An Unscented Kalman Filter Algorithm for Rebar Signal Processing Based on an Adaptive Forgetting Factor

In the inspection of concrete structures, the precise localization of rebar serves as the foundation for depth measurement and subsequent information analysis. Addressing the challenges of signal noise suppression and measurement accuracy enhancement in rebar detection, this paper proposes an unscented Kalman filter algorithm based on an adaptive forgetting factor. Conventional forgetting factor approaches exhibit poor adaptability and low filtering accuracy. When the noise is non-stationary or subject to abrupt disturbances, the accuracy becomes even less reliable, making it challenging to maintain stable filtering performance in complex environments. To overcome these shortcomings, this paper introduces an adaptive forgetting factor strategy that dynamically adjusts the noise covariance matrix based on real-time noise characteristics. This enables the filter to self-optimize its performance under complex conditions, thereby improving measurement accuracy. Furthermore, the algorithm incorporates a residual update mechanism, leveraging residuals to adjust the measurement noise covariance in real time, further enhancing the filtering efficacy. Simulation results demonstrate that the proposed algorithm significantly improves signal filtering accuracy in complex noise environments, outperforming traditional conventional forgetting factor approaches in detector position estimation and precise rebar localization.