基于改进遗传算法的单目相机标定和目标定位的研究

发布时间：2018-07-13 19:12

【摘要】：机器视觉自20世纪60年代中期出现以来已经有了较为长足的发展,其理论与应用研究也取得了丰富的成果。机器视觉已经在科学研究、社会生活、工业生产等等各个领域有了较为广泛和成熟的应用。在机器人视觉伺服,摄影测量学以及其他视觉应用领域中,相机标定是从图像中提取精确三维信息的先决条件,其标定结果的精度及算法的稳定性直接影响相机工作产生结果的准确性。本硕士文在国内外研究的基础之上,对相机标定技术进行了系统的理论研究与技术分析,包括相机标定过程中的图像特征提取算法改进、相机内参数的求解算法优化等等,并设计实验对所提算法的有效性和准确性进行验证;在此基础上,对单目相机的目标定位技术进行了探索研究。本硕士论论文主要工作和创新有:1.在对相机标定中常用特征提取方法分析基础上,提出了一种针对阵列圆形模板的特征提取改进算法。该算法利用圆形阵列模板的特殊性,通过三步过滤方法,快速的寻找到模板在图像平面上的投影,最后采用改进的Hu方法和亚像素方法,精确定位椭圆的圆心。2.提出了一种利用遗传算法对相机内参数进行优化求解的相机标定算法。该算法首先根据张正友的相机标定方法计算得到相机内参数作为内参初始值,然后通过改进的遗传算对初始内参数进行优化迭代得到更加精确的相机内参数。该方法分别用于两种不同的相机标定模板:棋盘格和圆形阵列模板,相机内参数标定的精确度较张正友相机标定方法有所提高。3.提出一种基于遗传算法的单目视觉的目标定位方法。该方法采用两幅图像进行定位,通过在图像中设置人工标志位,有效的避免因使用RANSAC图像匹配方法而导致的图像误匹配,并运用遗传算法实现优化求解基础矩阵,在此基础上,实现单目视觉的较为精确的目标定位。
[Abstract]:Since the emergence of machine vision in the middle of 1960s, it has made great progress, and its theory and application have also made a lot of achievements. Machine vision has been widely used in scientific research, social life, industrial production and other fields. In robot vision servo, photogrammetry and other visual applications, camera calibration is a prerequisite for extracting accurate 3D information from an image. The accuracy of the calibration results and the stability of the algorithm directly affect the accuracy of the camera work. Based on the research at home and abroad, this thesis makes a systematic theoretical research and technical analysis on camera calibration technology, including the improvement of image feature extraction algorithm in camera calibration process, the optimization of algorithm for solving camera internal parameters, and so on. Experiments are designed to verify the validity and accuracy of the proposed algorithm, and on this basis, the target location technology of single camera is explored and studied. The main work and innovation of this master thesis are: 1. Based on the analysis of common feature extraction methods in camera calibration, an improved feature extraction algorithm for array circular template is proposed. The algorithm uses the particularity of the circular array template, through the three-step filtering method, quickly finds the template projection on the image plane. Finally, the improved Hu method and sub-pixel method are used to accurately locate the center of the ellipse. A camera calibration algorithm based on genetic algorithm is proposed. According to Zhang Zhengyou's camera calibration method, the internal parameters of the camera are first calculated as the initial values of the internal parameters, and then the improved genetic algorithm is used to optimize and iterate the initial internal parameters to obtain more accurate internal parameters of the camera. This method is applied to two different camera calibration templates: chessboard lattice and circular array template respectively. The accuracy of calibration of camera parameters is improved by .3. compared with the calibration method of Zhang Zhengyou camera. In this paper, a method of target location based on genetic algorithm for monocular vision is proposed. In this method, two images are used to locate the image. By setting the artificial mark bit in the image, the mismatch caused by the use of RANSAC image matching method is effectively avoided, and the genetic algorithm is used to optimize the solution of the basic matrix. A more accurate target location of monocular vision is realized.
【学位授予单位】：扬州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP18

【参考文献】