研究目的
To develop a precise tracking method for image sensor communication in intelligent transport systems by modeling vehicle vibration and estimating parameters for inherent vehicle vibration and road surface irregularity.
研究成果
The simplified vehicle vibration model, using only Gaussian random processes for road surface irregularity, effectively reproduces transmitter displacement characteristics in image sensor communication for high frame rates (e.g., 1000 fps) on paved roads. This reduces the dependency on vehicle type and suggests that tracking systems can limit search areas to a few pixels without compensation, with road surface condition being a key factor. For unpaved roads, accuracy can be improved by including some inherent vibration components.
研究不足
The study is limited to I2V-ISC scenarios; vibration in the Z-direction was not measured precisely due to acceleration/deceleration effects. The model's accuracy decreases for lower frame rates (e.g., below 500 fps) and may require inclusion of inherent vibration components for unpaved roads. Generalization to other vehicle types and higher speeds beyond 60 km/h was not fully explored.
1：Experimental Design and Method Selection:
The study involved measuring actual vehicle vibration using a six-axis acceleration sensor in a smartphone during driving scenarios (paved and unpaved roads) to derive frequency characteristics and parameters. A simplified vehicle vibration model using only Gaussian random processes for road surface irregularity was proposed and validated through numerical analysis and experimental measurements of transmitter displacement in an image plane.
2：Sample Selection and Data Sources:
Vehicle vibration data were collected from a vehicle driven at constant speeds (40 km/h, 50 km/h, 60 km/h) on paved and unpaved roads in Nagoya, Japan, using a smartphone's acceleration sensor. Images of light sources (e.g., traffic lights, headlights) were captured with a high-speed camera for displacement analysis.
3：List of Experimental Equipment and Materials:
A six-axis acceleration sensor in an iPhone 6 (InvenSense MPU-6500), a high-speed camera (IDP-Express R2000-F by Photoron), and a smartphone for vibration measurement.
4：Experimental Procedures and Operational Workflow:
Vibration measurements were taken eight times per scenario, with data processed to remove offsets and analyze frequency characteristics. Transmitter displacement was measured by projecting world coordinates to image coordinates using a pinhole camera model and comparing differences between consecutive frames.
5：Data Analysis Methods:
Frequency analysis was performed on vibration data, parameters (variances of Gaussian processes) were estimated from components above 20 Hz, and probability densities of displacement were compared using Kullback-Leibler divergence to evaluate model accuracy.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
