以解算連續像對相對方位參數之單眼視覺里程計

Model by Using Satellite Images Monocular Visual Odometry by Solving Relative Orientation Parameters of Consecutive Image Pairs

林照捷(Chao-Chieh Lin) ； 曾義星(Yi-Hsing Tseng)

《航測及遙測學刊》 24卷2期 (2019/06) Pp. 111-121

https://doi.org/10.6574/JPRS.201906_24(2).0003

摘要

視覺里程計是透過分析連續像對讓機器人可以在陌生環境中持續且自動的定位。無論是在攝影測量或是電腦視覺皆存在該定位技術，但使用方法和目的有些不同，因此本研究結合這兩種方法完成單眼視覺里程計。首先提出合適的匹配策略及自動求解相對方位的方法，以第一張影像的透視中心為原點，且物空間座標系三軸和第一台相機座標系三軸相同，將相對方位依序疊加，獲得連續像對的連續相對方位。但相對方位的誤差會持續累積而產生路徑飄移，因此本研究提出針對相對方位的網型平差解決問題。實驗成果顯示本研究可以重建出合理的相機路徑和周圍的稀疏物點，且透過網型平差，可以有效地更新影像外方位參數及降低誤點偏差量。

關鍵字

單眼視覺里程計 ； 相對方位 ； 累積誤差

並列摘要

Through analyzing consecutive image pairs, the robotic can localize itself autonomously and continuously in an unknown environment without any human input, and it is known as Visual Odometry. Localization can be achieved in photogrammetry and computer vision, but the methods and objectives are little different. Consequently, combinable methods are used to reconstruct the camera path in this study. Firstly, an appropriate matching strategy and an automatic method for solving relative orientation parameters (ROPs) are applied in this study. Secondly, coherent relative orientation of consecutive image pair can be determined by stacking individual ROPs. Since the localization of consecutive images is incremental, the errors are accumulated over time. In this study, a preliminary study of network adjustment of ROPs is proposed to suppress the accumulated error in the navigation. In the tests, results show the estimated camera path and reconstructed 3D object points are reasonable, and the features in reconstruction can be recognized clearly. Comparing estimated camera path before and after with actual camera path, the accumulated errors is reduced significantly. Furthermore, the deviation of object points derived from different image pairs are decreasing. In conclusion, the network adjustment in this study is effective to reduce the accumulated errors of the camera path.

並列關鍵字

Monocular Visual Odometry ； Relative Orientation ； Accumulated Errors