Applied AI for Safer, Smarter Industry
We bridge the gap between cutting-edge AI research and real-world industrial challenges. OCST delivers rigorous scientific solutions for complex operational environments.
What We Do
Our research focuses on critical areas where AI can drive real impact in industrial and infrastructure systems.
Industrial AI for Manufacturing
Enhancing production efficiency through intelligent automation and real-time process control.
- check_circle Computer vision for quality inspection
- check_circle Predictive analytics for production optimization
- check_circle Real-time monitoring and control systems
- check_circle Human-robot collaboration frameworks
Smart Energy & Power Systems
Optimizing energy distribution and consumption through advanced analytics and automation.
- check_circle Smart grid management systems
- check_circle Energy storage optimization
- check_circle Demand response algorithms
- check_circle Distributed energy resource integration
Predictive Maintenance
Detecting anomalies early to prevent costly downtime using advanced time-series analysis.
- check_circle IoT sensor data analysis
- check_circle Machine learning for failure prediction
- check_circle Condition-based maintenance scheduling
- check_circle Digital twin integration
Optimization & Simulation
Modeling complex systems and digital twins for better strategic decision making.
- check_circle Digital twin development
- check_circle Supply chain optimization
- check_circle Resource allocation algorithms
- check_circle Scenario planning and simulation
Solutions
Practical AI solutions designed for real-world industrial applications, delivering measurable value in manufacturing and production environments.
Real-Time Object Tracking
Advanced multi-object tracking system for warehouse automation, logistics, and production line monitoring.
Learn More →
Automated Defect Detection
AI-powered visual inspection system for real-time quality control in manufacturing processes.
Learn More arrow_forward
Industrial OCR Solution
Rotation-aware OCR solution for automated recognition of engraved identifiers in manufacturing environments.
Learn More arrow_forwardReal-Time Object Tracking
Intelligent Vision System for Logistics and Manufacturing
Overview
Modern manufacturing and logistics operations require precise, real-time visibility of materials, products, and assets as they move through facilities. Traditional tracking methods using barcodes or RFID face limitations in coverage, reliability, and cost.
Our Real-Time Object Tracking Solution uses computer vision and AI to provide continuous, contactless tracking of objects throughout industrial environments, enabling smarter automation and operational insights.
The Problem
- Coverage Gaps: Barcode/RFID requires specific scanning points, creating blind spots
- Speed Limitations: Manual scanning creates bottlenecks in high-throughput operations
- Cost: RFID infrastructure is expensive for large-scale deployment
- Reliability: Occlusion, orientation, and environmental factors affect accuracy
- Integration: Difficult to integrate with existing automation systems
Our Approach
Vision-Based Multi-Object Tracking
We leverage cutting-edge computer vision algorithms to track objects continuously without physical tags.
Core Technologies:
- Deep learning-based object detection and classification
- Multi-object tracking (MOT) algorithms for identity preservation
- Re-identification (ReID) models for tracking across camera views
- Predictive modeling for handling occlusions and temporary losses
Tracking Capabilities:
- Simultaneous tracking of hundreds of objects
- Cross-camera tracking across facility zones
- Velocity and trajectory prediction
- Dwell time and path analytics
System Architecture
- Multi-Camera Network: Strategically positioned cameras for full facility coverage
- Object Detection: Real-time identification of tracked items
- Feature Extraction: Unique visual signatures for each object
- Track Association: Maintaining object identities across frames and cameras
- Data Fusion: Integration with other sensor inputs
- Analytics Engine: Real-time insights and anomaly detection
- API Integration: Data delivery to WMS / MES systems
Key Features
- Contactless Tracking
- Real-Time Performance
- Scalable Coverage
- Robustness
- Historical Analytics
Use Cases
Warehouse & Logistics
- Inbound/outbound verification and tracking
- Putaway and picking operation monitoring
- Pallet and container tracking
