Design Of Inspection Algorithm And System For External Surface Defects Of High Frequency Transformer Based On Machine Vision

The high-frequency transformer must be inspected and sorted for appearance defects before the delivery of the finished product,which is a key process of its product quality control.At present,the detection and sorting of the appearance defects of high-frequency transformers are completed by human observation at the scene.Manual detection methods have the disadvantages of high labor intensity,low efficiency,large influence of subjective factors,and easy to misjudge and miss detection.Therefore,the development of automatic and intelligent detection systems for appearance defects of high-frequency transformers is urgently needed.In this paper,the defect detection of high-frequency transformer external surface is the object,and the types and characteristics of external surface defects are analyzed.The corresponding image acquisition and processing system was constructed,focusing on the character recognition,broken and dirty defect detection algorithms of the external surface,and the corresponding recognition algorithm software and detection system monitoring and management software were developed.The main work and conclusions of this paper are as follows:(1)Aiming at the development of the hardware system for external surface defects of highfrequency transformers,according to the overall system scheme and technical requirements,the types and characteristics of external surface defects are analyzed,and a mechanical device with automatic loading and unloading of workpieces and position conversion functions is designed,combined with imagesthe image acquisition system,Combined with the image acquisition system,the hardware platform of the high-frequency transformer external surface defect detection system is constituted.(2)Aiming at the problem of multiple defects and interference information in the external glue image of high-frequency transformers,a regional positioning method based on the combination of template matching and differential algorithm is proposed.Applying color space conversion and median filtering,creating area templates and affine transformation,and difference calculation,smooth processing of external surface image,location and segmentation of defect areas are realized.Through the use of color space conversion and median filtering,creation of region templates and affine transformation,and difference calculation,smooth processing of the external surface image,location and segmentation of defective regions are realized.(3)Aiming at the detection of bad outer glue characters,broken outer glue and dirty outer glue,a small sample character recognition algorithm based on support vector machine is designed.The character segmentation method combining morphology and minimum enclosing rectangle is studied;through regularization coefficient and kernel function d optimization,a small sample set of outer glue character recognition classifier is designed;36 character classifiers can be better trained by using a small sample set of 36×50 defect images,and the character defect recognition rate is over 97%;a feature extraction and defect detection method based on the combination of area and roundness is designed.(4)Aiming at the development of the host computer software,a software system for external glue defects based on the MFC framework is designed.Under the development environment of Halcon and Visual Studio2017,C++ language and SQL Server database system are used to develop functional modules such as camera control,image processing and feature extraction,defect recognition algorithm,detection classification and corresponding data management of the external surface defect software system,and integrated with the hardware system.The test results show that the recognition rate of the external surface character recognition detection based on the support vector machine is 97.83%,and the surface quality detection accuracy rate of the external surface broken and dirty defects is 98.17%.The efficiency and accuracy of defect detection were greatly improved compared with manual inspection,which realizes the automatic detection of the defects of high-frequency transformer external surface,meets the quality inspection technical requirements of high-frequency transformer manufacturers,and has great market application prospects.