Research On Quality Detection Technology Of Aircraft Skin Riveting Based On Image Recognition

The riveting quality inspection process in the assembly process of the aircraft is a key part of the aircraft production.The quality of the assembled rivets directly affects the life and performance of the aircraft.At present,in the production workshop,the detection of skin rivet defects mainly relies on the manual sampling and simple instrument measurement by professional technicians.The detection efficiency is low and the missed detection error is large,which can not meet the requirements of the automatic detection system.In this thesis,the quality inspection technology of skin rivets is studied from the geometrical dimensions and surface defects of skin rivets.The main research work of this paper includes the following parts:(1)The riveting principle of aircraft skin rivets and the existing quality inspection technology are analyzed.The principle and implementation method of target detection algorithm are described in detail.Starting from the basic theory,this paper first introduces the target detection algorithm based on deep learning,and then introduces the circle detection algorithm based on Hough transform.The machine learning classifier is described.The two classification algorithms of support vector machine and logistic regression are introduced,which provides a theoretical basis for the classification and identification of surface defects of skin rivets in the following.(2)This thesis studies and analyzes the measurement method of the center and radius of the skin rivet for the detection of the geometrical dimensions of the aircraft skin rivet.In the random Hough transform circle detection algorithm,the geometric dimension detection of the skin rivet has uncertainty,easy to generate invalid accumulation and large detection error.In this thesis,a circular detection algorithm based on fuzzy set random Hough transform is proposed.By introducing fuzzy set mathematical theory,the sensitivity of random Hough transform circle detection algorithm to noise is reduced,and the circle in the image can be more accurately performed.Positioning significantly improves the detection accuracy of the circle.(3)In this thesis,the surface defects of aircraft skin rivets are detected and the corrosion defects of the surface of the skin rivets are studied and analyzed.In the traditional K-means clustering algorithm,the segmentation effect of the skin rivet corrosion region is extremely unstable and the false positive rate is high.In this thesis,the mean standard deviation K-means method is proposed to segment the corrosion area of the skin rivet.The initial cluster center is determined by calculating the sample mean and standard deviation,which greatly improves the segmentation accuracy.On this basis,the gray level co-occurrence matrix algorithm is combined to extract the characteristic parameters of the corrosion region.By comparing the accuracy of the SSD target detection algorithm and the support vector machine classification algorithm in the skin rivet corrosion defect detection task,it is found that the support vector machine has higher precision and faster speed,and has obvious advantages for the data set with less sample size.