Semantic Segmentation on Smartphone Motion Sensor Data for Road Surface Monitoring

Improving road safety is one of the critical issues for maintenance and management. Motion sensors embedded in smartphones to sense vibrations can be used detect rough surfaces when carried moving vehicles. Finding segments signal which reflect condition surface, however, a challenging task. This study proposes modified U-Net architecture with integrated bidirectional Long Short-Term Memory layers perform semantic segmentation on smartphone motion sensor data surface classification. Experiments show that using z-axis accelerometer gyroscope features, proposed method outperforms multiple existing algorithms.