Research Of Defect Recognition Method For Highlight Rotating Surface

The design quality inspection of mechanical parts is an important mean for modern industrial automation with the high performance and high-speed orientation during the manufacturing process.As an essential criterion for judging product quality and evaluating status,surface measurement directly affects the user experience and service life of mechanical products.The application of digital image detection technology has become the mainstream of surface defect detection due to the criterion demand for surface measurement on metal surface in which there exists many difficulties.Follow the principle of low-cost,low-delay and modularization,the research of defect recognition for metal surface with specular reflection characteristic is proposed in this paper.In this dissertation,based on the survey of the present research situation and development trends at home and abroad of some crucial techniques involved with the research project,by absorbing techniques such as precision instrument,image processing technology,computer science,etc.,the design of the defect recognition algorithm is realized independently.Meanwhile,the relevant key techniques on highlight surface defect detection,defect image preprocessing,image segmentation,image recognition and classification,defect characterization,data management and analysis have been studied.1.The image segmentation and recognition in image processing is the primary problem to the inspection based on image processing.Thus,in this dissertation,the traditional algorithm is ed as a number of links,and a novel defect recognition method based on adaptive global threshold for highlight metal surface is proposed and simulated so as to realize the recognition of defect on highlight surface under the premise of satisfying the testing requirement.The first step is about the structure of the compromise filtering pattern fully using the information of both spatial domain and range,which is used to smooth the image of highlight noise and preserve the edge information of target area.Following this,the first derivative of Gaussian function is constructed to construct the Canny optimal edge detector to complete the line differential filtering.The gradient non-maximal suppression operation of the image is used to complete initial image segmentation.In addition,based on the most common variances,the primary segmentation results are divided into the target defect and the noise background,and the binary recognition results of image are achieved.After that,the identification results are morphologically manipulated to obtain the final state of defect within the tolerance range.2.To realize the defect identification and classification,the characteristic of acquired measurement data is analyzed.The defect characterization parameters is defined and the relevant results are presented in the form of defects data tables.The converge example has been presented to analyze how to identify and deal with the image with defect by Oracle,thus the subsequent quantitative analysis could be supported by relevant data information.3.Aiming at the recognition accuracy and identification efficiency,etc.,on the basis of introducing three kinds of experimental images,the related defect recognition experiment was completed.Firstly,the complete defect recognition experiments based on general images captured on a conventional camera was completed,and the results showed that the proposed method was feasible.Secondly,the defect classification experiments based on defect characteristics and the experimental images with stripe captured on the experiment platform was completed,and the results showed that the proposed method,compared with other algorithms,preformed much better than other method.Finally,the contrast experiments based on imaged with standard binary mask was completed,and the results showed that the proposed method processed high precision.Moreover,the accuracy of segmentation results is verified quantitatively based on the index of misclassification error,and it showed that the recognition effect of proposed method is better.The proposed method in this work also has the important reference value regarding inspection for other work-pieces.