Image Mosaicking In Microscopic Images Of Nonwovens And Its Applications

Large panoramic images have great application values because they can be used to acquire macroscopic information of the targets accurately and completely.While large panoramic images can be obtained by image mosaicking technology,which has been widely applied in medical image analysis,pattern recognition,virtual reality,photogrammetry,remote sensing,military and other fields.In this paper,methods of using image mosaicking technology to obtain large panoramic images of nonwovens were studied,while the applications of large panoramic images in testing the structures and performances of nonwovens were carried out.First of all,original image sequences of nonwovens which contain overlap regions,were collected through an optical microscope equipped with an automatic slide platform.Secondly,the image sequences were stitched into a large panoramic image according to the improved Harris_SIFT algorithm.Finally,the properties of nonwovens,such as orientation and bonding spot area ratio,were tested with the large panoramic image.In order to collect microscopic image sequences of nonwovens,the slide platform of an optical microscope was moved toward a certain direction with a fixed step length,controlled by a computer.And then the images of nonwovens were preprocessed by gray scale and image segmentation,which made the subsequent image processing operations possible.The current image mosaicking methods can't be applied to the images of nonwovens very well.Therefore,two improved mosaicking methods to match the captured image sequences were proposed in this paper.Method A is an improvement of the NCC algorithm and SSDA algorithm.Introduced Harris corner detection into the NCC algorithm,the amount of computation of NCC algorithm was reduced and the efficiency of the algorithm was improved by combining the displacement parameters of microscopic images of nonwovens.And then the SSDA algorithm was used to select the best matching point.Method B is an improvement of Harris corner detection algorithm and SIFT image registration algorithm.The Harris corner detection does not have the scale invariant feature while the SIFT feature descriptor is high in dimension,large in computation quantity,slow in calculation speed.To solve the problems above,an improved Harris_SIFT mosaicking method was proposed.Combined with the characteristics of displacement parameters which exist in adjacent microscopic images of nonwovens,matching regions was reduced to an appropriate range,the matching time was reduced greatly and the matching efficiency was improved.On this basis,the automatic microscope image mosaicking scheme was designed.Images were collected and stitched into a large panoramic image by Harris_SIFT mosaicking method in real time.Combined with image segmentation,seed filling,area test,orientation analysis and other image processing technologies,the rolling point area ratio and fiber orientation structure were tested with the panoramic image.The experimental results show that the large panoramic images stitched by original image sequences not only keep enough image details,but also cover the entire outlines of the material structures.Therefore,they can be observed easily.In addition,it has been proved that the large panoramic images can be used to test bonding spot area ratio and orientations of nonwovens quickly and accurately.