結合影像暨感測器資訊之三維模型重建研究

Abstract

三維重建(3D Reconstruction)技術是利用多張多視角影像，將二維投影恢復物體三維空間之方法，類似人類雙目定位概念；從平面影像還原成立體模型如真實物體，表現出更豐富的細節資訊，而三維模型的呈現實現於生活廣泛之應用。 本研究以行動裝置為平台，對物體進行環繞拍攝取樣，透過運動恢復結構之影像演算法，無須事先校正相機參數，即計算出相機姿態與場景幾何相對關係；此外，加上感測器的地理資訊輔助，其強健穩定的特性，二次驗證定位方法，對齊校正座標，增加三維模型之精準度、提高運算效能。

3D reconstruction is the process of capturing the shape and appearance of real objects from the keyframe of different viewpoint. Through the projection of two-dimensional materials to restore three-dimensional space, which is similar to a binocular vision for the position. Nowadays, a 3D model is implemented in many applications, from an image reverts into a stereoscopic model as the original real object, that can be given more details of texture and structure.

First, based on the mobile device to scan around the object for video recording, using structure from motion(SfM) algorithm to calculate the relationship of camera position and scene geometric. Meanwhile, at the scanning stage, the sensor data are acquired along with tracks of features in the video. All these data are used to build a camera trajectory using above image techniques after scanning is completed. According to information support of sensor geography with robustness and stability, which can be demonstrated the second validation on positioning, not only enhance the accuracy of the 3D model, but also improve the efficiency.