Multimedia Sensor Dataset for the Analysis of Vehicle Movement
Wonhee Cho, Seon Ho Kim · 2017
With applications ranging from basic trajectory calculations to complex autonomous vehicle operations, detailed vehicle movement analysis has been getting more attention in academia and industry. So far, real-data driven analysis, e.g., utilizing advanced machine-learning, has used data from sensors such as GPS and accelerometer. However, such research requires quality datasets to enable accurate analysis. To that end, we have collected real vehicle movement data, Multimedia Sensor Data, that contain synchronized sensor data in fine granularity such as GPS, accelerometer, digital compass, gyroscope, and, most importantly, matching real video images recorded at driving time. These real video images provide a way to accurately label the sensor data in generating a quality dataset, e.g., a training dataset. Then, we performed preprocessing steps to clean and refine the raw data, subsequently converted the results into csv files, which are compatible with a wide variety of analysis tools. We also provided sample cases to demonstrate methods of identifying abnormal driving patterns such as moving over a speed bump. This dataset will be useful for researchers refining their analyses of vehicle movements.