In river management, it is important to obtain ice velocity quickly and accurately during ice flood periods. However, traditional ice velocity monitoring methods require buoys, which are costly and inefficient to distribute. It was found that UAV remote sensing images combined with machine vision technology yielded obvious practical advantages in ice velocity monitoring. Current research has mainly monitored sea ice velocity through GPS or satellite remote sensing technology, with few reports available on river ice velocity monitoring. Moreover, traditional river ice velocity monitoring methods are subjective. To solve the problems of existing time-consuming and inaccurate ice velocity monitoring methods, a new ice velocity extraction method based on UAV remote sensing technology is proposed in this article. In this study, the Mohe River section in Heilongjiang Province was chosen as the research area. High-resolution orthoimages were obtained with a UAV during the ice flood period, and feature points in drift ice images were then extracted with the scale-invariant feature transform (SIFT) algorithm. Moreover, the extracted feature points were matched with the brute force (BF) algorithm. According to optimization results obtained with the random sample consensus (RANSAC) algorithm, the motion trajectories of these feature points were tracked, and an ice displacement rate field was finally established. The results indicated that the average ice velocities in the research area reached 2.00 and 0.74 m/s, and the maximum ice velocities on the right side of the river center were 2.65 and 1.04 m/s at 16:00 on 25 April 2021 and 8:00 on 26 April 2021, respectively. The ice velocity decreased from the river center toward the river banks. The proposed ice velocity monitoring technique and reported data in this study could provide an effective reference for the prediction of ice flood disasters.