High-precision vision-based measurement of key pantograph parameters for high-speed trains using Siamese networks and MOTFSE

The pantograph transfers electric energy from the catenary to high-speed trains and is critical for electrified railways. However, excessive motion can cause pantograph-catenary malfunctions, including catenary breaks, arcing, and mechanical damage to the pantograph. Therefore, detecting pantograph head motion is crucial for reliable high-speed train operation. This paper employs an improved Siamese network to measure its horizontal/vertical displacement and estimate the dynamic envelope. A multi-branch Siamese network is also developed to extract displacements of multiple targets for stagger value estimation. Additionally, the proposed MOTFSE algorithm (Multi-Object Tracking and Optical Flow Speed Estimation) fuses tracking and optical flow results to estimate the vertical vibration speed. Experimental results show the proposed algorithm achieves over 90% tracking accuracy and precision for the pantograph head. It enables accurate measurement of the pantograph's dynamic envelope, contact wire stagger value, and head vertical oscillation velocity, offering a viable solution for practical application.