基于连续图像消影点的视觉定位算法研究  

Research on Visual Positioning Algorithm Based on Vanishing Point of Continuous Images

作　　者：王媛 刘小晗 王文杉 许忠雄 WANG Yuan;LIU Xiaohan;WANG Wenshan;XU Zhongxiong(China Academy of Electronics and Information Technology,Beijing 100041,China)

机构地区：[1]中国电子科学研究院,北京100041

出　　处：《火力与指挥控制》2022年第5期66-74,共9页Fire Control & Command Control

基　　金：国家自然科学基金资助项目(U19B2036)。

摘　　要：将消影点作为视觉定位的描述特征具有独特优势,针对消影点容易受到环境因素影响的问题,提出一种基于图像梯度场的消影点提取算法。利用优化的检测算法提取直线特征,对直线段筛选后估计消影点,并利用精度因子得到有效消影点,根据计算机视觉中的多视图几何实现视觉陀螺仪和视觉里程计。实验结果表明,通过过滤不可靠的消影点能够提高定位精度,视觉陀螺仪整体准确度优于惯性导航,部分数据点视觉里程计准确度优于惯性导航。The vanishing point as a descriptive feature of visual positioning has unique advantages.Aiming at the problem that the vanishing point is easily affected by environmental factors,a vanishing point extraction algorithm based on the image gradient field is proposed. The linear feature is extracted by the optimized detection algorithm. The vanishing point is estimated after the line segment is filtered,and the effective vanishing point is obtained by calculating the precision factor. Then realize the visual gyroscope and visual odometer according to the multi-view geometry in computer vision. The experimental results show that the accuracy of visual positioning is improved by improving the reliability of the elimination point,the accuracy of the visual gyroscope is better than that of the inertial navigation,and the accuracy of the visual odometer of some data points is better than that of the inertial navigation.

关 键 词：消影点 视觉陀螺仪 视觉里程计 视觉定位 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]