Monitoring Data Fusion Based On Visual And Inertial Sensing Systems

Bridges have been subject to vehicular and pedestrian loads,as well as natural loads such as wind and seismic activity,since they were constructed and put into economic use.These actions can lead to significant damage to the structure of the bridge over time.Additionally,building materials may deteriorate with time,resulting in issues such as aging and cracking,which further reduce the structural carrying capacity of the bridge and may lead to inadequate safety requirements.Therefore,it is crucial to establish a comprehensive bridge structural health monitoring system.By installing sensors and data acquisition equipment,the structure and functionality of the bridge can be monitored in real-time,and the monitoring data can be analyzed and processed to ensure the safety and reliability of the bridge.Displacement is an important indicator for evaluating the condition and performance of bridges,which is used to measure the dynamic response of the bridge under external loads.Compared with traditional measurement methods,computer vision recognition can obtain the displacement information of the structure without contact,which will not cause any influence on the measured structure.However,computer vision recognition has high requirements for lighting and environment,is susceptible to obstruction,its accuracy is limited,and the grayscale changes are unstable,which also restricts its application range.This paper aims to address these issues by fusing vision and inertial sensors to eliminate errors and biases in visual sensors,improve the accuracy of visual displacement monitoring,and enhance the robustness and stability of the monitoring system,providing reliable basis for bridge displacement monitoring.The main research contents of this paper include the following aspects:(1)An acceleration acquisition system based on Raspberry Pie has been established.Raspberry Pie collects vertical acceleration data of the structure by controlling inertial sensors,and plots the data into a data graph for visualization through the Matplotlib module.In the laboratory,a comparative verification has been conducted on pedestrian bridges under different working conditions using traditional sensor equipment to ensure the accuracy of the system.(2)Established a structure displacement recognition system based on Raspberry Pi.The system captures displacement videos by installing Open CV environment and camera device on Raspberry Pi,and filters the images to remove noise.The LK optical flow method is used to recognize the structural pixel displacement of specific target points on the bridge in the video,and the displacement is converted to real physical displacement through a scale factor.Through experiments,it has been verified that the system has high accuracy and reliability under different working conditions.(3)The Kalman filter principle was used to fuse data from visual and inertial sensors,and an estimation method for structural vibration displacement was achieved.In laboratory experiments on a pedestrian bridge,the two sensors mounted on the Raspberry Pi worked together at low-frequency vibrations to improve the accuracy of structural displacement estimation.At the same time,using a camera instead of the Raspberry Pi camera at highfrequency vibrations could estimate the structure’s displacement more accurately.