An automotive engineering program required a vision-based solution capable of monitoring driver behaviour, analysing road environments through external cameras, and identifying vehicles through automatic number plate recognition. The objective was to develop and deploy real-time computer vision applications that could operate reliably on edge hardware while supporting multiple automotive safety and monitoring functions.

The project spanned 12 months and was delivered by a dedicated ALTEN team of six engineers in India. The engagement focused on developing Driver Monitoring System (DMS) capabilities, external camera analytics, and an Automatic Number Plate Recognition (ANPR) solution, supported by optimized CNN models, image processing techniques, and deployment on NVIDIA edge hardware and Android platforms.

The scope included:

• In-cabin Driver Monitoring System (DMS)
• Head pose, yawn, drowsiness, and blink-rate detection
• Obstacle detection and safe passage area indication
• Traffic sign and traffic signal detection and recognition
• Automatic Number Plate Recognition (ANPR)
• CNN-based model development and optimization
• Edge deployment on NVIDIA hardware and Android devices
• Real-time video analytics pipeline development
• Image-processing optimization
• License plate recognition across multiple formats and operating conditions

Challenge:

Reliable Driver Monitoring

• Detecting head pose, yawning, blinking, and drowsiness indicators from live in-cabin video streams while maintaining consistent prediction accuracy
• Ensuring reliable monitoring across varying driver positions and viewing angles

Real-Time Road Environment Analysis

• Processing external camera feeds to identify obstacles, traffic signs, traffic signals, and safe passage areas in real time
• Maintaining consistent perception performance across dynamic road conditions

Edge Deployment Performance

• Deploying multiple computer vision models on NVIDIA edge hardware and Android devices with limited computing resources
• Achieving real-time inference without compromising detection accuracy

Number Plate Recognition Variability

• Supporting recognition from both stationary and moving vehicles
• Managing variations in plate orientation, fonts, styles, and environmental conditions, including twilight scenarios

Integrated Video Analytics

• Combining multiple vision models within a single processing pipeline while maintaining operational efficiency
• Delivering high-performance inference suitable for real-time deployment requirements

ALTEN Solutions

In-Cabin Driver Monitoring System

• Developed a video analytics solution incorporating head pose, yawn, drowsiness, and blink-rate detection for driver attention and fatigue monitoring
• Optimized image-processing modules within the DMS pipeline to improve prediction reliability

External Camera Analytics

• Developed computer vision capabilities for obstacle detection and safe passage area indication
• Implemented traffic sign and traffic signal detection and recognition using specialized CNN models
• Enabled continuous analysis of external camera feeds to support ADAS functionality

Edge-Optimized AI Deployment

• Optimized AI/ML models for deployment on NVIDIA edge hardware and Android devices
• Developed lightweight inference pipelines capable of supporting real-time video analytics across deployment platforms

Automatic Number Plate Recognition System

• Developed an ANPR solution capable of detecting, recognizing, and identifying vehicle number plates from live video streams
• Supported wide-angle recognition across horizontal and vertical plate orientations
• Enabled recognition of both stationary and moving vehicles under varying lighting conditions, including twilight environments
• Accommodated multiple license plate fonts and styles

High-Performance Vision Processing

• Implemented an ensemble of models to improve detection and recognition performance
• Integrated image-processing techniques and Lucas-Kanade Optical Flow-based motion detection to enhance inference accuracy and tracking performance
• Achieved processing speeds above 30 FPS with up to 95% accuracy on 1080p image inputs

Dedicated Delivery Team

• Maintained a focused team of six engineers throughout a 12-month engagement
• Delivered development, optimization, testing, and deployment activities from India under a software development model

Business Benefits

Real-Time Driver Monitoring

• Automated detection of fatigue and attention-related indicators enabled continuous monitoring of driver behavior from in-cabin camera feeds

Improved Road Scene Understanding

• External camera analytics provided reliable identification of obstacles, traffic signs, traffic signals, and safe navigation areas

Faster Edge-Based Processing

• Optimized AI/ML models enabled real-time inference on NVIDIA edge hardware and Android devices without reliance on cloud processing

Improved Recognition Accuracy

• Ensemble models and image-processing enhancements improved prediction reliability across DMS, external camera analytics, and ANPR applications

Scalable Computer Vision Deployment

• A unified video analytics pipeline enabled multiple vision applications to operate within the same deployment framework while supporting future expansion

Tools, Technologies & Expertise

Engineering & Development Tools

• Embedded C++
• Python
• TensorFlow
• Ubuntu
• GStreamer
• NVIDIA Edge Hardware
• Android Platform

Engineering Capabilities

• Computer Vision Development
• Driver Monitoring System Engineering
• ADAS Software Development
• CNN Model Development and Optimization
• Edge AI Deployment
• Video Analytics Pipeline Development
• Automatic Number Plate Recognition
• Traffic Sign and Signal Recognition
• Lucas-Kanade Optical Flow Integration

Domain Expertise

• Automotive ADAS Systems
• Driver Safety and Monitoring Applications
• Embedded Computer Vision Solutions
• Real-Time Video Analytics
• Vehicle Surveillance Systems
• Intelligent Transportation Technologies

Where ALTEN Adds Value

• Developing integrated computer vision solutions spanning driver monitoring, external camera analytics, and ANPR
• Optimizing AI/ML models for real-time deployment on Android and NVIDIA edge platforms
• Combining CNN-based perception, image processing, and optical flow techniques to improve performance and accuracy
• Delivering end-to-end development and deployment expertise for automotive vision applications

Why ALTEN
ALTEN combines computer vision, embedded software, and AI/ML expertise to deliver real-time vision solutions for automotive applications. By integrating Driver Monitoring Systems, external camera analytics, and Automatic Number Plate Recognition within a single edge-deployable framework, ALTEN enables reliable perception capabilities that support safety, monitoring, and operational efficiency.