RESEARCH ON THE SENSITIVITY OF THE DISTANCE MEASUREMENT METHOD BASED ON THE FACEMESH ALGORITHM

Abstract

Relevance. The article explores a method for estimating the distance to a user's face under varying lighting conditions, camera resolution, and head tilt angles. This method is significant for developing adaptive computer vision systems in medical, educational, and navigational applications. The goal of the research is to analyze the accuracy of distance measurement to an indoor object using a geometric approach based on the FaceMesh algorithm under variable external factors. The tasks include determining the impact of lighting quality, head position, and camera resolution on measurement accuracy; implementing a distance estimation algorithm using facial landmarks; conducting experiments; and formulating recommendations for the proposed method. The following methods were used: monocular geometry with the FaceMesh algorithm and real-time video processing techniques. As a result of the experimental studies, it was confirmed that the optimal conditions for accurate distance measurement are a frontal face position relative to the camera, a resolution of 640×480, and lighting above 200 lux. The error increases at tilt angles exceeding 15° and under low lighting. Under these conditions, the FaceMesh algorithm demonstrates an error of less than 3% at distances up to 90 cm. Conclusions. The method is suitable for use in real-time mobile systems with limited resources and has potential for further integration into multimodal recognition systems and adaptive navigation platforms.