中華民國內政部建築研究所中華民國內政部建築研究所

一、研究緣起
因應近年來人工智慧深度學習技術在電腦視覺、影像辨識上之快速發展，其技術核心之深度學習運算平台，也從雲端或伺服器平台延伸到人工智慧物聯網(AIoT)感測器上。使得原本必須透過雲端平台進行決策判斷之作業，已經可以由AIoT感測器端的邊緣運算來提供即時的影像辨識結果，不但加速決策之即時性，亦有助於改善資訊之安全性。本研究旨在探討智慧建築物導入AIoT 影像感測器之效益，並探討如何應用AIoT提升智慧建築之效能，以了解AIoT在未來智慧生活空間之應用潛力。
二、研究方法與過程
本計畫之主題為｢AIoT影像感測器在建築物安全應用與效益研究｣，其主旨包括：(1)透過文獻與技術現況回顧，調查國內外應用人工智慧物聯網(AIoT)影像感測技術於建築安全應用之案例、效益分析及相關資料；(2)提出AIoT影像感測器建築安全應用情境，說明人臉、車牌、寵物等影像辨識服務導入流程、因應隱私權保護課題之注意事項及推估可提升之建築物營運效益；(3)最後，本研究提出｢建築智慧科技應用成本效益分析模式(CAMITA)｣作為AIoT影像感測器建築安全應用之分析方法，再透過案例訪談及案例模擬試算方式，推估比較AIoT及傳統非AIoT影像感測器方案在建築物安全應用之量化及質化效益差異。
三、重要發現
經本研究調查與分析，獲致以下之重要結果與發現：
(一)完成國內外應用人工智慧物聯網(AIoT)影像感測於建築安全應用之案例及效益分析等相關文獻資料之蒐集與調查—本研究透過次級資料分析，完成AIoT影像感測相關軟硬體技術發展現況之回顧、AIoT影像感測技術於建築安全監測之應用之回顧、以及AIoT影像感測技術於建築安全監測之應用效益評價方法之回顧。【詳本報告｢第二章 資料蒐集與文獻分析｣】
(二)完成並提出AIoT影像感測器於建築各生命週期階段安全應用之可能應用情境及效益分析—本研究透過次級資料分析、專家訪談與效益分析模型設計，歸納AIoT影像感測器於建築生命週期各階段之18種可能應用情境類型，包括建築施工中之工地安全監控、施工中之結構安全監測、完工後之建築物安全監測、以及建築物使用營運階段之社區安全管理等；並提出｢建築智慧科技應用成本效益分析模式(Cost/benefit Analysis Model for Intelligent building Technology Adoption, CAMITA)｣之建構。透過決策實驗室分析法(DEMATEL)建立效益分析之層級架構，再經由網路層級分析法(ANP)完成CAMITA量化效益分析模式。本CAMITA分析模式雖以AIoT影像感測器於建築安全應用為標的，然其分析架構可延伸應用於其他智慧建築科技之效益分析。【詳本報告｢第三章 AIoT影像感測器建築安全應用情境及效益分析架構之研擬｣】
(三)完成兩個AIoT影像感測器在建築物安全應用效益之案例模擬試算與專家訪談—依據CAMITA分析模式，選擇兩個AIoT影像感測器在建築物安全之應用場域，作為本研究之模擬案例：(1) AIoT影像感測器應用於建築施工中工地安全之監控，共包含四種應用情境、(2) AIoT影像感測器應用於智慧社區之安全管理，共包含五種應用情境，以進行效益模擬試算以及專家訪談，並完成應用效益與潛在問題分析。經過三階段之專家訪談與模擬案例效益評估及模擬試算發現：模擬案例(一)｢建築施工中工地安全監控｣應用案例，淨現值指標(NPVI)可達到20.21%，而益本比(BCR)則可達到5.10；另外，在模擬案例(二)｢建築使用營運階段之智慧社區安全管理｣之應用案例，其淨現值指標(NPVI)可達到效益12.63%，而益本比(BCR)則可達到2.10；可見AIoT影像感測器在建築物安全應用確實具有明顯效益。然而透過專家訪談進行質性效益與可能潛在問題之分析亦發現，AIoT影像感測器於建築物安全應用也存在著資訊安全與個人隱私之潛在社會風險。不同利害關係人對於AIoT影像感測器於建築物安全應用之期待效益、成本接受度、未來發展潛力，以及社會風險之感受度亦皆有不同。本研究最後針對建築安全之利害關係人對於AIoT影像感測器於建築物安全應用之優先關切課題、期望效益以及可能面臨的限制與困難等，歸納出未來智慧建築與智慧生活空間應用AIoT影像感測器技術之具體建議，以提供建築工程業主、建築師、相關專業技師、施工廠商、社區管理委員會及保全等不同單位使用者在採用此一技術時之參考。【詳本報告｢第四章 模擬案例(一)：AIoT影像感測器於建築工地安全監控之應用｣及｢第五章 模擬案例(二)：AIoT影像感測器於社區安全管理之應用｣】
四、主要建議事項
(1)AIoT影像感測創新技術於社會住宅智慧建築應用之研究：立即可行之建議
主辦機關：內政部建築研究所
協辦機關：國家住宅及都市更新中心、各縣市社會住宅開發與規劃單位
