Developed 3D point cloud defect detection system software.In the whole system experiment process,for the 2D defect detection experiment,the 2D image defect detection algorithm was first used to detect defects such as cracks,scratches,holes,etc.,and compared with the classical detection algorithm,the results show that the effectiveness of the 2D defect detection algorithm;then the In and Sc defects are experimentally verified through the open source data set,and the results show that the defect detection success rate of the algorithm in this paper is higher than other traditional detection algorithms.Aiming at the 3D defect detection experiment,a comparative analysis method to verify the accuracy of the 3D defect detection algorithm was first designed.The point cloud data generated by the 3D model with the known defects size was used to evaluate the measurement accuracy of the 3D defect detection algorithm,and analyze the influence of setting different parameter thresholds on the measurement results and the error source of the algorithm;Then,for multiple defects in a single point cloud,use the clustering algorithm to divide the defects;Finally detect the defects size of the real pit point cloud data.The experimental results show that the minimum error of the 3D defect detection algorithm in this paper for defect depth is tens of microns,while the minimum detection error of the 2D size is sub-millimeter.