This paper presents a method for robust tracking of multiple pendulums using video analysis under imperfect lighting conditions. The experimental setup consists of five pendulums recorded by a stationary camera at 30 frames per second. Object detection is performed using color-based segmentation in the HSV color space implemented in Python with the OpenCV library. Two approaches are compared: a baseline method with a non-robust color mask and an improved method with optimized masking parameters. The quality of tracking is evaluated using temporal derivatives of the detected trajectories, specifically the second-order derivative, which highlights high-frequency noise caused by unstable detection. The results show a significant reduction in noise and improved trajectory smoothness when using the optimized mask. The proposed approach provides a simple and effective solution for motion tracking in suboptimal lighting conditions.
Science Journal 