基于鱼眼镜头的单目视觉定位系统

Monocular Vision Positioning System Based on Fisheye Lens

随着无人驾驶,扫地机器人的广泛应用,基于移动机器人技术的定位导航系统研究成为了一项科研热点。移动机器人达到自主导航的目的,需要具备精确的定位功能。视场角在定位系统中起着重要的作用,越大的视场角代表着更多的环境信息获取,而传统的单目相机的视场角小,在纹理较少的环境中,无法进行有效的定位。本文利用鱼眼镜头来替换传统的单目针孔相机,可以增加定位系统的视场角,有效利用冗余观测信息来保证位姿估计的鲁棒性和观测目标特征的长可见性。在本文中,设计了搭载鱼眼镜头的定位系统可以捕获更丰富的环境特征信息,并且引入了相机畸变模型改善图像畸变的问题,设计了基于特征点对图像信息进行处理的系统,实现了导航定位的功能,获得了更鲁棒的位姿跟踪效果。实验表明,基于鱼眼镜头的单目视觉定位系统定位性能提高,实现厘米级定位。

With the wide application of unmanned and sweeping robots,the research of positioning and navigation system based on mobile robot technology has become a scientific research hotspot.To achieve the purpose of autonomous navigation,mobile robot needs to have accurate positioning function.The field of view angle plays an important role in the positioning system.The larger the field of view angle represents more environmental information acquisition,while the traditional monocular camera has a small field of view angle and can not locate effectively in the environment with less texture.In this paper,fish eye lens is used to replace the traditional monocular pinhole camera,which can increase the field angle of view of the positioning system,and effectively use redundant observation information to ensure the robustness of pose estimation and the long visibility of observation target features.In this paper,a positioning system equipped with fish eye lens is designed,which can capture more abundant environmental feature information,and the camera distortion model is introduced to improve the problem of image distortion.A system for processing image information based on feature points is designed,which realizes the function of navigation and positioning and obtains a more robust pose tracking effect.Experiments show that the positioning performance of monocular vision positioning system based on fish eye lens is improved and centimeter level positioning is realized.