Research On Feather Measurement Based On Digital Image Technology

China is the largest producer of badminton in the world.The output of badminton produced by C hina accounts for up to 60% of the global output of badminton.In the current production and manufacturing workflow of badminton,the cutting of feathers is still a labor-intensive link.The cutting of feathers requires manual calibration with calibrators,which seriously affects the efficiency of this work link and restricts the overall efficiency of badminton production and industry development.Therefore,the non-contact measurement with machine vision technology can effectively guarantee the accuracy of measurement and improve the production efficiency.At the same time,work-related injuries caused by calibrating dimensions often occur in the shuttlecock cutting operation.It is also necessary to replace manual detection of feather thickness with computer technology.Using machine vision technology to automatically measure the thickness and size of feathers without staff involvement can significa ntly reduce labor costs and job risks,and can significantly reduce production time,improve production efficiency and reduce equipment depreciation costs.Compared with the traditional workflow of manually inspecting products,using this technology can automate mass production,effectively guarantee the quality,accuracy and efficiency of detection.This study mainly discusses the use of computer vision technology in the production process of badminton,instead of the traditional work link of manually cutting feathers,to achieve the automation of this link to improve feathers.The production efficiency of wool ball is as follows:(1)Research on image preprocessing methods for feather samplesThrough the analysis of some mainstream methods in image preprocessing module,such as image enhancement,image filtering,histogram equalization,etc.,according to the characteristics of feather image samples in this thesis,the dataset is preprocessed,and these algorithms are used to train multiple sets of datasets in this thesis,the training results are compared and analyzed,and the best fit is obtained from the angle of relative error rate of dimension measurement.Train the method of data set in this thesis to get the ideal data set processing effect.(2)Optimizing the Roberts algorithm model with Roberts interpolation algorithm InView of the characteristics and shortcomings of the Robrts algorithm model,the Roberts algorithm model is improved accordingly.In the model,the Roberts operator is used to obtain pixel-level image edges,and then the quadratic polynomial interpolation is used to obtain the high-precision coordinates of subpixels.In the new algorithm model,the subpixel edge detection mechanism is introduced to further improve the data accuracy.Relative error rate of dimension measurement is reduced by nearly 50% to 9.47% on the basis of Roberts algorithm model,and a set of ideal dataset results is obtained through Roberts algorithm optimization model.Machine vision technology can effectively support the computer integrated processing of information and data.It is one of the most important basic technologies in the computer integrated manufacturing system.Using computer vision technology in all industries can bring enormous benefits to the production and manufacturing of the industry.