The stability of power towers, which are crucial supports for transmission lines, can be compromised by factors such as geological subsidence, leading to potential tilting and even collapse if not promptly detected. Traditional manual inspection methods are inefficient and insufficient to meet current demands. In order to solve the above problems, this paper proposes a tilt estimation algorithm for power pole tower based on unmanned aerial vehicle (UAV) laser point cloud. The method first uses the semantic segmentation network MFNet-S to extract the categories of the towers in the power scene. Subsequently, the density clustering density-based spatial clustering of applications with noise (DBSCAN) is used to cluster the categories of the towers to obtain a single tower. After obtaining the tower, the tower is vertically segmented, and the bounding box of each layer of the point cloud and its center point are calculated. After obtaining the set of coordinates of the center points, random sample consensus (RANSAC) straight-line fitting is applied to determine the tower′s tilting vector. Finally, the tilt of the tower is calculated according to the formula derived in this paper. Through the experiments on the untilted original tower and the manually placed tilted tower, the experimental results show that the algorithm proposed in this paper has a higher accuracy of tilt calculation, with an average error of only 0.15% and 0.52%, which is better than the 0.24% and 6.59% of the comparison method.