Enhancement of Detection of Defective Roadside Fixtures Using AI Computer Vision Technology

I&T Wish Enhancement of Detection of Defective Roadside Fixtures Using AI Computer Vision Technology
(REF: W-0513)
Matched I&T Solution
Trial Project
Summary and Challenges
  • This project aims to develop a risk probability indicator system for the roadside fixtures (e.g. signboards, road signs, traffic lights, road lamps, or even trees). Inspections should be conducted regularly and after typhoon in order to spot out those defective objects which may pose threats to public safety, so that the authority could take immediate action to fix it.
  • In this project, "defective" defines as deformed/ distorted/ detached (with tilted angle) and corroded/ rusted.
Expected Outcome The risk probability indicator system utilizes image/video analysis (computer vision) and consists of three main modules:
  1. Identification of roadside fixtures: This module determines whether the roadside fixture in the street image is normal or defective.
  2. Defective factor analysis: This module identifies and quantifies the defective factors present in the fixtures. It includes sub-modules such as: 
    1. Deformed factor analysis: A model compares the deformed roadside fixture with a fixed object (i.e. ground) to determine the probability of deformation, considering the tilted angle of the fixture.
    2. Corrosion detection: A model trained on general corrosion data analyzes the image or other reliable means to distinguish rust and assess the level of corrosion. This helps determine the probability of a corroded fixture.
  3. Risk indicator consolidation: The probabilities obtained from the previous modules are consolidated into a risk indicator. This indicator represents the likelihood of defective roadside fixtures and enables the corresponding party to take appropriate action. In the expected system, the three modules should be designed in modular basis (via innovative system architecture) with the flexible of replacing one/all modules for serving different purpose of video-analytic based risk assessment in subsequent stages of the system development. Additionally, implementing edge devices is considered to reduce image processing time to generate risk probability of roadside fixture with location and highlighted in corresponding street view.
  4. Remarks: Solutions that involve software only, software and hardware, or turnkey solutions are welcome as long as they address the problem statement.
Expected Trial Duration 12-month
Contact Information
I&T Wish Proposer:Electrical and Mechanical Services Department (EMSD)
Contact Person:CHAN Wing Chung, Simon
Position:Information Technology Manager/Artificial Intelligence/1
Tel:3167 2895
Email: simonchan@emsd.gov.hk
Upload Date 2023-11-01
Closing Date 2023-11-15
Propose I&T Solution Closed