In this paper, the UAV photogrammetry data of two hills and two flat sampling areas are selected. The image data are processed by pixel Grid software through different operation methods, and the effect of the two methods is compared. The results show that: after the distortion correction, the image is rotated manually or the flight angle in the pos data is directly changed, which has no obvious effect on the error in the number of iterations of the automatic connection point. When the distortion correction is not corrected, the photo is rotated, and the angle of the Omega, Phi and Kappa angles is obtained from the UAV autopilot system, and makes the corner element higher precision when the automatic connection point is generated. Add Kappa to the POS data by 90, and then correct distortion at the same time choice of rotating pictures; Set the Omega, Phi is 0, Kappa angle from the IMU, this method avoids the influence of the angle deviation caused by the tilt or jitter of the flight process, so it is better to line elements when the automatic connection point is generated.