The construction industry, which is characterized by complex operations and dynamic environments, creates many safety hazards that demand continuous innovation for enhanced safety and efficiency. Traditional approaches to monitoring construction sites, which are often manual and error-prone, contribute to safety hazards and operational inefficiencies because they are usually manual operations. To address this, our research introduces the innovative use of the state-of-the-art You Only Look Once (YOLO) v8 architecture to develop a computer vision model specifically trained for the detection of construction vehicles with real-time detection and tracking abilities running on any low-wattage CPU. YOLO is a fast object detection architecture that treats detection as a single regression problem to predict bounding boxes around objects in images, which enables it to perform significantly faster than region-proposal-based methods. This study aims to address the limitations of traditional monitoring systems used at construction sites, particularly regarding real-time vehicle detection and tracking. It identifies a gap in the existing construction-site monitoring systems that lack AI integration, resulting in operational inefficiency and low safety risk detection. However, most AI models that support computer-vision-based operation monitoring require high-power processors, which are challenging to deploy in traditional monitoring systems. Although the recently introduced YOLO v8s has the potential to work with relatively low-power processors, its potential in construction operation monitoring has not yet been explored. Therefore, the primary objective of this study is to train a robust and accurate YOLO v8s model that can function effectively under diverse and challenging conditions at construction sites. Our methodology encompasses the comprehensive data collection of various construction vehicle images and rigorous model training and optimization to ensure high accuracy and real-time processing capability optimized by OpenVino from Intel. The findings reveal that our optimized YOLO v8-based model significantly outperforms traditional manual monitoring methods that do not have AI monitoring capabilities while demonstrating high accuracy in real-time vehicle detection and tracking. This research catalyzes the integration of advanced computer vision technology into construction management to enhance on-site safety and operational efficiency, contributing to its adoption in the field.