Accuracy Improvement of Vehicle Recognition by Using Smart Device Sensors.

Area of Science:

• Intelligent Transportation Systems
• Sensor Technology
• Machine Learning

Background:

• Smart devices are ubiquitous and possess sensors capable of collecting activity and physiological data.
• Existing research in intelligent transportation systems (ITS) can be expanded by integrating smart device sensor data.
• Vehicle recognition is a key component of ITS, with potential for enhancement through novel data sources.

Purpose of the Study:

• To explore the use of smart device sensors for vehicle recognition.
• To investigate how smartphone sensor data can be integrated into ITS research.
• To develop and evaluate a machine learning model for classifying vehicle types using sensor data.

Main Methods:

• Utilized smartphone accelerometer and gyroscope sensors to collect data from various vehicles (cars, buses, trains, bikes).
• Designed and implemented a 1D Convolutional Neural Network (CNN) model incorporating residual connections.
• Trained the CNN model on the collected sensor data for vehicle classification.

Main Results:

• The developed 1D CNN model achieved a prediction accuracy exceeding 98%.
• Demonstrated the feasibility of using common smartphone sensors for effective vehicle recognition.
• Sensor data from smart devices proved valuable for distinguishing between different vehicle classes.

Conclusions:

• Smart device sensors offer a viable and accurate method for vehicle recognition within ITS.
• The proposed CNN model demonstrates high performance in classifying vehicles based on sensor data.
• Future research directions include further refinement of sensor-based vehicle recognition techniques.