DIGITAL AI-POWERED
SURVEILLANCE ANALYTICS
IN OSH MANAGEMENT
PROOF OF CONCEPT (POC) OUTCOME REPORT UPDATE
PRESENTATION TO: Management Team
ORGANIZATION: Malaysia LNG Sdn. Bhd. (MLNG)
“Verifying Practical Capability of Digital AI Surveillance Analytics”
Presented by
Ir. Ts. Michael Anggie
Director, Sumira Engineering
AI
MLNG.POC
About Sumira Engineering
Established in 2021, Sumira is a Bumiputera-owned Sarawakian engineering and technology company founded and led by a team of siblings with complementary engineering and industry expertise.
Our multidisciplinary foundation brings together engineering knowledge, industry experience and practical project capabilities, enabling us to support end users, contractors and strategic industry partners through engineering design, project execution, technology integration and operational support.
11 Multidisciplinary Capabilities
Our Core Focus
DIGITAL AI, IoT & SMART TECHNOLOGY
Engineering Expertise // Intelligent Solutions
Digital AI Proof of Concept
The Digital AI-Powered Surveillance Analytics Proof of Concept (POC) was conducted at the Malaysia LNG Sdn. Bhd. (MLNG) Complex, Bintulu, specifically within the Contractor Yard (CY)_1 area, including the designated smoking area, to demonstrate the application of Digital AI technology in supporting Occupational Safety and Health (OSH), safety compliance, and security surveillance.
POC Execution Details
- Location:MLNG Complex, Bintulu (Contractor Yard CY_1 Area & Smoking Area)
- Client:Malaysia LNG Sdn. Bhd. (MLNG)
- Execution Date:13th August 2026
- Technology:MEGVII AI Solution
POC Objectives
To demonstrate and verify the practical applicability of the currently available Digital AI capabilities for supporting OSH, safety compliance, and security surveillance.
Demonstrate AI capability to analyze surveillance camera feeds and identify predefined events in real time.
Verify face detection and face recognition capabilities within the designated surveillance area.
Verify basic PPE compliance detection (missing safety helmets and face masks).
Verify mobile phone usage detection (operating a phone or engaging in a call).
Verify intrusion and access-related detection (movement into restricted areas, unauthorized persons).
Verify perimeter monitoring via AI-based detection across defined perimeter boundaries.
Verify smoking detection capability outside the designated smoking areas.
Demonstrate generation of AI events and alerts following successful detections.
POC Scope & Testing Boundaries
8 Functional Areas Assessed
Capability Demonstration
This POC was conducted to demonstrate AI detection capabilities available within the system configuration at the time of testing. It serves as a practical capability demonstration and does not constitute a full-scale operational deployment or final performance acceptance.
AI Parameter Evaluation Matrix
Demonstrated Parameters
0
/ 8Face Detection Performance
0
%Face Detection
Evaluation of detection & recognition capability.
Safety Helmet
Detection of personnel missing safety helmets.
Face Mask
Detection of personnel missing face masks.
Mobile Phone
Detection of personnel operating a mobile phone.
Intrusion
Movement into defined restricted areas.
Unauthorized Person
Persons without required authorization.
Perimeter
Movement across defined boundaries.
Smoking Detection
Detection within non-designated areas.
Visual Evidence & Captures
Real-time AI detection snapshots from the MEGVII AIoT Engine during the POC at MLNG Complex.
Face Detection & FR
Missing Safety Helmet
Mobile Phone Usage
Missing Face Mask
Intrusion - Personal
Intrusion - Vehicle
Perimeter Boundary
Formal detailed audit logs available in the POC Outcome Report Document.
Observations & Conclusion
Key Observations
- ▸Digital AI system successfully demonstrated all 8 identified AI detection parameters.
- ▸System automatically detected predefined safety/security events and generated corresponding alerts.
- ▸Observed face detection performance was approximately 99%.
- ▸Selected behavioral detection functions demonstrated an observed performance of ~75-80%.
Conclusion
The POC successfully demonstrated the currently available AI detection capabilities identified for the MLNG Complex CY_1 area.
It demonstrates the potential of Digital AI-Powered Surveillance Analytics to provide an additional intelligent monitoring capability in support of existing OSH, safety compliance, and security surveillance practices.
SUCCESSFULLY DEMONSTRATED
Added Value to Client
Transforming existing surveillance infrastructure into a proactive, intelligent, and highly efficient OSH management asset.
Capital Efficiency
The system leverages existing IP CCTV infrastructure (e.g., standard Hikvision cameras) and adds an AI intelligence layer without requiring unnecessary hardware replacement.
