Font Size: a A A

The Key Technology Of Matching Stitching Based On Residual Bamboo's Outline Of The Curvature

Posted on:2009-01-10Degree:MasterType:Thesis
Country:ChinaCandidate:Y ChenFull Text:PDF
GTID:2178360242493135Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
Bamboo is the first of China's ancient book form, made from wood chips or Bamboo, for the record of ancient military, cultural, judicial and household registration information. With a large number of precious cultural relics unearthed, the heavy workload also follow. Including the Bamboo cleansing and the collation of information. Because of the time over a long time,Bamboo slips of their own natural environment by erosion, and in the process of excavation easily fracture. To extract Bamboo on the integrity of the text messages also need to complete the restoration of a block Bamboo, we have this slips recovery process is known as matching stitching process. Slips matching stitching pattern recognition technology is a new technology development, has now applied to the heritage restoration, broken photo identification, machine parts interface, and other fields. In this paper, the characteristics of Bamboo slips, select a match based on Hausdorff distance splicing technology, the main study include:Bamboo picture in the collection or send the process, due to a photoelectric converter such as CCD coupled device, AID converters, such as the nature of the differences and the impact on the surrounding environment and the parameters of choice and other factors, collected the images may contain the different levels of noise and distortion, in order to effectively carry out feature extraction processing, it is necessary to eliminate noise, distortion correction, so that the output image with a clear outline of the benefit from the characteristics of slips. Acquisition to the text of the original Bamboo picture of a weighted value of filtering, this method makes up for the general lack of filtering can be a very good filter out noise while preserving the edge details. Floating threshold value of the two methods will be divided into a number of sub-block image, for each sub-block selected a threshold. This threshold method and fixed threshold value of the two methods, according to the local gray images automatically choose a suitable threshold, giving full consideration to the uneven nature of the image, can get the best value of the effect.Edge Image is one of the most basic features, including the information used to identify a useful, for people to describe or identify and explain the objectives of image provides a valuable and important parameters. In this paper, Roberts edge detection operator, Sobel edge detection operator, Prewitt edge detection operator, Canny edge detector, Laplacian edge detection operator for a comparative analysis, has been selected with the final second-order differential operator of the Laplacian edge detection operator . The operator of second derivative zero-crossing from the edge of the width of a pixel units, the results need not be on the verge of thinning treatment, so as to ensure the accuracy of positioning the edge.Because of the amount of information to reflect the image of local features, more reliable image of the match, corner detection technology has been widely used. Of the scale, from the border curve on the corner detection algorithm, including single-scale algorithm and multi-scale algorithm. Single-scale corner detection methods there are only focused on the dramatic changes of direction while ignoring the overall curve shape, so that corner of the test results or there is a false corner, or overlooked an important corner. To overcome these shortcomings, has been selected based on the Weibull function of the multi-scale corner detection methods. The method adopted Weibull function as a smooth convolution of nuclear, select a specific formula for calculating the curvature of the edge curvature, selected based on the curvature of the corner extraction methods from the corner.Residual Bamboo slips of the automatic pieced together from many scattered off Bamboo slips, through the use of computers, through the contour matching splicing technology to identify adjacent to each other residual Bamboo slips, then return the whole integrity of the original appearance of Bamboo slips .Bamboo slips is the outline of the characteristics of the feature points and their neighbors from the number of points within a point of the set point on the value of each element is the outline of the points on the curvature value. By calculating the curvature of the characteristics of Hausdorff distance between the characterization of their degree of similarity, Hausdorff distance is less that two characteristics of the higher level similar; to the contrary, Hausdorff distance value is more, that two characteristics of the degree of similarity Low. This algorithm can better match the premise of accuracy, can achieve practical computing speed.
Keywords/Search Tags:Bamboo, Photo pretreatment, Edge extraction, Corner detection, Matching stitching
PDF Full Text Request
Related items