Corner Sentinel is an advanced system designed to enhance road safety at sharp, blind corners, particularly in hilly regions. It combines Image Processing, Machine Learning, and Li-Fi Communication technologies to detect potential hazards, notify drivers in real time, and help prevent collisions. By leveraging these technologies, Corner Sentinel addresses the challenges posed by poor visibility, speeding, and wrong-side driving in dangerous areas.
This repository contains the complete source code, project report, and relevant technical documentation.
- Project Overview
- Features
- System Architecture
- Technologies Used
- Installation
- Usage
- Results
- Future Scope
- References
- Real-Time Vehicle Detection: Leverages the YOLOv8 model to detect and classify vehicles in real time.
- Wrong-Lane Detection: Identifies vehicles driving in the wrong lane and triggers alerts.
- Speed Estimation: Uses optical flow techniques to estimate the speed of detected vehicles.
- Accident/Crash Detection: Predicts and detects accidents based on sudden stops and abnormal vehicle movements.
- Li-Fi Communication: Sends real-time hazard warnings via modulated light signals to approaching vehicles.
- Scalable Architecture: The system's modular design makes it easy to upgrade and scale for future improvements.
The system is composed of two primary modules:
- Machine Learning Module: This module runs on a Raspberry Pi 5 and performs real-time vehicle detection, classification, and tracking using the YOLOv8 model. It analyzes traffic feeds from cameras positioned at blind corners.
- Li-Fi Communication Module: The module generates binary codes corresponding to detected hazards (e.g., wrong-lane driving, stationary vehicles, speeding) and transmits this data to approaching vehicles via Li-Fi.
The working of the system is as follows:
- Video Acquisition and Preprocessing: Cameras capture live traffic feeds, and frames are resized and normalized.
- Vehicle Detection and Classification: YOLOv8 model identifies vehicles and outputs bounding boxes and class labels.
- Vehicle Tracking: The ByteTrack algorithm assigns unique IDs and tracks vehicles across frames.
- Hazard Assessment: Based on factors like speed, lane usage, and the presence of emergency or stationary vehicles, Corner Sentinel assesses potential risks.
- Li-Fi Transmission: Binary code representing the detected hazards is transmitted using LEDs via Manchester Encoding.
- Warning Reception: Vehicles equipped with receivers demodulate signals and warn drivers of potential dangers.
The system uses image processing to acquire and analyze video streams in real time. Techniques such as Object Detection (via YOLOv8) and Lane Detection are employed to identify vehicles and lane boundaries. Optical Flow techniques are used for speed estimation.
A custom-trained YOLOv8 Nano model is employed for vehicle detection. The model was trained on a combination of Google’s Open Images Dataset and an Indian-Vehicles dataset sourced from Roboflow. The system tracks vehicles, detects their speed, and classifies them by type (e.g., cars, trucks, ambulances).
Li-Fi, short for Light Fidelity, is used for real-time communication between the system and approaching vehicles. The system encodes binary messages based on hazard assessments and transmits them using On-Off Keying (OOK) modulation and Manchester Encoding via high-power LEDs. Receivers on vehicles decode these signals and alert drivers about potential risks.
- Raspberry Pi 5: The core device that runs YOLOv8 and manages Li-Fi transmission.
- Camera: Captures live traffic feeds for real-time analysis.
- LEDs: Used for transmitting binary-coded warnings to vehicles.
- Solar Panels: Act as receivers for decoding the Li-Fi signals and extracting binary codes.
- Raspberry Pi 5 with Raspbian OS installed.
- Python 3.9+ with libraries:
opencv-python,ultralytics,numpy. - YOLOv8 weights.
-
Clone this repository to your Raspberry Pi:
git clone https://github.com/yourusername/corner-sentinel.git cd corner-sentinel -
Install the necessary Python packages:
pip install -r requirements.txt
-
Run the Program.
To start the system, run the following command:
python3 final.pyThe program will begin analyzing traffic feeds, detecting vehicles, and sending real-time alerts using Li-Fi transmission.
Data sent via Li-Fi is encoded in an 8-bit format:
- Bit 7: Stationary Vehicle Detection
- Bit 6: Accident Detection
- Bits 5-3: Vehicle Type (Ambulance, Car, Bike, Truck)
- Bit 2: Wrong-Lane Detection
- Bits 1-0: Speed Category (below 40 km/h, 40-60 km/h, above 60 km/h)
-
Vehicle Detection Accuracy: The YOLOv8 model achieved a high detection accuracy rate during testing.
-
Speed Estimation: Accurate speed estimation was achieved using the optical flow method.
-
Li-Fi Transmission: The system successfully transmitted hazard alerts using Li-Fi, which were received and decoded by equipped vehicles.
-
Enhanced Weather Adaptability: Improving the system to handle adverse weather conditions like rain and fog.
-
Range Extension: Research into extending the communication range of Li-Fi for broader coverage.
-
Optimized Machine Learning Models: Utilizing more efficient versions of YOLO and exploring alternative algorithms to enhance performance.
-
Li-Fi Circuit Enhancements: Introducing noise reduction filters and optimized encoding schemes to improve the robustness of data transmission.
The project reworked code from: Sending Text Messages with Visible Light Communication
For a complete list of references, please refer to the Project Report and the Project Presentation.