Research On Vision-based Defect Detection And Classification Algorithm Of Insulator Core Rod

Vision-based defect detection technology is widely used in industrial inspection for its high accuracy,high efficiency and other advantages.Quality control during manufacturing process of composite insulator core rod is of great importance.Tradition way of quality control relies on human eyes to manually inspect every part of the product,which is inefficient and unreliable.To boost productivity and to increase reliability,it's quite essential to design an automatic defect detection algorithm for insulator cord rod.The research target is to design a computer vision based defect detection and classification algorithm for insulator core rod.A comprehensive study on the properties of insulator core rod is carried out in this thesis.The content of this research includes:image extraction,feature extraction and analysis,classification algorithm,and system implementation.The main innovations and contributions of the paper are as follows,Firstly,a defect detection and segmentation algorithm based on visual saliency is proposed.Based on the study of the properties of insulator core rod and the environment of the production line,A defect detection and segmentation algorithm is designed.By connected component labeling,bounding box extraction,noise reduction and region merge,a good defect extraction is achieved.Then,the feature extraction and analysis are accomplished.11 features are extracted from the defect region based on prior knowledge of the defects.After that,the automatic defect classification method is studied.Based on the characteristics of the experiment condition,BP neural network is chosen as the classification model.The selection of model parameter and the training process of the neural network are also presented in this work.The experiment shows a good result with 80.4%average accuracy.Finally,the prototype of the defect inspection software is build.The main function modules of the system is accomplished standing on the analyzing of the requirement.In conclusion,the key technologies of the visual-based online defect detection system of insulator core rod are studied and implemented in this paper,and a good experiment result is achieved.