Vibration monitoring of rotating machines traditionally relies on contact sensors, but recent attention has shifted towards non-contact methods like vision sensors due to lower installation complexity and cost. However, existing research lacks exploration into environmental factors such as lighting conditions and the stability of sensor support. This study investigates the impact of different lighting conditions, frame rates, and unstable camera supports on vision sensor-based vibration monitoring accuracy. Two experiments were conducted: one measuring vibration from mass unbalance on a rotor under varied lighting, and another measuring mechanically induced vibration through a pipe using different sensor supports. Results indicate that despite the algorithm’s capability to extract accurate vibration data in terms of the frequency extracted, lighting variations and unstable support negatively affect the vibration amplitude results, even with pixel intensity tracking. Temporal requirements for the sensor were determined to be at least twice the frequency of interest. Future works could involve pre-processing methods for lighting, improved motion amplification to withstand video shakiness, and AI implementation for fault classification.