In the aviation manufacturing industry,many aviation parts need to spray special paint or rubber on the surface in the later stages of manufacturing to play a role in wear-resistant and protective.At present,most of my country’s spraying methods of aviation parts still use artificial spraying,which not only has low spraying efficiency and poor consistency,but also to harm the health of workers.Transformation needs.For the automatic spray operation of aviation parts,comprehensively use industrial vision,digital image processing technology,database technology and computer control technologies,and develop a software and hardware system that supports the automatic planning,classification,positioning,and spray path automatic planning of aviation parts.To realize the efficient and reliable automation spraying operation of aviation parts,it is of great value for accelerating the transformation and upgrading of the aviation manufacturing to intelligent.The main research content is as follows:(1)On the basis of fully investigating the current needs of aviation parts spraying operations,the overall solution based on machine visual positioning recognition is proposed.Analysis,selection and system integration of software and hardware systems that constitute industrial vision have been analyzed,selected,and system integration,realizing the acquisition of aviation component images,and based on the image principle of imaging systems,the visual system is correct and corrected,which improves the accuracy of image collection.(2)For the problem of multi-part and large field-of-view imaging,the study of key algorithms for image stitching is carried out,and the process steps of image stitching are given.Based on the comparative analysis of common image noise filtering methods and feature extraction algorithms,the median filtering is used to eliminate noise while keeping the edge features of the image well,and the SURF operator is used to achieve efficient detection of feature points;the fast nearest neighbor search algorithm is used for coarse matching of feature points,and the random consistent sampling algorithm is used for fine matching,so as to calculate the single-response matrix for image mapping conversion;Finally,the stitching gaps are eliminated by the image pyramid fusion method to obtain a panoramic image,and the stitching effect is evaluated objectively.(3)A random forest-based image classification algorithm is proposed to solve the automatic and efficient classification of multiple parts.A single-part image acquisition method with threshold segmentation and morphological processing is designed to expand the part image dataset by color perturbation and geometric transformation,improve its generalization and robustness,extract the shape invariant factor and geometric invariant moment of the part image to construct the feature matrix and complete the dataset preparation.The key parameters of random forest construction are optimized by Bayesian algorithm and the model classifier is trained.The dispersion degree of feature vector is visualized and Kruskal Wallis algorithm is used to calculate the feature importance and perform screening,which effectively improves the classification model accuracy.Two spraying path planning methods,reciprocal linear scanning and edge spiral filling for simple parts,are designed for the classification results.(4)Comprehensive utilization.NET Framework,SQLite database,digital image processing technology,MATLAB and other tools have developed the main functional modules and software systems for automatic visual testing for aviation parts.The automatic planning of the collection,identification,positioning,and spraying path of parts image has verified the feasibility and effectiveness of the system,and laid an important foundation for further realizing the engineering application of the automatic spray operation of aviation parts. |