Computer Vision applications toward Offshore HSE goals€¦ · • Metal-Mechanic Solutions •...
Transcript of Computer Vision applications toward Offshore HSE goals€¦ · • Metal-Mechanic Solutions •...
Grand Hyatt Rio de Janeiro • Rio de Janeiro, Brazil • 3-4 March 2020
Computer Vision applications toward Offshore HSE goalsPaulo Roberto FurioFIRJAN
© Global Offshore Brazil Summit and Gulf Quest LLC
SESI Innovation Centers
2
• Microelectronics
• Biomass
• Electrical Chemistry
• Structure engineering
• Longevity and Productivity
• Laser Processing
• Embedded systems
• Manufacture Systems
• Health Technologies
• Polymer Engineering
• Metal-Mechanic Solutions
• Psychosocial Factors
• Mineral
Technologies
• Information and Comms
Technologies
• Materials union
• Production Automation
• Logistics
• Incapacity prevention
• Surface engineering
• Metallurgy and special alloys
• Mineral processing
• Electrical Systems and Equipments
• Ergonomics
• Green Chemistry
• Virtual Production Systems
• Material Integrity and Inspection
• Biosynthetics
• Occupational Hygiene
• Advanced Manufacturing
• Advanced Materials
• Biotechnology
• Wellbeing
• Renewable Energy
SESI SENAINetwork
Economy for Health and Safety
• Worker Safety
Management Systems
© Global Offshore Brazil Summit and Gulf Quest LLC
SESI Innovation Centers
• The offshore setting is a dynamic and challenging workplace where it is of utmost
importance having well defined policies to ensure HSE goals.
• Such challenges include the movement of heavy loads, the handling of chemicals
and the risk of leaks and explosions.
• In this presentation we discuss three potential computer vision applications to
tackle these challenges:
- PPE and Ergonomics evaluation through Object Detection and Pose
Estimation
- Exposure timers
- Multispectral imaging for detection of leaks and spillage
3
© Global Offshore Brazil Summit and Gulf Quest LLC
Worker Evaluation in Real Time
4
© Global Offshore Brazil Summit and Gulf Quest LLC
Worker Evaluation in Real Time
• Offshore work presents challenges such as:
o fire, after ignition of released hydrocarbons;
o explosion, after gas release, formation and ignition of an explosive cloud;
o oil release on sea surface or subsea
• Adherence to PPE policies are key factors influencing worker safety in
hazardous conditions.
• Workplace PPE checklists may be use to guarantee worker compliance
5
✓✓✓
© Global Offshore Brazil Summit and Gulf Quest LLC
Worker Evaluation in Real Time
• Computers can be taught to detect objects and monitor workplace compliance to PPE policies.
• Such checklist procedures can be automated and implemented through computer vision methods such as object detection and instance segmentation
6
© Global Offshore Brazil Summit and Gulf Quest LLC
Real time PPE Detection
7
Instances of objects are classified
through the identification of
features of a given class through
machine learning techniques
© Global Offshore Brazil Summit and Gulf Quest LLC
Worker Evaluation in Real Time
Such approaches can be extended to also be used to evaluate human posture through the technique of pose estimation.
This opens up approaches that can evaluate and model ergonomics in the the offshore work environment.
8
© Global Offshore Brazil Summit and Gulf Quest LLC
Exposure Timers
9
© Global Offshore Brazil Summit and Gulf Quest LLC
Exposure Timers
• Some workplaces present challenges that required a well controlled amount of exposure to hazards such as loud noises or intense heat.
• Computer vision methods allow defining these areas and setting policies for timed exposure. Once a worker enters the exposure area, the computer will count the time and set off an alarm when the defined timer ends.
• That increases worker safety by enforcing a strict time policy around high risk exposure zones. More than one area can be defined, each with its own policy or set of policies to be followed.
10
Area
definition
Background
subtraction
Object
Identification
Exposure
Monitoring
© Global Offshore Brazil Summit and Gulf Quest LLC
Exposure Timers
11
• By using more than one camera set at once, it is possible to more accurately evaluate depth
• This allows for the building of dynamic “no-go.zones”, such as areas in the offshore where
overhead maneuvering of heavy loads may present a hazard.
• By checking the load’s shadow in respect with the ground, it is possible to define restricted areas under the load path.
© Global Offshore Brazil Summit and Gulf Quest LLC
Multispectral Imaging for hazard detection
12
© Global Offshore Brazil Summit and Gulf Quest LLC
Multispectral Imaging for hazard detection
• Light has many different interactions with materials depending on its frequency - i.e.: IR can signal heat signatures
• Through the use of Hyperspectral cameras, we can signal hazards such as small oil spills by defining sharp contrasts.
• By identifying such hazards a computer can quickly signal the occurrence of unwanted events and ensure a faster response time.
13
Vector map
representing
image edges
© Global Offshore Brazil Summit and Gulf Quest LLC
Thank you
14