Continuous Safety Visibility
Shifts Occupational Safety and Health (OSH) management from subjective, periodic manual audits to 24/7 autonomous monitoring with real-time alert generation.
Data-Driven Accountability
Converts visual events into measurable KPIs, providing centralized tracking of contractor safety records, compliance rates, and incident history.
Resource Optimization
Automates routine surveillance tasks, allowing safety personnel to focus on high-value incident response and critical decision-making rather than continuous screen monitoring.
Network System Design Architecture
Physical Connectivity & Edge-to-Admin Topology
Site Location (Field Capture)
POE SWITCH
Site Location Aggregator
POE SWITCH
Site Admin Central Switch
MEGVII (HEMETV)
MEGVII
Video Recorder
LAPTOP 1
MEGVII Mon.
LAPTOP 2
Hikvision Mon.
Project Limitations
Operating parameters and inherent considerations to acknowledge when scaling Digital AI surveillance platforms.
Environmental Dependencies
AI detection performance is inherently subject to external variables, including camera positioning, line-of-sight occlusions, lighting quality, and adverse weather conditions.
Behavioral Accuracy Ceilings
While facial detection achieves ~99% accuracy, complex behavioral anomalies (such as mobile phone usage) currently yield 75-80% precision and require ongoing optimization.
Validation for Complex Variables
Specific parameters, such as smoking detection in non-designated areas, are highly susceptible to environmental interference and require further field validation during active operations.
Decision Support Nature
The Digital AI is an intelligence and alerting layer; it does not physically prevent incidents and still requires human intervention for enforcement and emergency response.
Future Recommendations
Algorithm Customization
Train and deploy site-specific AI models tailored to emerging operational risks, such as chemical spill detection, specific heavy machinery proximity, or advanced ergonomic strain analysis.
Progressive Scalability
Expand the proven AI ecosystem from the initial pilot zones (e.g., Contractor Yard 1) to encompass critical manufacturing areas, storage terminals, and offloading perimeters.
Enterprise Integration
Synchronize the AI Command Dashboard with existing Human Resources databases, contractor permit systems, and automated access control gates for seamless identity management.
Biometric Telemetry
Enhance predictive safety capabilities by outfitting high-risk personnel with IoT wearables to continuously track Heart Rate Variability (HRV) and physiological data for proactive fatigue intervention.
WHY SUMIRA ?
Local Implementation
Sumira provides the local engineering, project coordination and client interface, ensuring effective communication, support and implementation in Sarawak.
Industry Understanding
Sumira understands the industrial, engineering and OSH environment, allowing the Digital AI solution to be developed around actual operational requirements.
AI Tech Partner
Sumira is supported by MEGVII, an established AI technology company with advanced computer vision and AI capabilities, bringing proven technology into industrial execution.
Practical Integration
The solution is designed to work with existing site practices, systems, workflows and infrastructure, with AI applied where it provides practical operational value.
Scalable Approach
The implementation can begin with targeted applications and progressively expand to additional facilities, use cases and digital capabilities based on MLNG's requirements and priorities.
“Sumira is the local implementation lead, supported by MEGVII's advanced AI technology, combining industrial understanding with AI capability to deliver a practical and scalable Digital AI solution for MLNG.”
INTEGRATED DASHBOARD
FROM DATA to SAFETY INTELLIGENCE
The Integrated Dashboard serves as your centralized safety intelligence platform, transforming raw data from multiple sources into actionable insights. This unified system visualizes all data—providing real-time monitoring, automated alerts, comprehensive analytics, and reporting across personnel, contractors, and vendors.
LIVE MONITORING
- - Camera status
- - Selected live feeds
- - Operational locations
AI EVENTS
- - PPE violation
- - Restricted area
- - Vehicle proximity
- - Intrusion & Smoke
EVENT INFO
Location | Date | Time | Camera | Event Type | Image Capture Evidence
PREVENTIVE ANALYTICS
Statistical summary of performance (daily, weekly, monthly) & anomaly trends.
ONE DASHBOARD. ONE VIEW. ACTIONABLE INFORMATION.
Management does not need to continuously monitor every camera. The system highlights events that require attention.
THE END
DIGITAL AI, IoT & SMART TECHNOLOGY
Engineering Expertise. Intelligent Solutions.
Successfully demonstrating practical AI capabilities for enhanced safety visibility, compliance verification, and operational intelligence.
Sumira Engineering Sdn. Bhd.
POC Verification - MLNG Complex Bintulu