Abstract: The accuracy and stability of high-precision outdoor visual measurement systems are often significantly degraded by the complex coupling and time-lag effects of environmental factors such as temperature,humidity,and illumination.To address this challenge,this paper proposes a data-driven disturbance compensation framework that integrates multi-dimensional environmental features.The method first identifies key environmental disturbance sources through systematic analysis and innovatively introduces a time variable to capture the diurnal periodicity of measurement errors,thereby enabling the construction of a high-precision compensation model.Experimental results demonstrate that temperature,humidity,and illumination are primary influencing factors,and their effects exhibit significant time lags.Introducing time variable is critical for enhancing model performance,improving the compensation accuracy of non-linear models by approximately 30%,which significantly outperforms linear models and achieves sub-pixel level accuracy.This study validates the effectiveness of the proposed framework,providing theoretical basis and practical pathway for solving the environmental adaptability problem in long-term outdoor visual measurement systems.

Key words: visual measurement, environmental disturbance, error compensation, data-driven, machine learning