Research On Television Connector Defect Detection System Based On Machine Vision

Today's era is an era of scientific speeding and technological advancement.The powerful traction of technology drives the development and upgrading of the industry.Some or all of the processes in the traditional art have been replaced by machines,and their efficiency and precision have been greatly improved.For example,in the case of quality control,it replaces the defect detection method used in the past.The machine's turnover rate is much lower than that of labor,and the efficiency has increased several times.Not only liberated the human hands,but also liberated the eyes.There are some of the drawbacks of traditional detection methods.In order to ensure high quality,it is necessary to spend a lot of time and layer inspection.It takes a lot of labor to spend.The standards are inconsistent,and there are hidden dangers of false detection and missed detection.People have emotions,which make the test results not objective.This is a long-term boring working environment and high-intensity uninterrupted work,causing physical and psychological damage to quality inspectors.Similarly,television connector detection also has a series of similar problems as described above.This paper takes it as the detection object,and studies and tests how to quickly identify and classify the defects in television connectors,and designs a set of safety stability coefficient and response.A visual inspection system that is extremely fast and capable of automatically and efficiently identifying defects.The system integrates machine vision,image processing technology and machine learning algorithms to meet real-time online detection,control automation and intelligent management.The main work of the thesis has the following aspects:First,the background of connector detection and its research significance in practical engineering applications are summarized.Here is a general problem with manual detection of connectors.The rise,development,growth and extension of machine vision and its research results at home and abroad are introduced;Secondly,it expounds the overall framework structure of the vision system and therequired performance.It emphasizes the composition and selection principles of hardware in the execution organization and the design software architecture,and schedules the hardware to perform actions;Thirdly,it describes the role of image preprocessing and the steps of common algorithms for processing images,including image filtering,enhanced images,and segmentation images,such that features of the connector are highlighted or separated from the background;Fourthly,The pin coplanarity and the improved local signal-to-noise ratio algorithm are introduced to extract the feature data.and the support vector machine design and problem equivalence conversion in machine learning algorithm are introduced;the method of pin-coplanarity and improved local signal-to-noise ratio(SNR)algorithm for extracting feature data.The idea of support vector machine and problem equivalence conversion in machine learning algorithm are introduced.Fifthly,the experimental analysis and evaluation of the construction and training of the mathematical model of SVM algorithm are carried out.The data evaluation methods of the two measures of confusion matrix and ROC curve are adopted,and the parameter tuning of the kernel function is introduced.The detection efficiency is much higher than that of manual.The detection accuracy is high and meets the actual needs of the project.Accuracy rate reached 98.63%?...