- Loading dock management
- Inventory reconciliation
Manufacturing
- Work-in-progress (WIP) tracking
- Assembly line monitoring
- Material flow optimization
- Cycle time analysis
- Bottleneck identification
Distribution Centers
- Cross-docking operations
- Sortation verification
- Route optimization
- Throughput analysis
Business Benefits
Operational Efficiency:
- Reduced manual scanning and data entry
- Faster throughput with automated verification
- Minimized search time for misplaced items
Accuracy & Visibility:
- Real-time inventory accuracy
- Elimination of tracking blind spots
- Enhanced traceability and compliance
Data-Driven Optimization:
- Detailed operational analytics
- Process bottleneck identification
- Resource allocation optimization
- Performance benchmarking
Cost Savings:
- Reduced labor costs for tracking activities
- Lower infrastructure costs vs. RFID
- Decreased inventory discrepancies and losses
Technical Specifications
Performance Metrics:
- Tracking Accuracy: >98% object identity preservation
- Detection Rate: >99% for target object classes
- Processing Speed: Real-time at 30+ fps per camera
- Scalability: Support for 50+ cameras per system
System Requirements:
- IP cameras (resolution: 1080p minimum, 4K recommended)
- GPU-accelerated compute servers (edge or centralized)
- Network infrastructure (1 Gbps minimum)
Integration:
- RESTful APIs for WMS/MES integration
- Standard protocols (HTTP, WebSocket, MQTT)
- Database connectivity (SQL, NoSQL)
- Dashboard and visualization tools
Deployment Options
Edge Deployment:
- On-premise processing for low latency and data privacy
- Suitable for facilities with local IT infrastructure
Hybrid Architecture:
- Edge detection with cloud-based analytics
- Balances performance with advanced analytics capabilities
Custom Configurations:
- Tailored to specific facility layouts and requirements
- Integration with existing automation systems
Implementation Roadmap
- Site Assessment: Facility walkthrough and requirement gathering
- Pilot Installation: Limited deployment to validate coverage and accuracy
- Model Customization: Training on specific object types and facility conditions
- Full Deployment: System-wide rollout with comprehensive coverage
- Integration: Connection to WMS, MES, or analytics platforms
- Optimization: Continuous improvement based on operational feedback
Why This Solution?
- Proven Technology: Based on state-of-the-art computer vision research
Automated Defect Detection
AI-Powered Visual Inspection for Quality Control
Overview
Manufacturing quality control traditionally relies on manual visual inspection, which is time-consuming, inconsistent, and prone to human error. As production speeds increase and quality standards become more stringent, automated inspection systems have become essential.
Our Automated Defect Detection Solution leverages deep learning to provide real-time, reliable, and scalable quality control for diverse manufacturing environments.
The Challenge
- Inconsistency: Human inspectors have varying attention levels and judgment criteria
- Speed Bottlenecks: Manual inspection cannot keep pace with modern production lines
- Cost: Skilled inspectors are expensive and difficult to retain
- Coverage: 100% inspection is often infeasible manually
- Subtle Defects: Microscopic or low-contrast defects are easily missed
Our Approach
Deep Learning-Based Visual Inspection
We employ state-of-the-art computer vision models specifically adapted for industrial defect detection.
Technical Framework:
- Custom-trained convolutional neural networks (CNNs) for defect classification
- Multi-scale feature extraction for detecting defects of varying sizes
- Real-time inference optimized for production line speeds
- Adaptive threshold tuning to balance false positives and false negatives
Key Capabilities:
- Detection of surface defects (scratches, dents, cracks, discoloration)
- Dimensional anomaly detection
- Texture and pattern irregularities
- Assembly and alignment verification
System Architecture
End-to-End Pipeline:
- Image Acquisition: High-resolution industrial cameras with controlled lighting
- Preprocessing: Image enhancement and normalization
- Defect Detection: Multi-stage deep learning models
- Classification: Categorization of defect types and severity
- Decision Logic: Automated pass/fail determination
- Data Logging: Comprehensive defect tracking and reporting
The system integrates seamlessly with existing production lines and quality management systems.
