Accurate registration of photogrammetric point clouds is essential for reliable geometric monitoring of slope instabilities. Although low-cost imagery provides an easy and inexpensive option for data acquisition, it often suffers from doming and drift-induced errors during reconstruction, resulting in geometric distortions within the point cloud. To mitigate these effects, a method for improving point cloud registration based on an segment-wise Iterative Closest Point (ICP) algorithm is presented. The approach subdivides the point cloud into small, locally rigid segments that are each aligned individually with a reference model. To ensure smooth transitions between neighboring segments, an interpolation of the local transformations is applied. The method is evaluated using data from Mt. Hochvogel, where several video sequences of the summit area are captured with low-cost cameras. For registration and reference, terrestrial laser scanning (TLS) data and a photogrammetric point cloud from an earlier epoch are used. The latter is used for change analysis based on M3C2 distance computation. The results demonstrate that the described distortions can be substantially reduced, particularly in comparison to a single global ICP registration, providing an effective means of improving registration accuracy for low-cost photogrammetric monitoring, especially when ground control point (GCP) measurements are not feasible. The overall registration quality depends strongly on the reconstruction quality and the chosen part size. Segments that are too small may not provide sufficient spatial structure to ensure a stable ICP solution in all three dimensions, whereas larger segments limit the local adaptability.