A New Vehicle Localization Scheme based on Combined Optical Camera Communication and Photogrammetry

Table of Contents

• 1. Introduction
• 2. Proposed System Architecture
• 3. Methodology
• 4. Experimental Results
• 5. Original Analysis
• 6. Technical Details and Mathematical Formulation
• 7. Case Study: Highway Platooning Scenario
• 8. Future Applications and Directions
• 9. References

1. Introduction

Localization is the process of identifying the position (x, y in 2D space; x, y, z in 3D space) of an object at a specific time. With the rise of Internet of Things (IoT) and autonomous vehicles, accurate localization has become critical. Traditional GPS provides line-of-sight solutions but suffers from accuracy issues in urban canyons and tunnels. This paper proposes a novel scheme combining Optical Camera Communication (OCC) and photogrammetry to achieve high-precision vehicle localization without modifying existing transportation infrastructure.

2. Proposed System Architecture

The system classifies vehicles into two categories: Host Vehicles (HVs) which estimate positions of other vehicles, and Forwarding Vehicles (FVs) which move in front of HVs. FVs transmit modulated data from their tail lights, which is received by the HV's camera using OCC. Additionally, streetlight (SL) data is used to improve HV position accuracy.

2.1 Key Components

• Optical Camera Communication (OCC): Uses modulated light from FV tail lights and SLs to transmit data.
• Photogrammetry: Measures distance by calculating the occupied image area on the image sensor.
• Data Fusion: Combines OCC and photogrammetry data for robust localization.

3. Methodology

The HV determines its own position using SL data, then calculates the relative position of the FV by comparing distance changes between HV-SL and HV-FV. The distance between FV or SL and the HV camera is computed using photogrammetry: $d = \frac{f \times H}{h}$, where $f$ is focal length, $H$ is actual height, and $h$ is image height.

3.1 Distance Calculation

Using the pinhole camera model, the distance $d$ from the camera to an object is given by:

$d = \frac{f \times W}{w}$

where $W$ is the actual width of the object and $w$ is the width in pixels on the image sensor.

3.2 Position Estimation

The HV position is first estimated using triangulation from multiple SLs. Then, the relative position of the FV is determined by:

$\Delta P_{FV} = P_{HV} + \Delta d \cdot \cos(\theta)$

where $\Delta d$ is the change in distance and $\theta$ is the angle of arrival.

4. Experimental Results

The experimental setup used a camera with 640x480 resolution, focal length 3.6 mm, and a tail light of 0.15 m diameter. Results showed distance measurement error less than 5% for distances up to 30 meters. The proposed scheme achieved positioning accuracy within 0.5 meters, significantly outperforming GPS-only solutions which typically have 2-5 meter errors.

5. Original Analysis

Core Insight: This paper presents a clever fusion of two mature technologies—OCC and photogrammetry—to solve a critical problem in autonomous driving: reliable vehicle localization without expensive infrastructure upgrades. The key innovation is using existing tail lights and streetlights as communication beacons, turning passive infrastructure into active positioning aids.

Logical Flow: The authors logically progress from problem identification (GPS limitations) to solution design (OCC+photogrammetry), then to mathematical modeling and experimental validation. The flow is coherent, though the paper could benefit from more rigorous comparison with state-of-the-art methods like LiDAR-based SLAM or V2X communication.

Strengths & Flaws: The main strength is the low-cost, infrastructure-light approach. However, the scheme assumes clear line-of-sight and good lighting conditions, which may not hold in fog, rain, or at night. Additionally, the reliance on tail light modulation could be affected by dirty or damaged lights. Compared to LiDAR-based systems (which cost thousands of dollars), this camera-based approach is far cheaper but less accurate in adverse conditions. As noted by Geiger et al. (2012) in the KITTI dataset, camera-based methods often degrade in low-light scenarios.

Actionable Insights: For practitioners, this scheme is best suited for highway platooning and parking assistance where lighting conditions are controlled. Future work should explore hybrid approaches combining OCC with radar or ultrasonic sensors for all-weather operation. The paper's photogrammetry model could be enhanced using deep learning-based depth estimation, as demonstrated by Eigen et al. (2014) in their work on single-image depth prediction.

6. Technical Details and Mathematical Formulation

The photogrammetry model uses the pinhole camera equation:

$\frac{x}{X} = \frac{f}{Z}$

where $x$ is image coordinate, $X$ is world coordinate, $f$ is focal length, and $Z$ is depth. For a known object size $S$ and image size $s$, distance $D$ is:

$D = \frac{f \times S}{s}$

The OCC modulation uses On-Off Keying (OOK) at frequencies above 100 Hz to avoid visible flicker. The received signal strength (RSS) is used for distance estimation as a secondary method:

$P_r = P_t \times \frac{A_r}{\pi D^2} \times \cos(\phi)$

where $P_r$ is received power, $P_t$ is transmitted power, $A_r$ is receiver area, and $\phi$ is incidence angle.

7. Case Study: Highway Platooning Scenario

Scenario: A platoon of three vehicles traveling at 80 km/h on a highway. The lead vehicle (FV) transmits its speed and braking status via modulated tail lights. The middle vehicle (HV) uses OCC to receive this data and photogrammetry to measure distance.

Implementation Steps:

1. FV tail light modulates data at 200 Hz (OOK).
2. HV camera captures frames at 30 fps, demodulates the signal.
3. Photogrammetry calculates distance: $D = \frac{3.6mm \times 0.15m}{h_{pixels} \times 0.006mm/pixel}$.
4. HV adjusts speed to maintain safe distance (2-second rule: ~44m at 80 km/h).
5. If FV brakes, HV receives the signal within 33 ms (one frame) and reacts.

Outcome: The system maintains platoon formation with 0.5m accuracy, reducing air drag by up to 15% and improving fuel efficiency.

8. Future Applications and Directions

The proposed scheme has several promising future applications:

• Autonomous Parking: Using OCC from parking lot lights for precise positioning.
• Intersection Management: Vehicles communicate with traffic lights to optimize flow.
• Fleet Management: Real-time tracking of delivery vehicles in urban areas.
• V2X Integration: Combining OCC with DSRC or 5G for redundant localization.
• Smart City Infrastructure: Streetlights become multi-functional communication nodes.

Future research should focus on deep learning-based object detection to improve robustness, and integration with inertial sensors for seamless operation during OCC outages.

9. References

1. M. T. Hossan et al., "A New Vehicle Localization Scheme based on Combined Optical Camera Communication and Photogrammetry," IEEE Access, 2021.
2. A. Geiger, P. Lenz, and R. Urtasun, "Are we ready for autonomous driving? The KITTI vision benchmark suite," CVPR, 2012.
3. D. Eigen, C. Puhrsch, and R. Fergus, "Depth map prediction from a single image using a multi-scale deep network," NeurIPS, 2014.
4. World Health Organization, "Global status report on road safety 2018," WHO, 2018.
5. J. Y. Kim et al., "Optical camera communication for vehicular applications: A survey," IEEE Communications Surveys & Tutorials, 2020.