政府規劃在未來八年內完成20萬戶只租不售之社會住宅，以解決青年及經濟弱勢族群在都會地區之居住問題。截至2020年底，正在興建及完工之社會住宅共計四萬六千餘戶，預計未來由政府主辦興建之社會住宅將達到12萬戶，如此龐大之住宅新建計畫提供智慧建築創新技術最佳之實證應用場域。經由本研究結論發現，應用AIoT影像感測器於社會住宅之社區安全管理可達到極高之經濟效益。尤其未來社會住宅之規模有可能朝向小型化 (戶數在35~80間)，若採用傳統24小時人工保全系統進行管理，勢必提高住戶之保全管理費用負擔，如此將與社會住宅開發之初衷(減輕青年及經濟弱勢族群之經濟負擔)相違背。若能透過AIoT影像感測器之導入，則可以大幅降保全管理人力需求；尤其是夜間之巡檢，甚至可以達到無人化之智慧社區管理目標，大幅減低住戶保全管理費用之負擔，實為立即可行之政策。
(2)AIoT影像感測創新技術於建築施工安全與災害監測應用之研究：立即可行之建議
主辦機關：國家住宅及都市更新中心、各縣市社會住宅工程主辦單位
協辦機關：內政部建築研究所
延續前一項建議，本研究之結論發現，應用AIoT影像感測器於社會住宅施工過程之工地安全管理，對於減低工地意外發生，改善人員安全，具有明顯效益。除此之外，在國外之建築安全應用中，亦有將AIoT影像感測器應用於施工中之鄰房下陷或傾斜之監測、工地水/火災之監控，以及完工後建築物之結構位移監測、住宅邊坡安全監測、建築物表面結構損壞檢測、外牆損壞檢測等應用案例。建議可以以我國堅實之資訊產業基礎結合目前正在大力興辦的社會住宅工程為策略夥伴，進行上述AIoT影像感測器應用之測試與實證，以建立本土研究範例與數據。如此，不但可以鼓勵國內AIoT影像感測器相關產業投入研發以提升我國智慧建築科技水平外，對於建立相關產業鏈、行產品驗證與技術提升，以及長期之智慧建築設備生產輸出等，皆有所助益。蓋因社會住宅完工後，將失去絕佳之技術實踐場域，因此，本建議事項亦具有立即之時效性。
(3)建築物應用AIoT智慧物聯網技術效益評估之實證研究：短期可行之建議
主辦機關：內政部建築研究所
協辦機關：各大學院校及學術研究機構
本研究之一重要成果在於參考國外學術文獻方法，提出｢建築智慧科技應用成本效益分析模式(CAMITA)｣。此一模式之分析架構並不侷限於AIoT影像感測器，而可適用於所有智慧建築科技應用之成本效益分析。其所得到之淨現值指標(NPVI)與益本比(BCR)指標，皆屬經濟決策分析之效益指標，不但提供業主或使用者經濟決策之支持數據，更可與國際相關研究之數據互相比較。然而目前國內對於創新智慧建築科技之應用，尚未累積足夠之參考數據，以支持公私部門建築工程業主對於採用創新技術與否之決策判斷；因此，建議未來建研所可以參考CAMITA架構，進行更多智慧建築與智慧社區之實證案例成本效益分析研究，以累積更多本土案例數據資料，供公私部門相關單位決策參考。此外，本研究僅進行模擬案例之分析，有關AIoT影像感測器實際應用時與建築自動化系統(BAS)之連動、促發與回應系統間之連動整合，以及此一整合系統所可能產生之加成效益，建議可以於後續實證研究中進行詳細規劃與驗證。
(4)AIoT影像感測器於智慧建築智慧創新應用與相關資訊安全規範之研究：短期可行之建議
主辦機關：內政部建築研究所
協辦機關：各大學院校及學術研究機構
我國｢智慧建築標章｣之｢智慧建築評估手冊｣八大評估項目中前七項目之評估項目相對完整，唯有｢8.智慧創新｣項目，目前尚未有明確而詳細之規範。經由本研究發現，AIoT影像感測器於建築安全之應用不但涵蓋｢智慧建築評估手冊｣中第5項｢安全防災｣之所有項目，且其設備建置，亦牽涉到｢1.綜合佈線｣與｢2.資訊通信｣；而健康照護應用其境，亦涵蓋｢7.健康舒適｣。有鑑於AIoT等智慧資訊設備在智慧建築與建築安全領域應用之快速發展，提供未來智慧生活空間更多創新之想像，對於人居生活亦產生革命性之改變。然而，所有創新技術之發展皆需要有足夠之養分與支持環境，尤其是關乎公共安全與人民權益之建築科技，更需要政府有力的支持與相關資訊安全及隱私法令規範監督。因此，如何將AIoT影像感測器這類突飛猛進之智慧科技，應用於未來居住生活空間之智慧創新應用，並制定相應之資訊安全與個人隱私規範，值得政府學術研究單位投入研究。一來可以充實｢智慧建築評估手冊｣第八項｢智慧創新｣之實際內涵，另一方面則可進一步更新與完善其他評估項目之評估內容。
(5)AIoT影像感測器於智慧建築智慧創新應用人才培養及產學研合作推廣：短期可行之建議
主辦機關：台灣建築中心
協辦機關：各大學院校及學術研究機構、AIoT設備產業
本研究經由兩個模擬案例－模擬案例(一)：AIoT影像感測器於建築工地安全監控之應用，模擬案例(二)：AIoT影像感測器於社區安全管理之應用－均證明AIoT影像感測器於建築生命週期不同階段較目前傳統方案具效益。但經由三階段專家訪談暨問卷調查及兩場專家座談會，發現對於系統介面友善性、成本透明度等均尚有發展空間。然而，有鑑於後續推廣AIoT等智慧資訊設備在智慧建築與建築安全領域之應用，須培養營建安全與資訊工程雙專長人才。因此，由政府、研究機構、學術研究單位、營建產業及物業管理單位投入研究與實踐，一來可提升更佳的建築智慧化效益，更可藉此培養出產業實務人才。

