BCA and HDB Singapore seeking AI-assisted drone technology for building facade inspection
Singapore’s Building and Construction Authority (BCA) and Housing & Development Board (HDB) have launched a Call for Proposal on using drones for Building Facade Inspection. This is supported by the Partnerships for Capability Transformation through Government Lead Demand initiative (Gov-PACT). The call is open to all local SMEs/startups, Institutes of Higher Learning (IHL) & Research Institutes (RI).
(Under Gov-PACT, the Government serves as the large organisation with which SMEs/startups will work with to develop and test-bed innovations. S$80 million was allocated for Gov-Pact under SPRING’s Partnership (PACT) programme as part of MTI-SPRING’s initiatives announced at the Committee of Supply (COS) 2017.)
The objective of this Call for Proposal is to develop an integrated inspection system, using advanced image-capturing drone and computing technologies, assisted by artificial intelligence (AI) to perform autonomous defect detection, based on image recognition and machine learning. This is expected to aid the automation of building façade inspection, improve workflow, promote collaboration between inspection parties as well as improving the process of reviewing and assessing the condition of building facade.
Building facades need to be regularly inspected and maintained to ensure that the façade elements (e.g. wall plasters, tiles and claddings) or any installations (e.g. awnings, sunshades and air-conditioner units) that attach to the facade will remain intact so that they do not fall and pose a danger to the people and properties below them. However, inspection of facades can be a challenging task and the problem is further aggravated by the increasing building height and complexity of the facade element.
Currently, inspecting building facades is a slow, resource and manpower intensive process. It usually involves a long, tedious and costly process that requires the inspector to work at height using scaffold, gondola, boom lift or even rope access. There are concerns over personnel safety. Moreover, the reliability of traditional inspection method relies heavily on experience and judgement of the inspector. Such inspection may also be delegated to lowerskilled workers who may not be equipped with the competency required for the inspection.
BCA and HDB feel that there is an urgent need of a new inspection model, using innovative solutions to improve the inspection efficiency, reduce safety risk to personnel, save time and at the same time being cost effective.
The scope of inspection under this Call for Proposal covers all types of building façades, exterior features, fixtures and appendages to the external walls which shall include, but not be limited to, plaster, concrete, tiles, siding, bricks, cornices; curtain walls; claddings; awnings and sun-shading devices and other architectural features and projections. Inspection priority shall be given to plaster, concrete, tiles and any exterior projections from the building footprint.
Condition of the facade in this context refers to its safety, stability and integrity. Quality and functionality of the facade i.e. energy loss and air leakage are beyond the scope of inspection.
The project can be categorized into 2 main phases – Phase 1 comprising of Phase 1A & 1B, and Phase 2 comprising of Phase 2A & 2B.
Image mapping and processing software developed concurrently with drone field trials
According to the Problem Statement document, Phase 1A consists of the development of drone image mapping & processing software. The software should provide mapping and image indexing for traceability (position, geolocation, date, time of survey, etc.).
Images should be organised to allow easy browsing and navigation, identify and mark the defects or anomalies, make annotation and present the results for reporting purpose. It is envisaged that the image databank would contain images and of various façade elements and defects commonly found in building facade. Daylight visual and thermal images should be integrated to be examined in tandem and providing side-by-side comparison.
The solution should also enable Visual 3D reconstruction of building for general building visualization and localisation of defects with the ability to pan, rotate and zoom. The 3D rendered model should be linked to an individual or group of high resolution images to allow closer examination by inspectors. The model should allow overlaying of defects or anomalies on its surfaces.
The program’s basic functions such as viewing and browsing of images, marking of defects and anomalies, and making annotations should be made accessible via mobile devices using apps.
The programs developed shall apply AI to perform autonomous defect detection, based on image recognition and machine learning. The proposed artificial intelligence module, together with the image databank of the defects, are to be developed as an extension (add-ons/plugins) to the program.
The aim is to design a program which could autonomously identify and analyse the facade elements and defects, determine and classify the anomalies according to the type, confidence level and measurement of the severity or magnitude. For Phase 1A, the defects required for auto-recognition should minimally include a) cracks, b) spalling (include early indication and tell-tale sign leading to spalling) on concrete, plaster and tile surfaces.
To ensure the accuracy of the defects detection, Inspector shall conduct close-up inspection to validate defect images collected during the field trials (under Phase 1B). The image recognition system shall be updated and retrained to improve the detection accuracy.
Phase 1B involves drone field trials to be conducted on 2 existing buildings, assigned by BCA / HDB, on 2 separate occasions. This will conducted concurrently with Phase 1A. Drones are to be deployed to perform survey on the condition on the entire façade areas of the building using both visible light and infrared thermal cameras (other alternative scanning technology can be proposed).
During Phase 2A, the autonomous defect recognition scope and capability of the program developed in Phase 1 will be extended to things such as void/ hollow/ delamination; efflorescence/discolouration on concrete, plaster and tile surface; and metallic corrosion. Additionally, in this phase, the defects have to be assessed, by measuring the magnitude and/or analysing the patterns and they have to be classified according to the severity of the defects.
In addition to visual imaging, Phase 2A should include the image recognition and analysis using thermal infrared imaging (or the adopted alternative scanning technique) in providing a more accurate detection of the defects. Applicants will conduct one field trial on an existing building to assist in developing the program and defect recognition capability. The defects should be verified by Inspector with a close-up site inspection and results further retrained.
Finally, in Phase 2B, one existing building will be assigned by BCA / HDB to demonstrate the effectiveness of the solution through a pilot implementation.