[Objective] Slope gradient data was accurately obtained to address the limitation of slope gradient underestimation using the freely downloaded 30 m resolution digital elevation model (DEM) for farmland in Northeast China, in order to provide important data support for quantitatively evaluating soil erosion in the rolling hilly regions. [Methods] A 5 cm resolution DEM was generated from drone survey images and resampled to obtain 1, 5, and 12.5 m DEM resolutions. Combined with the 30 m DEM resolution, the optimal DEM resolution for slope gradient extraction in the study area was identified. Additionally, the histogram matching method was used to establish a slope gradient conversion model between the 30 m DEM resolution and the optimal DEM resolution for each slope gradient category. [Results] ① The slope gradient distributions derived from the five DEM resolutions indicated that the 1 m and 5 m DEM resolutions exhibited a strong similarity to the slope gradient distribution of the 5 cm DEM. Given that the 5 m DEM resolution corresponds to a 1∶10,000 scale topographic map, the 5 m DEM resolution was optimal for constructing the slope gradient conversion model. ② Using the histogram matching method, a univariate linear model and a univariate quadratic non-linear model were developed for slope gradient conversion between the 30 m and 5 m DEM resolutions across different slope gradient segments. The linear conversion model was suitable for slopes less than 7°, while the non-linear model was more appropriate for slopes greater than 7°. ③ After applying both linear and non-linear conversion models, the frequency distribution of slope gradients extracted from the 30 m DEM resolution closely matched that of the 5 m DEM resolution, significantly improving covariance and correlation coefficients. This reflected that the slope gradients after conversion from the 30 m DEM resolution can accurately represent ground undulation; additionally, the optimization results from the non-linear conversion model were superior to those from the linear conversion model. [Conclusion] The 5 m DEM resolution is the optimal resolution for extracting slope data in the study area. The developed conversion model for low-to-high resolution slope gradients showed that the non-linear slope conversion model has a better optimization effect than the linear slope conversion model.