A.Research Background
In response to the successful commercial implementations of artificial intelligence deep learning technologies in computer vision and image recognition in the last decade, the core deep learning computation device has shifted from the traditional cloud or server platform to Artificial Intelligence Internet of Things (AIoT) sensors. Instead of cloud computing platform, real-time image recognition can be performed directly through edge computing on the AIoT sensor. Such a paradigm shift has not only improved the efficiency of image recognition and decision-making, but also enhanced the information security of image transfer. The current research aims to explore the applicable scenarios and the associated pros & cons of implementing AIoT image sensing in smart buildings to realize the potential benefits of AIoT in future smart living environments.
B.Research Method and Procedure
The topic of this study is entitled “Application and benefit evaluation of AIoT image sensing for building safety monitoring”. To achieve the planned objectives, the following research tasks are conducted as follows: (1)the previous works published in the literature on the application and benefit evaluation of AIoT image sensing are reviewed to address the potential application scenarios of AIoT image sensing to safety monitoring of building, e.g., face recognition, car license plate recognition and missing pet finding; (2)development of the method for benefit evaluation of the associated application scenarios, discussion on data privacy and security issues related to AIoT image sensing, and the use cases simulation and benefit evaluation of AIoT image sensing application; and (3)a "Cost/benefit Analysis Model for Intelligent building Technology Adoption, CAMITA)” for benefit analysis of AIoT image sensing application in building is proposed for comparison of both quantitative and qualitative benefits between AIoT and non-AIoT alternatives for building safety monitoring of the simulated use cases.
