The Image Process And Mosaic Of Monofilament For Large View Scene

In the early 1920s, people has transported the first digital image across the Atlantic Ocean by utilizing Bartlane Cable picture transmission system, it reduced the time to transport from more than one week to less than three hours. The technology of image process was applied to the remote transport of image for the first time, and it made people experience the might power of the technology of the digital image.With the constant development of digital image process and computer technology, especially the maturation/improvement of the microcomputer, the technology of image mosaic is an increasingly popular study field, as a research branch of the robot vision,missile picture guidance,virtual reality modeling based on picture,video compression and video search, it has become a focus of photo graphics, computer vision, image disposing and computer graphics. The purpose of image mosaic is to raise the resolution of image and get larger vision, it can stitch several images of high-resolution and small rangeinto a whole large view scene image according to image acquisition sequence. Basing on the information of pixel dot in the overlapping area of neighborhood image, it also enabled the accurately matching, the stitching and retrieval of the panoramic images and the process and technology of build a vivid large-scale scene image. It has absolutely high theoretical and practical value.This paper aimed at the monofilament in bath monitoring system research of Jilin chemical fiber 50,000 spindles viscose filament. On the foundation of digital image researching, and starting with theory of image mosaic, the paper has made deep and painstaking research on matching algorithm, and explored new method of the template pick-up, matching algorithm, geometry transform and images uniting combining with the image geometry and real-time thinking. It also solved the problem of detection speed system that the once measuring range of CCD is limited at certain resolution, and then realized the gathers, measures and analyses to the sample online, and finally applied the technological of image mosaic to the monofilament in bath monitoring system research of Jilin chemical fiber 50,000 spindles viscose filament.At present, there is no precedent of theory research and practical application of the fibre filament automatical measure in domestic and international field. The exploiture of the detection system realize the mount of monofilament measured automatically instead of measured manually. Not only filled the blank of relevant research, but also created certain social benefits.The main content of this text includes:(1)The design of fibre filament detection system According to the demand of project, the systematic structure of the image detection system were designed, and the correct collection of the digital image of the fibre filament were achieved. Detailed introduction is listed as following:a)Light source: Adopting the annular fluorescent lamp as the light source, its relative spectrum power curve is showed as Fig.1: Fig.1 Relative spectrum power curve of the annular fluorescent lampThe fluorescence lamp presents four linearly band spectrums at 405nm, 430nm, 540nm and 580nm, peak value at 615nm, and the energy drops at long wave section (crimson) place, it shows the light source of fluorescence Wavelength(nm) Relative Spectrum Power have higher radiation energy in green wave band (550nm-560nm) than in the red wave band (650nm-700nm). It is consistent with the relative spectrum response of CCD.Adopting the annular light source, it can overcome the problems such as insufficient intensity, uneven intensity and the directional sensitivity of the receiving device caused by single or several light sources, it can make the light even, drop the directional sensitivity to the lowest, and improve the quality of the image in large extent .b)Geometry condition of measurement: geometry condition of measurement is 45/0: c: 45°lighting annularly, observes vertically. As following picture shows: c)Choice of the lensThe parameter of the lens:Object distance (From the central point of the thing): 40mmDepth of field: 5mmObject visual field range: 16mmOptical amplification: 0.3 timesMany times'testing show that the great depth of field is more practical to silk form thing.d) CCD picture sensorThis system used the ICX204AL type 1/3 inch CCD image sensor with square pixel for B/W cameras made by Sony Company. e) Image gather cardThis detection system adopting the type OK_M20B of Beijing Joinhope image technology LTD. It's 8 bits top-grade image gather card that based on PCI assembling which can collect standard and non-standard interlacing video signal as well as the signal source separated with signals(such as VGA), it's the second generation products of OK series.(2)Choosing and improving of match algorithmsThe total calculation amount of any kind of match algorithm is the calculation amount of algorithms multiply the cout of searching position It's mean that(The total calculating amount)=(the calculation amount of algorithms)×(count of searching position).In order to reduce the total calculation amount, improve speed of measurement, we can either reduce calculation amount of algorithm or count of searching position, or both of them.This text selected the Fixed Mask Correlation Algorithm because of it high precision, high reliability, well mosaic effect, anti-interference strong and not restrain by content;For reduce total calculating amount , we select the tactics that reduce the count of searching position by analyse: a)Reduce the area of the individual template, increase the number of the template, fetch several dispersed small one in the big original template, in other words, the whole area of the template to reduce, but has not reduced the cover area of the template, hardly exist influence to the matche precision;b)In searching for the area, according to the characteristic of the image, restrict the searching area in the overlap area of two image, reduced the search area of the template, helped to reduce both operation time and the probability of mistake matches.In this paper, we use the Normalized Cross-Correlation Value R~G T (m,n) as the matching rule:Among them, the size of real-time image G(x,y) is M×N, the size of template T(x,y) is J×K, J≤M,K≤N.Presume the coordinate of template T as (j,k), the area image G(x,y) within the range of template T as G (j, k), the first pixel dot of G (j,k) on up-left positions is (m, n), the coordinate m and n are speak to G (x,y).In the procedure, door limit LR (m, n) of closing value each other of normalization Fetch 0.7 as judging the condition. R~G T (m,n)>0.7,template is propose tomatching. This number can be changed directly in the procedure code.(3) Software realizinga) The programme codes compile under VC++6.0, the procedure of the software realizes as the following chart to show: (4)Image mosaic example:According to large number of experiments, this kind of match method, have certain feasibility, procedure have applied on the monofilament in bath monitoring system research of Jilin chemical fiber 50,000 spindles viscose filament which the key item of Jilin Provincial Science and Technology Department (Serial number of the project:20050305). It have certain practicability to improve the speed of filaments count.This thesis discussed mainly the image match base on the shift transformation varied, when add some variable and coordinate transformation, this method can be extend to other affine transformation like rotation transformation and proportion varying and so on.