Client: Department of Transportation (Dishub) of East Java Province — Regional Government Services: Computer Vision Year: 2024 Technology: Next.js, AI, Computer Vision, Data Analytics
Background
East Java Province has one of the busiest road networks in Indonesia, connecting major cities such as Surabaya, Malang, and Sidoarjo with industrial zones and strategic ports. The Provincial Department of Transportation is responsible for monitoring traffic flow across this network for transportation planning, traffic management, and infrastructure development decisions. With continuously increasing vehicle volumes, the need for accurate and efficient monitoring systems has become increasingly urgent.
Challenge
Before AIGLE, traffic monitoring in East Java relied on manual methods with significant limitations:
- Inefficient manual surveys: Officers were stationed at monitoring points to count vehicles visually — a slow, error-prone process that could not operate 24 hours a day
- No real-time data: Traffic data was only available after surveys were completed and manually processed, preventing rapid response to congestion events
- Limited coverage: Reliance on human resources severely constrained the number of locations that could be monitored simultaneously
- No vehicle composition analysis: Manual counting could not consistently distinguish and classify vehicle types (motorcycles, cars, buses, trucks)
- Underutilized CCTV footage: CCTV cameras were already installed at many locations, but their recordings were used only for passive surveillance without any traffic data extraction
Solution
GTM developed AIGLE, an AI-powered traffic detection and vehicle counting system that transforms raw CCTV footage into actionable traffic intelligence.
Real-Time Vehicle Detection
Deep learning models automatically detect and classify vehicles from live CCTV feeds. The system distinguishes between motorcycles, cars, buses, and trucks — providing granular traffic composition data at each monitored intersection.
Continuous Automated Counting
Computer vision algorithms count vehicles passing through designated monitoring zones continuously, 24 hours a day, 7 days a week. Automated counting eliminates the need for manual surveys and delivers consistent, accurate volume data around the clock.
Interactive Dashboard
A web-based dashboard presents traffic volume trends, peak hour analysis, and historical comparisons through interactive charts and maps. Operators can filter data by location, vehicle type, and time range to quickly identify congestion patterns.
Multi-Camera Management
The platform supports simultaneous feeds from multiple CCTV cameras deployed across the road network. Administrators can add, configure, and monitor camera streams from a centralized interface.
Development Process
AIGLE was developed systematically:
- CCTV infrastructure analysis: Auditing the existing camera network to determine compatibility and priority monitoring points
- AI model development: Training object detection models with Indonesian vehicle datasets covering various lighting and weather conditions
- System integration: Connecting live CCTV feeds to the AI pipeline and presenting results on a real-time web dashboard
- Calibration and validation: Testing detection accuracy under various field conditions and fine-tuning the model
- Deployment and training: System launch and operator training for Dishub personnel
Results
- 24/7 real-time monitoring across East Java's CCTV-equipped road network
- 99.5% detection accuracy in identifying and classifying vehicles from CCTV feeds
- 4 vehicle types classified automatically: motorcycles, cars, buses, and trucks
- Manual surveys eliminated — vehicle counting that previously required field officers is now fully automated
- Historical data enabling long-term traffic trend analysis for infrastructure planning
- Report exports in standard formats for policy documents and inter-agency coordination
Technology Stack
- Computer Vision — Deep learning models for real-time vehicle detection and classification
- Next.js — Frontend framework for the interactive dashboard
- AI — Inference pipeline for continuous CCTV video processing
- Data Analytics — Historical traffic data aggregation and visualization