C.Important Results and Findings
The research team attained the following results and important findings for the project:
(1)Review of the relevant research works, including: use cases and associated benefits of AIoT image sensing in building safety both in Taiwan and other countries, the most advanced AIoT image sensing related software and hardware technologies, the application of AIoT image sensing technology in building safety monitoring, and the benefit evaluation method for AIoT image sensing application in building safety monitoring. [Please refer to Ch. 2 of this report for details.]
(2)Summarization of the 18 applicable scenarios and development of a benefit analysis method for AIoT image sensing application in building safety monitoring, including the site safety monitoring of construction process, the safety monitoring of building structure during construction, the building structural safety monitoring after construction, and the community safety management during usage and operation stage. A preliminary "Cost/benefit Analysis Model for Intelligent building Technology Adoption, CAMITA)” for benefit analysis of AIoT image sensing application is also proposed. The proposed CAMITA model comprises of two analysis stages: firstly, the Decision Making and Trial Evaluation Laboratory (DEMATEL) is used to construct the framework required for cost/benefit analysis; then the Analytic Network Process (ANP) is adopted to determine the evaluation weights of the assessment items of the cost/benefit analysis framework constructed previously. Although the proposed CAMITA model was developed to evaluate the benefit of AIoT image sensing application in building safety monitoring, its application can be extended to evaluate the benefit of other smart building technologies. [Please refer to Ch. 3 of this report for details.]
(3)Completing of the trial calculation of the benefits of two AIoT image sensing use cases, including: (1) Case (I)—AIoT image sensing for construction site safety monitoring—including 4 different application scenarios; (2) Case (II)—AIoT images sensing adoption in the security control of smart community—including 5 different application scenarios. After a 3-stage domain experts interview and questionnaire survey, it is found that, when compared to traditional building safety monitoring alternatives, the AIoT image sensing alternative is more beneficial with Net Present Value Index (NPVI) of 20.21%, and Benefit/Cost Ratio of 5.10 in Case (I), and with NPVI of 12.63% and BCR of 2.10 in Case (II). The interviews with the domain experts also pointed out that there exist potential threats and risks while adopting AIoT image sensing in building safety monitoring. Different interviewees (stakeholders of buildings) have different perspectives on the benefits and risks of such an innovative technology from the angles of their priority issues, expected benefit, application limitations, and implementation obstacles. Finally, application recommendations of AIoT image sensor technology in smart buildings and smart living spaces are summarized to provide a reference for future users of such a technology, e.g., construction project owners, architects, related professional consultants, construction firms, community management committees, security, and other users. [Please refer to Ch.4 & Ch.5 of this report for details.]
D.Suggestions
Suggestion I:“Research on the Application of AIoT Image Sensing Innovative Technology in Social Residential Intelligent Buildings”—could be adopted immediately.
•Principal Executor: The Architecture and Building Research Institute (ABRI)
•Associated Executor: National Housing and Urban Regeneration Center (NHURC), the social housing development and planning agencies of cities and counties.
The government has planned to build 200,000 units of social housings to solve the accommodation problem of young people and economically disadvantaged groups in metropolitan areas. Such a huge residential project offers the best empirical field for the application of smart building innovation technologies, including AIoT sensors. The application of AIoT image sensors in the safety management of small social housing communities of households between 35 and 80 can achieve extremely high economic benefits by reducing the cost of the traditional 24-hour manual security system. In particular, AIoT image sensors for night safety inspections can reach the goal of unmanned smart community management, which greatly reduces the burden of household maintenance and management costs.
Suggestion II:“Application of AIoT Image Sensing Innovative Technology in Construction Safety and Disaster Monitoring”—could be adopted immediately.
•Principal Executor: National Housing and Urban Regeneration Center, the social housing development and planning agencies of cities and counties.
•Associated Executor: The Architecture and Building Research Institute (ABRI).
