Self-Updating Vehicle Monitoring Framework Employing Distributed Acoustic Sensing towards Real-World Settings

Abstract

The recent emergence of Distributed Acoustic Sensing (DAS) technology has facilitated the effective capture of traffic-induced seismic data. The traffic-induced seismic wave is a prominent contributor to urban vibrations and contain crucial information to advance urban exploration and governance. However, identifying vehicular movements within massive noisy data poses a significant challenge. In this study, we introduce a real-time semi-supervised vehicle monitoring framework tailored to urban settings. It requires only a small fraction of manual labels for initial training and exploits unlabeled data for model improvement. Additionally, the framework can autonomously adapt to newly collected unlabeled data. Before DAS data undergo object detection as two-dimensional images to preserve spatial information, we leveraged comprehensive one-dimensional signal preprocessing to mitigate noise. Furthermore, we propose a novel prior loss that incorporates the shapes of vehicular traces to track a single vehicle with varying speeds. To evaluate our model, we conducted experiments with seismic data from the Stanford 2 DAS Array. The results showed that our model outperformed the baseline model Efficient Teacher and its supervised counterpart, YOLO (You Only Look Once), in both accuracy and robustness. With only 35 labeled images, our model surpassed YOLO's mAP 0.5:0.95 criterion by 18% and showed a 7% increase over Efficient Teacher. We conducted comparative experiments with multiple update strategies for self-updating and identified an optimal approach. This approach surpasses the performance of non-overfitting training conducted with all data in a single pass.

Figures (14)

• Figure 1: Map of DAS Stanford 2 Array. In this study, the segments in use are denoted by red lines, whereas the unused ones are marked with blue.
• Figure 2: Stages of data preprocessing. (a) Raw data with only detrending. The horizontal axis represents the channel number, while the vertical axis corresponds to the time. (b) Images with model-based signal filtering. (c) The outputs of STA/LTA selection. (d) Final preprocessed images after rotation. (e) In the absence of STA/LTA selection, the preprocessed images focus solely on detecting negative amplitude values. The red box emphasizes the comparative clarity of traces, and the brown box illustrates the contrast between vehicle trace and background.
• Figure 3: Vehicle detection using STA/LTA for a single channel within the Stanford 2 Array (Sand Hill Road Array). (a) Amplitude-time domain; (b) STA/LTA-time domain. The red line signifies the STA/LTA value reaching the initiation threshold, marking the onset of an event, while the blue line indicates the STA/LTA value meeting the termination threshold, denoting the end of the event.
• Figure 4: Architecture of our vehicle detection model with shape prior loss. The basic architecture is an Efficient Teacher network xu2023efficient. Blue arrowed lines indicate training with unlabeled data, while orange arrowed lines indicate that with labeled data.
• Figure 5: Apparent false detection in the top right corner erroneously identifies one trace as even three in (a) and two in (b). The black numbers represent the confidence scores, which indicate the model's certainty regarding the accuracy of each prediction.