Applying wavelet packet decomposition and one-class support vector machine on vehicle acceleration traces for road anomaly detection

摘要: Road condition monitoring through real-time intelligent systems has become more and significant due to heavy road transportation. conditions can be roughly divided into normal anomaly segments. The number of former should much larger than the latter for a useable road. Based on nature monitoring, detection is applied, especially pothole in this study, using accelerometer data riding car. Accelerometer were first labeled segmented, after which features extracted by wavelet packet decomposition. A classification model was built one-class support vector machine. For classifier, some segments used train classifier left all potholes testing stage. results demonstrate that 21 detected reliably study. With low computing cost, proposed approach promising application.