For safety on construction sites, it is important to track the location, posture and movement of construction equipment (e.g. excavators, trucks, cranes, and bulldozers) and workers. Currently with the aid of surveillance cameras on construction sites, computer vision techniques can be applied to process the captured videos and images captured, thereby identifying the site conditions and potential hazards. This project will develop a methodology framework based on computer vision techniques that can automatically identify the posture of construction equipment and workers from videos captured on construction sites. The same techniques will also be leveraged for enhancing walkability on construction sites and public areas.
(1) Build image library for training. (2) Label objects and posture in images in the image library. (3) Develop an engine for image processing and object identification based on computer vision and deep learning techniques. (4) Test the engine and evaluate its performance.
(Students with computer programming background are preferred.)
(1) Students will be able to understand and apply state-of-the-art computer vision and deep learning techniques, with applications on construction and building environments. (2) Students will be able to compare and evaluate different computer vision and deep learning techniques. (3) Students will be able to recommend setups for surveillance cameras with computer vision and deep learning techniques, to enhance automated construction site monitoring and walkability assessment.