Road vehicle recognition and classification using magnetic field measurement

Publication Type:
Thesis
Issue Date:
2018
Metrics:
Full metadata record
Files in This Item:
Filename Description Size
01front.pdf165.93 kB
Adobe PDF
02whole.pdf1.92 MB
Adobe PDF
This dissertation presents a road vehicle detection approach for intelligent transportation systems. This approach uses a roadside-installed, low-cost magnetic sensor and associated data collection system. The system measures magnetic field changes to count, detect, and classify passing vehicles into a number of vehicle types. We compare each vehicle using dynamic time warping (DTW), then extend Mel-Frequency Cepstral Coefficients to analyse the vehicles’ magnetic signals and extract them as vehicle features using the representations of cepstrum, frame energy, and gap cepstrum of magnetic signals. There are three directions (X-axis, Y-axis, and Z-axis directions) in the earth’s magnetic field. We design one- (X-axis direction) and three-dimensional (i.e. X-axis, Y-axis, and Z-axis direction) map algorithms using Vector Quantisation to classify the vehicle magnetic features according to four typical vehicle types for the Australian suburbs: sedan, van, truck, and bus. We also compared experimental results between these two methods. Results show that our approach achieves a high level of accuracy for vehicle detection and classification. In the end, we found that filtering raw magnetic measurement signals can significantly influence vehicle recognition accuracy. Compared with the one-dimensional map, we reached the highest accuracy of vehicle classification in our test data using the three-dimensional map.
Please use this identifier to cite or link to this item: