應用影像辨識技術於營造業工程進度管控之研究

Abstract

目前營造業之使用3D相關技術發展已漸臻成熟，但實務上應用多侷限於初期規劃設計，現階段僅作為與業主溝通或準備投標文件時的單純視覺化工具。規劃設計階段所建立之3D模型，於施工中未能配合實際情況進行即時更新，無法達到進度監控目的。本研究所發展之影像辨識比對匹配技術，可比對工地拍攝之施工影像與設計階段建置之3D模型影像，藉由影像間建築元件的差異，以協助專案人員瞭解工程實際進度與預期進度之落差。 本研究視覺化進度監控模式之建立程序如下：(1)辨識建築元件：分離影像背景與前景，辨識出工地照片以及3D模型照片不同時間點的新增建築元件。(2)比對物件特徵：比對、匹配兩組照片的前景建築物件之特徵，作為比較預期進度與實際施工進度差異依據。(3)分析容許誤差範圍：以合成之理想條件工程施工照片，確立影像辨識誤差容許範圍。(4)案例實證：實際驗證工地施工進度照片，建立影像辨識進度管控應用模式。 國內營造業目前未見以視覺化方式管控施工進度，透過本研究建立之影像辨識邏輯連結工程進度管理，可半自動比對施工現況與預計進度差異，提供即時進度資訊，期能達到自動化電腦視覺管控工程進度之目的。

Although the 3D technology has been researched and developed during last two decades in construction industry, the application of 3D are still limited to the planning and designing stage of a construction project. 3D model are mainly used as visual-based tool for contractor to communicate with owner and prepare for bidding document. The schedule information in the 3D system cannot fulfill the need of monitoring the progress in the construction field because it cannot be updated in real time. This research tend to apply the well-developed image recognition and matching algorithms to compare different construction unit appearing in photos shoot in construction field with images retrieved from 3D as-planned model. The result of comparison can demonstrates the schedule difference in field and early planning expectation. The research is conducted as following procedure: 1. Identify the construction unit appearing in both field photos and 3D images. 2. Comparison of characteristic description of every object identified. 3. Decide the allowed error range for each description factors. 4. Examine real construction photos and establish visual-based construction progress monitoring model. In Taiwan, there are few researches related with visual-based progress monitoring in construction field. This research constructed a progress measurement model through semi-automatic comparison of field photos and 3D images applying image recognition technology. The results are expected to provide the project manager with real-time information about actual field completion percentage and automatic measurement of construction progress.