Broken Defects Detection Of Surface Label Based On Machine Vision

Labels are widely used in food,medicine and many other areas of production.The image detection methods based on the machine vision can detect label defects automatically,which not only reduce costs and improve efficiency,but also ensure the stability and reliability of testing.Besides,automatic detection methods improve the automation of bottle industry and provide an increase of product competitiveness.This paper describes the development status of defect detection at home and abroad,and uses the curve detection of surface label as the research target,analyzes the demand of online label defect detection,and proposes an automatic detection scheme of label defect defect based on machine vision.The program is not sensitive to the small displacement and rotation of the target,and the design of a multi-block multiple defect detection scheme based on the similarity measure takes into account the detection accuracy and processing efficiency.The main work of the paper includes:?The overall design scheme of defect detection system for surface tag based on machine vision is proposed.The scheme mainly includes four steps:image acquisition,preprocessing,template extraction and defect detection based on similarity measure.The algorithm flow and working principle of the system are introduced,and the algorithm of the key steps is introduced in detail.?An improved surface correction method is used.The image is mapped from the 3D surface to the two-dimensional plane by a cylindrical backprojection method based on the combination of the quadratic curve fitting and the camera calibration.The image information is complemented by the interpolation algorithm to facilitate the subsequent steps.This method is not sensitive to object displacement,more suitable for practical application environment.?For the case where there may be slight rotation of the target,this paper uses the image registration method to extract the template image.Based on the analysis of the common template matching algorithm,a SSDA algorithm based on image gray scale is adopted for the characteristics of the label image.The algorithm has high precision and fast processing.?A scheme of block multiple defect detection based on similarity measure is proposed.Through the analysis of the common similarity measure,the two indexes of image hash and structural similarity are selected for the characteristics of the label image.The image is segmented and the image is divided into defective and defective regions by using the image hash similarity calculation method with faster processing speed.The defective and defective regions may be used.Determine,in order to obtain the final test results.