Key Strengths
- High Accuracy: Detection rates exceeding 99% with low false positive rates
- Real-Time Performance: Inspection speeds matching or exceeding production line throughput
- Adaptability: Easily retrained for new defect types or product variations
- Scalability: Deployable across multiple production lines and facilities
- Traceability: Complete documentation of inspection results with image archiving
Industry Applications
- Metal Manufacturing: Surface defect detection on steel and aluminum products
- Electronics: PCB inspection and component placement verification
- Automotive: Paint quality and body panel inspection
- Textiles: Fabric defect detection and pattern matching
- Packaging: Label verification and seal integrity checking
Return on Investment
- 60–80% reduction in manual inspection labor costs
- 95%+ defect detection rate improvement over manual inspection
- Reduced rework costs through early defect identification
- Enhanced brand reputation from improved product quality
- Data-driven insights for process improvement
Implementation Process
- Assessment: Evaluation of current inspection processes and requirements
- Pilot Deployment: Limited deployment to validate performance
- Model Training: Custom model development using client production data
- Integration: Full system deployment and integration with existing infrastructure
- Optimization: Continuous tuning and improvement based on production feedback
Technical Requirements
Hardware:
- Industrial-grade cameras (resolution based on defect size requirements)
- Compute platform (edge device or server-based)
- Appropriate lighting systems for consistent image acquisition
Software:
- Compatible with standard industrial protocols (OPC UA, Modbus, etc.)
- Integration APIs for MES, ERP, and quality management systems
Why Choose This Solution?
- Production-Proven: Validated in high-volume manufacturing environments
- Customizable: Adapted to specific product types and defect characteristics
- Supportable: Comprehensive training and ongoing technical support
- Future-Ready: Continuous model improvements and updates
This solution is ideal for manufacturers seeking to modernize quality control, reduce costs, and maintain competitive advantage through superior product quality.
Get Started
Ready to transform your quality control process?
👉 Contact us for a consultation and proof-of-concept demonstration.
Industrial OCR Solution
Rotation-Aware OCR for Manufacturing Traceability
Overview
In modern manufacturing environments, reliable identification of metallic components is essential for traceability, quality control, and automation. However, engraved or stamped identifiers on metal surfaces are often difficult to recognize due to low contrast, reflections, surface degradation, and arbitrary orientations.
Our Industrial OCR Solution is designed to address these challenges with a robust, deployment-ready vision pipeline, validated in real factory environments.
Key Challenges in Industrial OCR
- Low-contrast engraved characters on metallic surfaces
- Strong reflections and uncontrolled lighting
- Arbitrary object orientation on production lines
- Strict formatting requirements for serial numbers
Conventional OCR systems, originally designed for documents or natural scenes, often fail under these conditions.
Our Approach
Rotation-Aware Industrial OCR Pipeline
- Orientation-aware localization of engraved text regions
- Geometry-normalized text extraction
- Robust recognition under rotational ambiguity
- Confidence-based decision logic for reliable operation
This approach ensures stable performance without relying on heavy preprocessing or language models.
System Architecture
- Industrial image acquisition
- Rotation-aware detection of objects and text regions
- Orientation normalization
- Dual-direction text recognition
- Confidence-based result selection
The architecture is optimized for real-time or near real-time deployment in manufacturing systems.
Key Strengths
- Rotation Robustness: Stable recognition at arbitrary angles
- Industrial Reliability: Designed for harsh factory conditions
- Deployment-Oriented Design: Lightweight edge/server inference
- Confidence-Aware Output: Safe automation integration
Validated Use Cases
- Automated traceability of metallic components
- Serial number recognition on steel bars and cylindrical products
- Quality control in smart factory environments
The solution has been validated using real industrial data, demonstrating high accuracy and stable runtime behavior.
Integration & Deployment
- Manufacturing Execution Systems (MES)
- Quality Control (QC) systems
- Factory automation platforms
Flexible deployment options are available depending on production requirements.
Why Choose This Solution?
- Proven performance in real manufacturing environments
- Designed by engineers with industrial deployment experience
- Focused on reliability, not experimental complexity
Interested in a demo or proof-of-concept (PoC)?
👉 Contact us to discuss how this solution can be adapted to your production environment.
Spotlight Projects
Explore our latest breakthroughs and how they are reshaping industrial landscapes.
CASE STUDY
Vision-Based Steel Coil Inspection
Deployed deep learning–based computer vision on hot and cold rolling lines to detect surface defects on steel coils in real time, integrated with existing PLC control systems.
TECHNOLOGIES
Impact
40% reduction in manual inspection effort and significant scrap reduction in continuous production lines.
PILOT PROJECT
Adaptive Production Line Optimization
Built a real-time analytics and optimization layer for a multi-stage manufacturing line, continuously tuning process parameters based on live quality and throughput metrics.
