Novel blockchain deep learning framework to ensure video security and lightweight storage for construction safety management

In construction management, video data tampering behavior like manual forging and deletion can negatively impact on-site safety and accident accountability. Blockchain technology holds the potential to address this issue by leveraging distributed ledger characteristics. However, blockchain's limited storage capacity and block size make it difficult to upload large-sized construction data such as daily monitoring video. Furthermore, it is unnecessary to store all construction data in any case. Therefore, this study proposes a blockchain deep learning framework that focuses on how to efficiently extract and securely store key information (i.e., video summarization that involves worker's unsafe behavior) on-blockchain for data traceability. The framework involves a novel data-driven and rule-based keyframe extraction (DRKE) model to lightweight large-sized construction video in the nascent field of deep learning and blockchain combination. To define parameters for the DRKE model, specific construction rules (e.g., people's unsafe behavior-based and people-based rules) have been predefined. This framework has been evaluated, and the results demonstrate its capability for effective video security storage, facilitating practical needs in construction management. The study extends existing research and provides a practical solution for large-sized construction video storage with security and lightweight considerations. The proposed video security storage and data lightweight process offers substantial benefits to construction management, such as streamlined accident investigation and accountability and improved on-site work efficiency, contributing to the smooth progress of construction projects.