Uncompliant behavior regarding the use of Personal Protective Equipment (PPE) is one of the major factors contributing to workplace injuries within the construction field. Meanwhile, current static camera surveillance systems are limited by a fixed field of view, rendering them unable to continuously track mobile targets. To address this challenge, this research proposes an active surveillance system that integrates the YOLOv11 object detection algorithm with a Nonlinear Model Predictive Controller (NMPC) designed for Pan-Tilt camera mechanism. The approach automates PPE violation recognition and target tracking, factoring in the physical motion limitations of the hardware. Numerical simulation results indicate that the NMPC significantly surpasses the conventional PID controller. Notably, NMPC achieves fast and stable system response, while the PID controller fails to stabilize the system, resulting in persistent oscillations. This study validates the feasibility and effectiveness of a unified framework, paving the way for automated, intelligent, and continuous safety monitoring systems in dynamic construction environments.