TECHNOLOGIES
Impact
Up to 12% throughput increase and 8% reduction in off-spec product across the pilot line.
RESEARCH PAPER
Self-Healing Smart Grid Control
Developed reinforcement learning agents that dynamically reroute power flows in distribution networks to isolate faults and maintain service during local failures.
TECHNOLOGIES
Impact
Simulated 45% reduction in outage duration and improved resilience to cascading failures.
INNOVATION
Substation Condition Monitoring Platform
Aggregated IoT sensor streams from transformers, breakers, and busbars to perform anomaly detection and health scoring for high-voltage substations.
TECHNOLOGIES
Impact
Reduced unplanned interruptions by 20% across monitored substations.
CASE STUDY
Cross-Plant Predictive Maintenance Hub
Implemented a unified predictive maintenance platform aggregating vibration, temperature, and process data from multiple plants.
TECHNOLOGIES
Impact
30% reduction in critical equipment failures and improved model reuse across plants.
OPTIMIZATION
End-to-End Supply Chain Digital Twin
Built a digital twin of a multi-echelon supply chain to simulate demand surges, capacity constraints, and transport disruptions.
TECHNOLOGIES
Impact
Up to 18% reduction in inventory carrying costs while maintaining service levels.
Publications
Our research contributions to the scientific community, advancing the field of industrial AI.
Automated Detection and Recognition of Engraved Serial Numbers on Metallic Surfaces
Authors: Thanh-Dat Nguyen; Sachin Ranjan; Le-Anh Tran; Kichul Lee; Moonseok Kang; Hoon Kim
Venue: Unpublished
Automated reading of engraved or printed serial numbers on metallic components remains a challenging problem in industrial manufacturing due to low contrast, reflective surfaces, arbitrary orientations, and surface degradation. Conventional optical character recognition (OCR) systems often fail in such environments, particularly when text is embedded within geometrically structured objects and appears under uncontrolled rotations. This paper presents a robust end-to-end vision pipeline for accurately recognizing numeric series on industrial metallic surfaces. The proposed approach f irst employs an oriented object detection model to localize target objects and their associated text regions using oriented bounding boxes. The detected text areas are then rotation-normalized through geometry-aware cropping and square padding, enabling consistent downstream recognition. A YOLO-based character recognition model is subsequently applied, with dual-orientation inference used to resolve rotational ambiguity. To further improve reliability, a confidence-driven post-processing strategy suppresses duplicate detections, enforces valid text-length constraints, and selects the most consistent recognition results. Experiments on real-world industrial data demonstrate that the proposed system achieves high end-to-end recognition accuracy and robustness under challenging imaging conditions, outperforming baseline configurations without orientation handling or post-processing. The proposed pipeline offers a practical and deployable solution for automated traceability and quality control in manufacturing environments.
Read full paper arrow_forwardPOCS-based Image Compression: An Empirical Examination
Authors: Truong-Dong Do; Le-Anh Tran; Thanh-Dat Nguyen; Nghe-Nhan Truong; Dong-Chul Park; My-Ha Le
Venue: Proceedings of 2024 7th International Conference on Green Technology and Sustainable Development (GTSD)
This paper investigates the applicability of the Pro-jection onto Convex Set (POCS)-based clustering algorithm to image compression tasks. The POCS-based clustering approach treats all data points in a given dataset as non-intersecting convex sets and performs POCS-based parallel projections from each cluster prototype onto corresponding member data points to minimize an objective function and update cluster prototypes. The POCS-based clustering algorithm has been proven to be able to yield promising results against other prevailing clustering approaches in terms of convergence time and clustering error on general clustering tasks. In this study, a comparison of various clustering schemes for image compression applications has been conducted. The evaluations and analyses on various standard test images verify that the POCS-based clustering algorithm can perform competitively against other conventional clustering methods in image compression problems.
