Font Size: a A A

Application Of Active Structure-light Vision Technique Into Object's Dimension Measure And Three-dimension Reconstruction

Posted on:2007-11-12Degree:MasterType:Thesis
Country:ChinaCandidate:X T LiFull Text:PDF
GTID:2178360185954615Subject:Optical Engineering
Abstract/Summary:PDF Full Text Request
In recent twenty years, with the development of computer vision, laser, precision optics,video-data capture and image process technique, objects' dimension measure andthree-dimension reconstruction have been well developed by the structure light active vision.Objects' precise dimension measure is an useful issue which is being used widely and has greatactive value. Computer-vision method for Object-dimension precise measure has been wellutilized by its kinds of merits such as no-touch, high precision, easy process by computer,simple connection with control system etc.Object's three-dimensionreconstruction can bebroadly used by kinds of fields such as robot navigation, vision surveillance, producing forbuildings, producing for model, on-line product inspection etc. In resent years, it is also wellused by computer-based virtual reality, Image Based Rendering, three-dimension animation etc.It can be realized by lots of methods, which can be distinguished into two kinds of methods(touch methods and no-touch methods). Among no-touch methods, the optical method is verypopular. The main feature of optical no-touch method is that the surface of object is not touchedduring the measure by optical method. What's more, the optical no-touch method has the meritsof high speed, good automatization and the wide-measured soft or hard objects withcomplicated surfaces. In recent years, the optical no-touch method of structure light has beenwidely attached importance to by its merits of no touch, easy operation and high precision etc.According to the existent theories, we build up objects' dimension measure andthree-dimension reconstruction system respectively by one-camera vision and two-cameravision. These two systems are both realized by line structure-light active vision. Thecamera-calibration method is supported by random-plane method which is created anddeveloped in recent a few years. The camera video data is collected by the video capturetechnique based on Directshow.The line structure-light vision system, Directshow technique and the random-plane calibrationmethod are separately described in the following content.(i) line structure-light active vision systemFrom the aspect of structure light's form and figure, structure light can be distinguished by dotstructure light, line structure light, circle structure light, multiple-line structure light andmulticolor structure light etc. This paper's one-camera and two-camera systems are alldeveloped by line structure-light vision system. The main operation steps for linestructure-light ,laser-triangle method are that: ①object is lighted by line structure light;②theimages for the lighted object at different object's part are collected by camera and computer;③processing the images, reconstruct the object's three-dimension figure, analyzing thereconstruction figure.The main principle for line structure-light triangle method is that: ①there is a triangle amongobject, projective point and observation point;②when the line-light stripe is projected ontothe surface of object, the stripe will be curved by the scraggly surface of object. Because thecurved stripe is made by the projective line stripe and the surface shape of object, this stripemust contain object's three-dimension information. The object's three-dimension figure can bededuced from the stripes' shape by building up a correct model for the relationship betweencurved stripe and object's surface shape.(ii) Random-plane calibration methodRandom-plane calibration method is a method that inherits part from traditional calibrationmethod and part from self-calibration method.As for one-camera and line structure-light system, only one picture of calibration plane needsto be captured by the camera.As for two-camera and line structure-light system, each camera has better capture at least tenpictures of random-laid calibration plane in order to resolve the calibration of inner parametersfor each camera. Then, the calibration plane should be laid in public view field of both twocameras ,and the picture of plane should be captured by each camera in order to resolve thecalibration of outer parameters for each camera.This calibration method use a calibration plane , but the plane's space position doesn't need tobe measured by operator. The dots positions in plane is the only thing that need to be known.(iii) Directshow video capture techniqueIt is a video capture technique that has no relationship with hardware (such as video capturecard).It dismisses the strong conjunction between the program makers and hardwaremanufactures. It doesn't need the support of further-development program package of video'shardware. So the program made by Directshow is more transplantable and can be used by allkinds of video capture cards or other kinds of video devices.According to this paper's detailed content, the line structure-light system for one camera andtwo camera are introduced separately.(I) three dimension reconstruction system using single cameraUsing the existed theories upwards and experiment condition, we design and set up the systemof measurement. The hardware includes the CCD camera, zoom lens set, four-port imagingcollecting, 20mW neon-helium laser, the beam expander, slit-shaped pore, the flat surfaces offixed points, PC machine, the precision angle turn set and the measured objects etc. Thesoftware was programmed by VC together with Matlab. As respect for calibration, randomplane method was used in order to get the relationship between the curved stripe and theobject's three-dimension figure. This method can be easily and flexibly used. No professionaloperator is needed. What's more, this calibration method is good for adjusting and improvingthe experiment's precision. A useful coordinate-revising method was built up according to theknowledge of analytic geometry and the property of experiment's hardware system (therotation axis of the rotating platform can not be consistent with each of the calibration axis).Objects' geometry parameters, such as dimension and surface area etc, were measured. Themeasure errors of this system were analyzed. The mean single-dot precision is higher than0.1mm/24mm.In final, kinds of objects such as metal objects, plastic objects, rocks,half-transparent glass objects, were measured by this structure light system and reconstructedby CATIA. The reconstructed result is good.(II) Three dimension reconstruction system using two camerasThis paper introduced line structure-light system with two cameras. The main principles forrotated platform and line move platform were separately described. The main operation stepsare: (i) calibrating the two cameras by random plane method;(ii) capturing object's pictures;(iii)After processing the pictures and matching feature points by the polar-line equation, thecorrespondence points can be matched.(iv)According the feature of rotated platform (or linemove platform), object's dimension measure and three-dimension reconstruction can be done.Finally, line structure-light calibration and correspondence points' match were elementarilyresearched according to the experimental condition.
Keywords/Search Tags:Structure-light
PDF Full Text Request
Related items