The Research On Classification Of Defects In Transmission Hub Based On Feature Point Detection

With the continuous improvement of people's quality of life,motorcycle products have been widely popularized.At this stage,people are often concerned about the quality of motorcycles.One of the important parts is the transmission hub.If there is a problem with the transmission hub,it will seriously affect the motorcycle gear control.Therefore,it is of great significance to detect the defect of the transmission hub.This thesis will analyze the defects of the transmission hub.And the technology involved in feature extraction and classification of defects is introduced in detail.Aiming at the problem that the traditional SIFT algorithm has low efficiency in detecting feature points,an improved algorithm is proposed.The algorithm introduces FAST multi-scale space,by adding scale parameters to traditional FAST algorithm,and continuously making Gaussian blur on the image according to scale.The algorithm detects the extreme points and compares the different scales to find the optimal corner points and improve the efficiency of detection at key points.The experimental results show that the algorithm also has the characteristics of scale invariance and rotation invariance.At the same time,the efficiency of generating feature description vector is greatly improved,and the accuracy of feature description is high.Aiming at the problem that the number of feature vectors is uncertain,K-means clustering is used to form the final feature vector,which is convenient for classification and training.However,the feature points are local features.In order to effectively improve the detection accuracy,the algorithm extracts the color features such as rust defects,and introduces the global feature color moment,and integrates the local features with the global features,and adopts the "one to all" type SVM multi-class classification model.So it can improve the classification and recognition rate.Experiments show that the recognition accuracy will be improved by using multi-feature fusion classification model.According to the surface defect characteristics of the transmission hub,the corresponding machine vision detection platform is built.Through the comparative experiment,the detection efficiency of the improved SIFT algorithm is improved.And the recognition accuracy of the classification model based on the fusion of local features andglobal features is also improved compared with that of the single feature classification model.