Study Of Positioning Error Correction Methods For Stereo Vision Camera

The objective world is a three-dimensional spatial world, to get access to the location information of space three-dimensional scenes is an important step to understanding the world. At present, there are a lot of methods and techniques to get access to the location information, three-dimensional visual positioning method is the most important location-aware technology of computer passive ranging, which can directly simulate the way that how the human visual processes the scene, it can measure the three-dimensional scene information in a variety flexible conditions, it can transform obtained two-dimensional image datum into three-dimensional spatial information, at last, achieving the purpose of understanding the world from the images.Stereo vision system in robot vision positioning, aviation mapping, the military field, the medical diagnosis and industrial inspection has become an increasingly widespread application. If these applications can achieve or not accurately, and if a vision system can obtain high-precision feedback information, relate to a basic and important question, namely, the positioning accuracy of the three-dimensional visual system. Positioning accuracy of the three-dimensional visual system has become one of studies on the visual navigation and positioning system. Therefore in order to let three-dimensional visual positioning system obtain more high-precision positioning, this paper will use error correction methods to further improve the positioning accuracy of the three-dimensional visual positioning system, to meet the requirements of a high-precision operation.From the realization of the method of error correction, the error correction technologies conclude software error correction and hardware error correction. Compared with hardware error correction, software error correction technology is more flexible and has some advantages as follows:Firstly, software error correction technology can correct more errors, have the wider range and the higher precision, and can operate overall error correction including systematic errors and random errors at the same time; Secondly, it is cost savings, without investing a lot of money can improve the positioning accuracy of stereo vision camera. Using software error correction technology to improve the positioning accuracy of the three-dimensional visual positioning system is presented in this paper.This article is on the basis of applying the binocular stereo vision system for navigation and positioning, after analyzing and modeling of a large number of data for positioning errors, and then studies the correction method of stereo vision camera positioning error. It mainly includes several aspects as follows: to further improve the visual three-dimensional positioning system, to analyze the formation mechanism of the positioning error, to establish of the location error model and analyze the regular pattern of the error, research methods of error compensation. The main contents of the research are as follows:1. To further improve the visual three-dimensional positioning system: First of all, we compare and analyze the related spatial location technologies of the three-dimensional visual camera, including space-positioning technology and camera calibration techniques; then we study the binocular stereo vision positioning experiment, and add the latter data-processing modules in the visual navigation system, which includes two parts of error correction and accuracy enhancement. At the meantime of perfecting the system, we propose the use of error compensation method to further improve the positioning accuracy.2. Analyze the formation mechanism of the positioning error: through the binocular stereo vision positioning experiments we get a large number of space point positioning coordinates, space point is mainly selected reconstruction of points within the scope of the 64mm×90mm×70mm space, separately 13 groups of space point coordinates datum, each group concludes 176 positioning coordinates. Then we study the classification of measurement error, including systematic errors, random error and gross error. Combining experiments we analyze the causes of errors which include software algorithms, hardware errors, as well as human-induced error.3. Establish the positioning error model and analyze the regular pattern: Get 13 groups of positioning error datum through contrasting the reconstruction of point coordinates of the real world. Analysis the largest and the smallest and the average error the positioning error, propose the amendment results which this article hopes to achieve. We use curve fitting method to establish the two-dimensional mathematical functions error model, as the regular pattern of the error was not obvious in the depth direction ,we also use interpolation method to establish a three-dimensional statistical model of the error. Combined with the error model and the qualitative analysis of the cause of the error in the second chapter, we further analyze the error. Due to the existence of positioning errors regularity, we determine that the location error belong to the scope of systematic errors .The regular pattern of the error in X/Y direction is linear, the error in the depth direction of the middle part of the edge is the largest, and the error on the four corners and in the center of the calibration plane are the smaller ones. And also the regular patterns of the error in the various'surface are similar to each other. With the calibration target parallel moves, the positioning error in the depth direction also moves with the regular patterns of the error in the various error surfaces.4. Based on the analysis of the location error model and regular patterns of the error, we study positioning error correction methods. For the errors of X and Y direction of which the regular patterns of the error are linear, we use the function method combined with schedule method to amend; For the error of the depth we use the method of fuzzy reasoning to amend, where the fuzzy control mainly includes fuzzy processing of the variables, the choice of membership function, the establishment of the fuzzy rules, fuzzy reasoning and selecting the fuzzy decision methods.5. Compiled a three-dimensional visual positioning system software, applying VC + + compiler documentation procedures, we have added a stereo vision camera error correction module. Through starting the Matlab engine, we realize calling the error correction procedure in the Matlab compiler. It could correct the error of any point in the horizon amended, and show the curved surface contrast of before and after error correction in the direction of the depth at the same time; achieve mixture programming of using VC and Matlab , improve the efficiency in the use of the procedure.The experimental results show that after the error correction the positioning accuracy is from 0.6mm to 0.15mm, the positioning accuracy increased by 75% than before, at the meanwhile, the positioning error of X and Y directions is controlled in less than 0.1mm, the positioning error of the depth direction is limited in less than 0.15mm, all the average location error data is less than 0.04mm.It can be better to meet the three-dimensional visual positioning system's requirement. Although after using the error correction technique, the precision of the three-dimensional visual positioning systems have been greatly improved, this research is still in the basic stage, there are still many problems which need to be thoroughly studied and resolved, also the error correction algorithm need to be further improved. However, with the application of the three-dimensional visual positioning system has become increasingly widespread, through the establishment the positioning error model of three-dimensional visual system, it will be of great significance to analyze regular patterns of the error, and use error correction method to improve positioning accuracy .