It is found in this study that the application of AIoT image sensors in the construction site safety management of social housing construction has significant benefits in reducing accidents on the site and improving personnel safety. Moreover, the AIoT image sensor application cases in construction safety of other countries also suggested that other potential fields such as monitoring the sinking or tilt of neighbor houses during construction, monitoring water/fire at the construction site, and monitoring structural displacement after the completion of the building, monitoring slope safety, detecting building surface structure damages, and detecting exterior wall deterioration. Such applications can be strongly supported by the highly competitive Information and Communications Technology (ICT) industries of Taiwan. By incorporating the smart building industry with the ICT industry, it will help for the establishment of related industrial chains, product verification and technology improvement, and long-term production and export of smart building equipment and products. As the field of excellent technical practice will be lost after the completion of the social security housing, this proposal should be considered immediately.
Suggestion III:“An Empirical Study on the Benefit Evaluation of Building Application AIoT Smart Internet of Things Technology”—could be adopted in the short term.
•Principal Executor: The Architecture and Building Research Institute (ABRI).
•Associated Executor: Universities and Research Institutes.
One of the important outcomes of this project is the proposal of CAMITA benefit analysis model for smart building technologies. Such a model can be applied to evaluate other smart building technologies. The cost/benefit indexes derived with CAMITA can be compared with the other literature, the domestic application data for innovative smart building technologies has not yet been accumulation to support the construction project owners in the public and private sectors in deciding whether to adopt innovative technologies. Our research suggested applying CAMITA model for evaluating more public smart building projects to accumulate more domestic reference data for the decision-makers in public and private sectors. Moreover, this research only analyzes simulated cases. The linkage between AIoT image sensors and the Building Automation System (BAS) and the potential benefits of the integration between AIoT sensors and the triggering and response systems are not researched yet. It is recommended to carry out detailed planning and verification in follow-up empirical research.
Suggestion IV:“Research on the Innovative Applications of AIoT image Sensing Technology for Smart Buildings”—could be adopted in the short term.
•Principal Executor: The Architecture and Building Research Institute (ABRI).
•Associated Executor: Universities and Research Institutes.
The first 7 evaluation items of the 8 major evaluation items in ABRI’s "Smart Building Evaluation Manual" are relatively complete, only the "8. Smart Innovation" project has no clear and detailed specifications. It is found from the study that the application of AIoT sensors covers wider than that defined in the “Item 5: Disaster Prevention” of the ABRI’s "Smart Building Evaluation Manual". It is also relevant to “Item 7: Health and Comfort”. In view of the rapid development of the application AIoT in the field of smart buildings and building safety, it offers unlimited imaginations for future smart living spaces and also brings revolutionary changes to human life. However, the development of all innovative technologies requires sufficient nutrients and supporting environments, especially construction technologies involving public safety and people’s rights. It requires strong government support and relevant information security and privacy laws and regulations. Therefore, how to apply intelligent technologies such as AIoT image sensors to future residential living spaces, and to formulate corresponding information security and personal privacy regulations, is a topic worthy of research by government academic research organizations. We suggest the relevant organizations can enrich the contents of the "Item 8: Smart Innovation" of the "Smart Building Evaluation Manual", and they can further update and improve the evaluation content and the items of the "Smart Building Evaluation Manual".
Suggestion V:“Academia-Industry Collaboration on Training and Development of AIoT-IS Application Talent”—could be adopted in the short term.
•Principal Executor: Taiwan Architecture and Building Center (TABC).
•Associated Executor: Universities, Research Institutes, and AIoT-IS Manufacturers.
Both of the two simulated case studies have confirmed that AIoT image sensors are more effective than traditional solutions at different stages of the building life cycle. The research was also found through three-stage expert interviews and questionnaire surveys and two expert forums that there are still improvements and developments to be made in terms of system interface friendliness and cost transparency. In order to promote the application of intelligent information devices such as AIoT in the field of smart buildings and building safety, it is necessary to cultivate talents with dual expertise in building safety and information engineering. Therefore, research investment and practice application by government agencies, research institutions, universities and colleges, construction industries, and property management contractors can cooperate to improve the benefits of smart building and cultivate practical talents in the industry.