Error modeling and subregion correction method for close-range photogrammetry based on unidirectional accumulation

Close-range photogrammetry (CRP) is a well-established technique for shape 3D measurement, valued for its accuracy and flexibility. However, the increasing accuracy demands of modern manufacturing reveal a persistent limitation: accumulated errors. This paper analyzes and simulates error propagation in CRP based on the collinearity equation and experimental data, showing that the errors exhibit a tendency of unidirectional accumulation. To address this issue, an accumulated error prediction model (AEPM) is developed to quantitatively estimate error distribution. Building on AEPM, a subregion correction method is proposed, in which measurement data are divided using subregion descriptor (SRD) and corrected independently under 3D alignment constraints. Experiments conducted over a range of 6.8m×6.5m demonstrate that AEPM achieves an estimation accuracy of 0.09 mm, while the correction method improves local and global accuracies to 0.016 mm and 0.05 mm, representing improvements of 75.9 % and 60 % compared with uncorrected results. These findings verify the effectiveness of the proposed approach in mitigating accumulated errors and enhancing the accuracy of CRP measurements.