Font Size: a A A

Research On The Error Detection Algorithm Of Cylindrical Surface Based On Image Processing Technology

Posted on:2017-08-06Degree:MasterType:Thesis
Country:ChinaCandidate:J F TianFull Text:PDF
GTID:2348330515463932Subject:Microelectronics and Solid State Electronics
Abstract/Summary:PDF Full Text Request
In order to guarantee the qualified rate of products,it is necessary to inspect products before they leave the factory.One of the detection is to inspect whether there are stain or character error on the product appearance.For a long time the detection is done manually,this reduces production efficiency and adds costs.With the development of image processing technology,image processing technology plays an important role on industrial production and people's daily life.One of the roles is the image matching and error detection.Most of the algorithm is designed for the plane image which exist linear transformation,and is not suitable for the image which exist nonlinear transformation.For this reason,the research and improve the error detecting algorithm of cylindrical surface is significant.In the first of this dissertation,the error detecting algorithm difference of the plane image and the cylindrical image is analyzed.Then the error detection algorithm of cylindrical surface is divided into two steps.The first step is to rectify the cylindrical surface image.The second step is to match the rectified image and the template image.In order to rectify the cylindrical surface image,we acquire the geodesics form the middle line of string lines by curve fitting.Then we rectify the cylindrical surface image by means of rectifying geodesics.As there may be illustrations and short string lines on the image,we choose to discard them by connected components labeling.Then we obtain the direction of rectifying by applying Hough to geodesics.In the matching processing between the rectified cylindrical surface image and template image,the matching algorithm based on SURF was adopted.There may be mismatching in the matching algorithm based on SURF.As the traditional RANSAC algorithm is only fit for discarding the matching between linear transformation images,and there is nonlinear transformation between the rectified image and the template,the result of discarding mismatching is poor.Then the Euclidean distance and the slope are applied by the curve fitting to get the nonlinear transformation trend.The matching does not meet the trend is discarded.This will improve the matching result.In the end of this dissertation,a summary of the whole dissertation is given,and future research topics are suggested.
Keywords/Search Tags:Cylindrical image rectifying, Image denoising, Geodesic, Curve fitting, Image matching
PDF Full Text Request
Related items