Digital close-up photogrammetry technology can save time for construction project managers by eliminating the need to visit field sites for power line foundation quality inspection. However, existing projection density algorithms for power line foundation point cloud classification suffer from low accuracy. This paper proposes a multi-view grid-based point cloud classification algorithm for power line foundation. Firstly, the elevation difference features of the grid are constructed to coarsely classify the point clouds; then the plane point clouds and elevation point clouds obtained from the coarse classification are projected to the XOZ and XOY planes and divided into grids; finally, according to the elevation relationship between different types of point clouds in the point cloud, the seed grids for various types of point clouds are selected for neighborhood boundary tracking, and the boundary grids are distinguished according to the projected density of the grids, so as to realize the classification of the point clouds. The performance of the algorithm is evaluated by manually discriminating and counting the number of various types of point clouds as the real value, and the results show that the classification accuracy of the algorithm is better than that of the classification algorithm based on the projection density. The maximum measurement error of the length of the two groups of power line foundation is 2.137 mm, and the maximum measurement error of the width is 2.714 mm, which does not exceed the specification requirement of ±10 mm, and the maximum measurement error of the spacing of the same group of ground bolts is 3.124 mm, which does not exceed the specification requirement of ±3.6 mm. this method can realize the quality inspection of the power line foundation, and the measurement precision meets the construction and acceptance specification requirements.