Edge computing-based imagery data preprocessing strategy

The advent of sensing and edge computing technologies has presented novel prospects for contemporary structural health monitoring. Structural health monitoring systems are collecting vast amounts of data over the time. However, the collected data may be polluted with inaccurate, missing, or irrelevant information, presenting a big challenge for data processing and analysis to retrieve structural conditions. To address this issue, our study introduces a novel edge computing-based data preprocessing strategy. An automated algorithm is developed to detect and discard anomalies for each sensor data type. Crucially, for image data, we perform frequency analysis on local devices to evaluate image clarity, determining the suitability of the gathered images for future analysis. Our methodology ensures that only photos meeting our requirements are uploaded to the central server, significantly reducing network congestion. This localized preprocessing not only diminishes the data transmission volume but also improves source data quality and usability, thereby alleviating the computational burden on the centralized system. Experimental results reveal that our strategy markedly elevates the efficiency and accuracy of structural health monitoring systems, providing a potential technological groundwork for upcoming practical applications.