Read full paper arrow_forwardEmbedding Clustering via Autoencoder and Projection onto Convex Set
Authors: Le-Anh Tran; Thanh-Dat Nguyen; Truong-Dong Do; Chung Nguyen Tran; Daehyun Kwon; Dong-Chul Park
Venue: 2023 International Conference on System Science and Engineering (ICSSE)
Projection onto Convex Set (POCS) is a powerful signal processing tool for various convex optimization problems. For non-intersecting convex sets, the simultaneous POCS method can result in a minimum mean square error solution. This property of POCS has been applied to clustering analysis and the POCS-based clustering algorithm was proposed earlier. In the POCS-based clustering algorithm, each data point is treated as a convex set, and a parallel projection operation from every cluster prototype to its corresponding data members is carried out in order to minimize the objective function and to update the memberships and prototypes. The algorithm works competitively against conventional clustering methods in terms of execution speed and clustering error on general clustering tasks. In this paper, the performance of the POCS-based clustering algorithm on a more complex task, embedding clustering, is investigated in order to further demonstrate its potential in benefiting other high-level tasks. To this end, an off-the-shelf FaceNet model and an autoencoder network are adopted to synthesize two sets of feature embeddings from the Five Celebrity Faces and MNIST datasets, respectively, for experiments and analyses. The empirical evaluations show that the POCS-based clustering algorithm can yield favorable results when compared with other prevailing clustering schemes such as the K-Means and Fuzzy C-Means algorithms in embedding clustering problems.
Read full paper arrow_forwardEfficient Infrared and Thermal Imaging Fusion Approach for Real-time Human Detection in Heavy Smoke Scenarios
Authors: Authors: Nghe-Nhan Truong; My-Ha Le; Truong-Dong Do; Le-Anh Tran; Thanh-Dat Nguyen; Hoang-Hon Trinh
Venue: 2023 International Conference on System Science and Engineering (ICSSE)
Fire is considered one of the most serious threats to human lives which results in a high probability of fatalities. Those severe consequences stem from the heavy smoke emitted from a fire that mostly restricts the visibility of escaping victims and rescuing squad. In such hazardous circumstances, the use of a vision-based human detection system is able to improve the ability to save more lives. To this end, a thermal and infrared imaging fusion strategy based on multiple cameras for human detection in low-visibility scenarios caused by smoke is proposed in this paper. By processing with multiple cameras, vital information can be gathered to generate more useful features for human detection. Firstly, the cameras are calibrated using a Light Heating Chessboard. Afterward, the features extracted from the input images are merged prior to being passed through a lightweight deep neural network to perform the human detection task. The experiments conducted on an NVIDIA Jetson Nano computer demonstrated that the proposed method can process with reasonable speed and can achieve favorable performance with a mAP@0.5 of 95%.
Read full paper arrow_forwardPOCS-based Clustering Algorithm
Authors: Authors: Le-Anh Tran; Henock M. Deberneh; Truong-Dong Do; Thanh-Dat Nguyen; My-Ha Le; Dong-Chul Park
Venue: 2022 International Workshop on Intelligent Systems (IWIS)
A novel clustering technique based on the projection onto convex set (POCS) method, called POCS-based clustering algorithm, is proposed in this paper. The proposed POCS-based clustering algorithm exploits a parallel projection method of POCS to find appropriate cluster prototypes in the feature space. The algorithm considers each data point as a convex set and projects the cluster prototypes parallelly to the member data points. The projections are convexly combined to minimize the objective function for data clustering purpose. The performance of the proposed POCS-based clustering algorithm is verified through experiments on various synthetic datasets. The experimental results show that the proposed POCS-based clustering algorithm is competitive and efficient in terms of clustering error and execution speed when compared with other conventional clustering methods including Fuzzy C-Means (FCM) and K-Means clustering algorithms.
Read full paper arrow_forwardAbout OCST AI Research
We bridge the gap between cutting-edge AI research and real-world industrial challenges. OCST delivers rigorous scientific solutions for complex operational environments, combining academic rigor with practical deployment capabilities.
Our Mission
To advance the state of industrial AI through rigorous research, practical applications, and meaningful partnerships that drive real-world impact.
Why OCST for AI
We combine academic rigor with practical deployment capabilities, setting us apart from traditional labs.
Deep Industrial Expertise
Decades of combined domain knowledge in core heavy industries, ensuring our solutions address real-world challenges.
Real-World Deployments
Proven AI integration in live industrial environments, from prototype to production.
Safety & Reliability
Focus on safety-critical systems and explainable AI for industrial trust.
Long-Term R&D
Sustainable, long-term research beyond short-term cycles, building foundational technologies.