ABSTRACT
Bridge weighing in motion (BWIM) detects heavy vehicles passing a sensor or multiple sensors. Sensor technology with rapid development enables us to monitor bridges with multiple types of sensors and cameras. These sensors are installed at different positions and require synchronization. In this paper, we propose a novel vehicle detection framework for the multiple strain sensors BWIM systems based on a novel deep sensor-fusion approach. The proposed framework is composed of three components. The first component collects and denoises the strain sensor signal. Then, the cleaning signals are processed by the Wavelet transform in the second component to obtain the corresponding coefficient matrices. Finally, the third component feeds the coefficient results into an improved deep neural network. It extracts complex features through multiple neural layers and can detect vehicle properties from multi-view signals. The performance of the proposed framework is evaluated on real-world bridge health monitoring datasets. The results of the experiments show that the proposed framework is reliable and effective for vehicle detection in the BWIM system.
