基于虚拟平面靶标的大视场摄像机标定技术研究

发布时间：2018-04-20 23:06

本文选题：虚拟大平面靶标 + 小尺寸靶标组合　；参考：《南昌航空大学》2017年硕士论文

【摘要】：基于立体视觉的测量系统被广泛应用于工业、农业、国防等领域,摄像机标定是其关键步骤,标定的精度直接决定整个测量系统的精度,因此,研究高精度的摄像机标定技术对提高立体视觉测量系统的精度具有重要意义。本课题组在开展直升机旋翼桨叶运动参数测量研究时,发现传统较高精度的摄像机标定方法在大视场环境下,标定精度不理想。据此,本文分析了影响基于立体视觉大视场摄像机标定精度的影响机理,采用虚拟大平面靶标(Virtual Large Planar Target,简称VLPT)研究了大视场摄像机标定技术,并系统分析了标定精度的影响因素。主要工作内容及其研究成果如下:(1)构建了大视场摄像机标定方法的实验验证系统。根据标定方法的验证要求,分别从硬件、软件和靶标三个方面,设计出实验验证系统。硬件平台包括视觉模块和数据处理模块;软件平台分为图像采集模块、摄像机标定模块和标定精度测试模块;分别从靶标的类型、标记点的类型、标定点的个数、靶标的定位、标定点的排序五个方面,设计出8*8圆形标记点二维,且带有靶标定位点和排序定向点的靶标。该实验验证系统为验证本文方法的有效性和分析标定精度的影响因素提供基础保障。(2)提出了基于VLPT的大视场摄像机标定方法。首先,利用多个相互独立的小尺寸靶标(Mutually Independent Small Target,简称MIST),获取多张标定图像;其次,针对每张标定图像,采用定位标记点圆心的方法,找到各个MIST,并为其上标定点进行排序和编号;再次,设定虚拟平面,通过在该平面上找到每个标定点对应虚拟点的方式,将多个MIST联系到一起,构造出每张标定图像相应的VLPT;接着,利用已获得的多个VLPT,计算出摄像机参数,并通过非线性优化算法对标定参数进行全局优化;最后,将本文方法与传统标定方法进行对比实验,结果表明:本文方法的标定精度,与大尺寸靶标的标定精度相近,且明显高于小尺寸靶标的标定精度。(3)分析了VLPT标定方法精度的影响因素。根据研究本文方法的过程,分别从标记点的定位精度、靶标的摆放方式、MIST的大小、标定图像对的数目四个方面分析其对标定精度的影响机理,采用理论分析与实验验证相结合的方式,得出了以下结论:1)标记点的定位精度对标定参数的获取过程影响不大,但严重影响了待测点的三维信息计算结果;2)通过“米字”形摆放方式,选取靶标面积为245mm*245mm的MIST,通过8对标定图像,可以获得较高的标定精度。这为本文方法的合理使用提供了指导性的建议,进一步提高了本方法的标定精度。
[Abstract]:The measurement system based on stereo vision is widely used in the fields of industry, agriculture, national defense and so on. Camera calibration is the key step. The precision of calibration directly determines the accuracy of the whole measurement system. It is very important to study the camera calibration technology with high precision to improve the precision of stereo vision measurement system. In the research of helicopter rotor blade motion parameter measurement, our group found that the traditional camera calibration method with high accuracy is not ideal in large field of view. Based on this, this paper analyzes the influence mechanism of camera calibration based on stereo vision, and studies the calibration technology of large field camera by using Virtual Large Planar Target, a virtual large plane target. The influencing factors of calibration accuracy are analyzed systematically. The main work and research results are as follows: 1) an experimental verification system for large field camera calibration method is established. According to the verification requirements of calibration method, the experimental verification system is designed from three aspects: hardware, software and target. The hardware platform includes visual module and data processing module, the software platform is divided into image acquisition module, camera calibration module and calibration precision testing module. There are five aspects in the ranking of marked points, and the two dimensional target with target location and sort orientation point is designed for 8 ~ (8) circular marking points. The experimental verification system provides a basic guarantee for verifying the effectiveness of the proposed method and analyzing the factors affecting the calibration accuracy. (2) A large field of view camera calibration method based on VLPT is proposed. Firstly, multiple calibrated images are obtained by using several independent small Independent Small targets (MISTs). Secondly, for each calibrated image, the method of locating the center of the mark point is used. Find each MIST, sort and number its superscript points; third, set the virtual plane, and associate multiple MIST together by finding the corresponding virtual point of each marker point on that plane. The corresponding VLPTs of each calibrated image are constructed. Then, the camera parameters are calculated by using the obtained VLPTs, and the calibration parameters are optimized globally by nonlinear optimization algorithm. The experimental results show that the calibration accuracy of this method is similar to that of large scale target. The accuracy of VLPT calibration method is obviously higher than that of small size target. According to the process of studying the method in this paper, the mechanism of influence on calibration accuracy is analyzed from four aspects: the positioning accuracy of marking points, the size of MIST and the number of calibrated image pairs. By combining theoretical analysis with experimental verification, it is concluded that the accuracy of tagging points has little effect on the acquisition of calibration parameters. However, the result of 3D information calculation is seriously affected. By means of "meter word" arrangement, the high calibration accuracy can be obtained by selecting 245mm*245mm whose target area is 245mm*245mm, and by using 8 pairs of calibrated images. This provides guidance for the rational use of this method and further improves the calibration accuracy of the method.
【学位授予单位】：南昌航空大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】