Some Key Research Of Camera Stretcher And Solar Panel Surface Detection

With the rapid development of computer hardware and software, highperformance,andhighdistinguishratevisioninputdeviceisinvented.Soitispossible that the machine vision technologyis widely applied in the fields ofindustrialautomation,industrialonlinetest,precisionmeasurementandetc.It started early abroad with the application of machine vision in industrydomain. Now many domestic domains are also in bustling with this aspect,meanwhile insufficiencies and shortcoming still exist in this field. Theultimate goal of the machine vision is to make the computer observe thesurrounding environment just like a human being, have the abilityto adapt tothe environment, and finallymake its own judgment. To achieve such a goal,we still need a very long way to go. Currently so-called machine vision ispeople let the computer observe some specific objects, and then input thespecificjudgeability.Baseonvisualsensitivityandfeedback,itcancompletesome specific task in a certain intelligence level. For example, expresswayautomaticcamerasnowcanmonitorsomespeedingvehicles.The machine vision implements according to the convex ImagingPrinciple. First we image an object on CCD. For CCD has the electric chargememory and the electric charge shift characteristic, we can translate opticalsignals received into electrical signals and transmit it. So light intensity istransformed to current intensity so that it is easy to measure or detect. CCDhas two types: one is linear CCD and the other is area CCD. According totheir characteristics, linear CCD is used in non-contact length measure andarea CCD is used to make the character of surface examination or thereal-timemonitoring.The third chapter narrates an example of realizing the real-timenon-contact length measure and roughly demonstrates the processes. Wepoint several places which need to pay more attention in the article. First isthe choice of lighting equipment, for it relates the quality of image and themeasuring accuracy; second is the linear CCD component. It is unnecessarytousethemostexpensiveone.Itisgoodsolongasitcansatisfyownrequest,otherwise it will enormouslyincrease the cost, and this is the most unwillingthingtobeseenintheindustrialapplication.The color of material measured is generallywhite or black, and we drawtwo lines with a color that has a clear contrast with the material. Under thephotosource illumination, as a result of the color difference, the reflectedlight intensity is different. After the light passes the camera, it forms imageon linear CCD, and it is transformed to current .Then through rectifier circuitand capture card, the data is transmitted to the computer. On the terminalmonitor, the data is displayed as a curve, in which there are two prominentparabola, and what we have to do is getting the distance between the twoparabola.We first smooth the primitive image in order to remove the unnecessarynoise. We set the width of smooth window to 3 pixels, and then seek for theedges on the smoothed image. The grey scale has very a big sudden changenearby two parabola edges, and it reflects the approximate positions of twolines on the workpiece. Based on it, we seek for the more precise edgeposition. Now there are many related algorithms on non-contact lengthmeasure, and the most typical algorithm is line fitting least square methodand spline interpolation method. The article has introduced their detailedreasoning formulas. Regarding experimental data, we get results using thesetwo algorithms, and compare with the manual survey result. After analysis,we discover line fittinglease square method is more applicable to our systemthansplineinterpolationmethod.The fourth chapter elaborates a example of non-contact surfacemeasurementusingareaCCDintheindustrydomain.Theworkpiecehas twotypes: one is larger(4cm×3cm) and its color is blue-brown; the other issmaller and its color is yellow-green. Through observing the image we findthere is a series of parallel white stripes and a lot of small noise points,besides the cracks needed to be examined. It has brought the difficultyto thedetection.Ourideaisfirsttoremovetheseparallel whitestripes,thenremovethenoisepoints,andfinallyobtainthecrackdetected.The ways of recognizing white stripes are two kinds. The first is toestablish a reference frame. Then we try to find the slope of these straightlines, then remove these white stripes through the method of fitting line.Through experiments we find that this approach would not be feasible.Becausetheslopes ofthesewhitestripes lookedliketheparallels areactuallydissimilar, we can only detect the images taken at normal incidence, and itdemandshighshootingcondition.Another idea is to use Hough transform. With a certain point on theimage of these white strips, in the predefined scope, we get each straight lineacross the point. And under the slope-intercept coordinate system, the seriesof line correspond to a straight line. Then we do the same thingto each pointofawhitestrip,andthewhitestripcorrespondtoacertainpointunderthetheslope-intercept coordinate system. In other words, under the Cartesiancoordinate, a series of straight lines across a point correspond to a certainstraight line under the slope-intercept coordinate system. And we transformthe problem of seeking the straight lines to a cumulative count of the cells.We get the specific line through seeking the maximum of the units under theslope-intercept coordinate system. The experimental results shows that themethodisfeasible,wecanprocessthedifferentimages.Regarding removing the noise points, we first remove some isolatednoises through the algorithm of auto-adapted threshold value, and thenthrough seeking the edge and calculating the area, length and width of theconnected region, we further remove the larger noise spots. Finally we findthe crack to be detected. From the experimental results, we find this methodis extremely satisfying. We use the OpenCV library to implement thealgorithm. OpenCV is an open source library that developers can freelyinvoke. Moreover, due to OpenCV has a good transplant, developers cancarryonthedevelopmentinMS-WindowsandtheLinux.At last, we conclude with a summary of the result and someshortcomings,andmakeaprospectforourfuturework.