應用多來源影像進行影像式模型重建及精度評估指標建立

以往影像式建模的影像多從高解析度的相機取得，隨著智慧行動裝置的普及影像來源愈趨多樣化。本研究建立精度指標以焦距估算建模後的體積精度，快速評估各來源影像的模型精度。實驗使用各款智慧型手機與數位相機拍攝水泥柱以重建模型，將相機率定的焦距視為真值，評估SfM (Structure from Motion)的焦距精度，並計算模型體積與量測體積的誤差，接著以迴歸分析焦距與體積誤差，建立精度評估指標。實務上大多案例無法得知待測體積之真值與誤差，當使用任何具拍攝功能的設備獲取影像時，便可以透過本研究所建立精度評估指標，以焦距精度初誤待測物之體積差，作為後續分析資訊的參考依據。

關鍵字

影像式建模；相機率定；焦距

並列摘要

Most images for Image-Based Modeling (IBM) were taken from high resolution cameras in the past. With its widespread, cameras on smart mobile devices can be one type of image sources. The aim of this study is to build up an index of reconstruction quality which can calculate the error of the volume through the focal length. Thus, the index can estimate the accuracy of reconstructed models rapidly by multi-source images. This study uses various kinds of smartphones and a digital single lens reflex (DSLR) to take photographs of concrete specimens and to reconstruct 3D models. Focal lengths calculated by camera calibration are considered as the true value to estimate focal lengths accuracy by structure from motion (SfM) and then the volume error of the 3D models. Finally, the accuracy index of reconstructed models was established through a linear regression based on focal lengths and volume error. Owing to unknown true volume in most actual conditions, it becomes convenient to apply various mobile devices with the camera to take photographs and reconstructed models. According to the accuracy index, the error of earthwork quantity (Y) can be estimated by focal length (X), and the result can be the basis of disaster analysis and prevention measures aprior.

並列關鍵字

Image-Based modeling ； camera calibration ； linear regression

國際替代計量

應用多來源影像進行影像式模型重建及精度評估指標建立

全文下載

主題瀏覽