4.1.2. Experimental Results

The calibration results of both methods for one frame showing 30 GCPs are displayed in Figure 6. The calibration results of the two methods in this case could be distinguished by the naked eye, even when details were closely examined (Figure 6).

**Figure 6.** The calibration results of (**a**) the proposed method and (**b**) the RFM method.

To make accurate comparisons between the two methods, the calibrated residuals of the RCPs were employed in the statistical analysis. The mean value and root mean square error (RMSE) of the calibrated residuals were used as the measurable indicators. The variation trends of the mean value and the RMSE for the total GCPs across the two methods are shown in Figure 7. At least 19 GCPs were required to solve the coefficients of RFM; the RFM method did not work with 5, 10, or 15 GCPs. As the number of GCPs increased, the mean value of the two methods gradually decreased, as shown in Figure 7a. Similarly, the RMSE of the two methods gradually decreased in Figure 7b as the number of GCPs increased, i.e., the calibration performance of the two methods improved when the number of GCPs increased. The RFM method was superior in calibration performance when

presented with enough GCPs. However, the proposed method did well when insufficient GCPs were available. Therefore, the proposed method achieved a better calibration performance in situations with insufficient GCPs compared with the RFM method.

**Figure 7.** The variation trends of (**a**) mean value and (**b**) root mean square error (RMSE) of the number of GCPs across the two methods.
