基於電腦視覺與BIM輔助工地現場室內定位之方法

Construction site indoor positioning using computer vision and BIM

黃伯凱(Bo-Kai Huang)

指導教授 ： 謝尚賢

國立臺灣大學/工學院/土木工程學研究所/碩士(2020年)

https://doi.org/10.6342/NTU202003403

摘要

近年室內定位在無線通訊發展下產生許多良好的應用及技術，然而營建施工現場因通訊設備通常不完善，將使得依靠通訊設備的室內定位技術無法被正常使用在施工現場，因此本研究認為基於影像之室內定位技術在施工現場應具有更好之可行性。
本研究針對施工現場之環境限制，提出一套基於電腦視覺與BIM輔助工地現場室內定位之方法，方法流程為：(1)透過BIM模型產生大量場景影像，並且使用預訓練之深度學習模型提取影像之特徵值，建立BIM場景影像資料庫；(2)拍攝施工現場影像後，同樣使用深度學習模型提取其特徵值，並與BIM影像資料庫進行相似度匹配，計算得到最相似影像之空間位置；(3)透過特徵點對解算查詢影像以及最相似影像之本質矩陣，進而得到拍攝相機位置與姿態；(4)將BIM模型投影至施工現場，並透過視覺慣性里程計不斷計算使用者之相對位置與姿態；(5)透過施工現場偵測到之水平面及垂直面，持續校正位移及姿態之觀測誤差。最終經案例測試後，驗證確實可達到定位之效益，本研究亦針對實驗結果改善所提之方法，使方法更符合施工現場之應用可行性。

關鍵字

電腦視覺 ； 室內定位 ； BIM ； 深度學習

並列摘要

Due to the development of wireless communication, indoor positioning has produced many good applications and technologies. However, the lack of communication equipment at the construction site, there's indoor positioning technology that can not be used normally. Therefore, this study believes that image-based indoor positioning technology should have better feasibility at the construction site.
In consideration of the environmental limitations of the construction site, this study proposes a method that based-on computer vision and BIM for construction site indoor positioning. The method consists of the following: (1) Generate a lot of images through the BIM model, and use the pre-trained deep learning model to extract the image features, and establish a BIM images database; (2) Taking a real image on the construction site and using the deep learning model to extract the features of the image and calculate the similarity with BIM images database, then the user get the most similar image; (3) Estimate the essential matrix of the real image and the most similar image through feature point pairs to obtaining the camera pose; (4) Project the BIM model to the construction site with the mobile device, and continuously calculates the relative camera pose of the user through the visual-inertial odometry ; (5) Detect Horizontal and vertical planes from the construction site, and continuously adjust the observation error. Finally, after the case study, it was verified that the benefits of indoor positioning can be achieved in the construction site. This study also improving the proposed method to make the method more in line with the application feasibility of the construction site.

並列關鍵字

computer vision ； indoor positioning ； BIM ； Deep Learning

參考文獻

Acharya, D., Khoshelham, K., Winter, S. (2019). BIM-PoseNet: Indoor camera localisation using a 3D indoor model and deep learning from synthetic images. ISPRS journal of photogrammetry and remote sensing, 150, 245-258.

Acharya, D., Ramezani, M., Khoshelham, K., Winter, S. (2019). BIM-Tracker: A model-based visual tracking approach for indoor localisation using a 3D building model. ISPRS Journal of Photogrammetry and Remote Sensing, 150, 157-171.

Acharya, D., Singha Roy, S., Khoshelham, K., Winter, S. (2019). Modelling uncertainty of single image indoor localisation using a 3d model and deep learning. ISPRS Annals of Photogrammetry, Remote Sensing Spatial Information Sciences, 4.

Andrews, H., Patterson, C. (1976). Singular value decompositions and digital image processing. IEEE Transactions on Acoustics, Speech, and Signal Processing, 24(1), 26-53.

Atzori, L., Dessi, T., Popescu, V. (2012, May). Indoor navigation system using image and sensor data processing on a smartphone. In 2012 13th International Conference on Optimization of Electrical and Electronic Equipment (OPTIM) (pp. 1158-1163). IEEE.