Research On An On-line Glass Defect Inspection System Based On Machine Vision

Glass defect online detection system has played an important role in the development of the national economy.Machine vision is the core of the glass defects on line detection sy stem.With the rapid development of computer technology,information technology and net work technology.The superiority of on line detection system based on machine vision glass defect is more obvious,and has become the mainstream of the development of automatic d etection system and research hot spot.On line detection system of glass defects based on m achine vision has be designed in this paper,an improved dynamic threshold algorithm and k ernel function selection method based on support vector machine and its parameter optimiz ation method are proposed,feature vectors of the glass defects are extracted and their recog nition and classification are carried out.Firstly,in this paper,the structure of machine vision as the guiding ideology,through th e detailed introduction and analysis of the function of each module of the hardware and sof tware,Complete the development of suitable for on-line glass defect detection system.Secondary,According to the characteristics of Glass defect images,The preprocessing method of the defect image is studied,and the enhancement of the defect image is realized.This paper focuses on the in-depth study of the defect extraction module,and discusses the shortcomings of the traditional defect extraction algorithm,improved dynamic threshold alg orithm is proposed.Through experiments the algorithm was analyzed and determined.The experimental results show that improved algorithm can fast defects the glass detects,the ac curacy rate of up to 97%,can meet the requirements of Glass online quality detection.Finally,studied the defect classification algorithm based on support vector machine,de signed and implemented of the multi-class classifier based on support vector machine,and s elect the defect sample classification experiment,obtained better classification results.Expe rimental results show that the algorithm is more accurate and efficient than the existing cla ssification algorithm.