The Spot Image Extraction And Error Compensation For Real-Time Three-Dimensional High-Precision Measurements

The3D measurement based on computer vision is widely applied in aeronautical manufacturing,celestial navigation, automobile research and exploitation and other fields。Some key problems of3Dcoordinate real time measurement for dispersed targets based on stereoscopic vision are studied in thispaper. Especially the acquisition and center location of the light spots with2D Gaussian distributionmodel is deeply discussed including the forming principles and the design of hardware equipments. Anew method called Gauss Surface Analyzing for center location of the light spots is described here; thesource and compensation of systematic error in center location are also involved. The main contents andresults are as follows:1, Combined with the datum planes calibration problem in the aircraft assembly process, the keyissues of the dynamic measurement of pose based on target panel and monocular vision are studied,especially the fundamental principles and the flow of algorithms are thoroughly discussed, as well asthe design and manufacturing of hardware equipments.2, By analyzing the camera imaging principles, it follows that the light spot of point light source fcomplied with Gaussian distribution is caused by the defocused shooting and diffraction by small holes.On base of this conclusion, the calculating formula of the critical diffraction aperture is derived,followed with the design and manufacturing of circular aperture. Combined with the camera lens andCCD camera and infrared LED, an experiment which could be use to get the light spots with2DGaussian distribution model was invented.3, By dividing the coordinate of the spot center into integer pixel value and sub pixel value, a newfast center location algorithm, namely Gaussian Surface Analyzing (GSA) is deduced, in which thecenter is located with analytical formulas and the intensity values of all pixels in the region of interest(ROI). The experiments have confirmed that compared to the GSF, the GSA is accurate and robust allthe same, however the running time is no more than1/278of the GSF respectively. The GSA can beimplicated in the field of satellite navigation, visual measurement and astronomical observations.4, By analyzing the CCD imaging principle, we find that the center of the light spot is notcompletely equivalent to the target center, which is an important source of systematic errors. In addition,the difference between the real gray level of spot and the mathematical models which the center locationalgorithm based on is also one of the sources of systematic errors. Based on those conclusions, thedistributions of the system errors of six different center location algorithms are estimated withexperimental methods, and the compensation methods are given here as well.