Video-image based mobile scaffolding safety detection using deep learning and internet of things (IOT) | |
| Author | Chowdhury, Md. Rakibul Islam |
| Call Number | AIT Thesis no.CM-23-06 |
| Subject(s) | Internet of things Deep learning (Machine learning) Neural networks (Computer science) Video surveillance |
| Note | A thesis submitted in partial fulfilment of the requirements for the degree of Master of Engineering in Construction, Engineering and Infrastructure Management |
| Publisher | Asian Institute of Technology |
| Abstract | Deep neural networks (DNNs) have excelled in the field of object detection. When it comes to speed and accuracy, YOLO is competitive with other state-of-the-art DNN based object identification methods. A scaffold is a vital part of every construction site, but it may also be a source of serious injury or even death. However, it takes a lot of time and effort for safety managers to monitor the scaffolding based on their own subjective observations. This research aims to solve this problem by creating a deep learning-Internet of Things (IoT) based mobile scaffolding safety detection system with 12 individual parts. The process has three stages: dataset creation using actual mobile scaffolding photos, web application development, and NodeMCU and Blynk library connectivity for signal processing. Following picture enhancement and processing, the dataset has 3040 images. Prediction accuracy for the classes was over 60% prior to the dataset extension. However, with the improved and expanded dataset, class prediction accuracy increased by almost 20%. Four different type of lighting conditions used in the pilot study, which were broken down into "Natural," "Artificial," "Low" and “Night Light” categories, and also “untrained” image test also applied to assess the performance of the model. False positives were found for just one scaffolding but performance significantly dropped for night image. The novel aspect of the study is the identification of movable scaffolding safety by use of twelve principal components. NodeMCU's capacity to create reports and interpret work signals sets it distinct from other IoT devices. This report may be used by the safety engineer as evidence in the case of an accident. The report may be used as a reference for workplace safety by the safety engineer or the project authorities. Everyone participating in the building project, not only the safety engineer, will benefit from this research. |
| Year | 2023 |
| Type | Thesis |
| School | School of Engineering and Technology |
| Department | Department of Civil and Infrastucture Engineering (DCIE) |
| Academic Program/FoS | Construction Engineering and Infrastructure Management (CM) |
| Chairperson(s) | Hadikusumo, Bonaventura H. W. |
| Examination Committee(s) | Avirut Puttiwongrak;Sarawut Ninsawat |
| Scholarship Donor(s) | Indorama Ventures (IVL) Foundation Scholarship |
| Degree | Thesis (M. Eng.) - Asian Institute of Technology, 2023 |