Accelerometers are widely used in modern vehicular technologies to automatically detect and characterize road anomalies such as potholes and bumps. However, measurements from an accelerometer are usually plagued by high noise levels, which typically increase the false alarm and misdetection rates of an anomaly detection system. To address this problem, we have developed in this paper an adaptive threshold estimation technique to filter accelerometer measurements effectively to improve road anomaly detection and characterization in vehicular technologies. Our algorithm decomposes the output signal of an accelerometer into multiple scales using wavelet transformation (WT). Then, it correlates the wavelet coefficients across adjacent scales and classifies them using a newly proposed adaptive threshold technique. Furthermore, our algorithm uses a spatial filter to smoothen further the correlated coefficients before using these coefficients to detect road anomalies. Our algorithm then characterizes the detected road anomalies using two unique features obtained from the filtered wavelet coefficients to differentiate potholes from bumps. The findings from several comparative tests suggest that our algorithm successfully detects and characterizes road anomalies with high levels of accuracy, precision and low false alarm rates as compared to other known methods.
Anomaly Detection of Network Traffic Based on Prediction and Self-Adaptive Threshold
