光流應用於自我移動估測

Abstract

自我移動估測對於載具的自動控制是必要的資訊之一，而隨著電腦視覺領域的相關技術應用於載具愈來愈廣泛，若能開發出以電腦視覺為基礎的自我移動估測系統，不僅能夠在傳統自我移動估測方法無法使用的情況下也能工作，而且也可以用來搭配其他電腦視覺的應用，因此本研究的目標為開發基於電腦視覺的自我移動估測系統。 本研究目標為以影像光流法來實現即時的自我移動估測，因此本文會探討不同影像光流的計算方法，找出最適合即時運算的方法，然後描述如何利用相機模型，將影像移動量轉換成實際的移動量，最後再以實驗驗證系統的性能表現。

Egomotion estimation is an essential information for autopilot. As the massive development of computer vision in the modern world, there are a lot of applications for vehicles which based on computer vision. There are many advantages to develop egomotion estimation technique based on computer vision. For example, it can be used for egomotion estimation when traditional methods are not available. At the same time, it can cooperate with other computer vision techniques. Therefore, we propose an egomotion estimation system based on computer vision in this research. The goal in this research is to develop a real time egomotion estimation system via optical flow. In this paper, we discuss and compare different calculation methods of optical flow to find the moderate technique for our objective. Also, we describe how to transform optical flow to physical translation based on camera model. Finally, we design an experiment for the performance verification.