The increasing amount of 3D data collected accumulates in datasets that can be used for change detection. In particular, multi-epoch datasets representing a single scene enable assessing rates of change. To track individual changes, points have to be identified. However, in unstructured point clouds, identifying co-located points in different epochs is computationally demanding. In the proposed method, horizontally co-located points are identified across epochs by discretising point clouds and employing Morton code spatial indexing. Identifying changed points enables estimating change dimensions at a point level. Additionally, identifying points enables masking points in change visualisation, where points in earlier epochs occlude newer points in subsequent epochs. Finally, discarding duplicate points from unchanged parts of a dataset mitigates file-size problems in sharing and visualising large datasets. The method was verified with two datasets. Based on the presented results, the method can estimate depth, area, and volume with high accuracy. With sufficiently dense and accurately registered laser scanned test data, the Mean Absolute Percentage Errors (MAPE) in volume estimates amounted to 2.9%. Total MAPE for all estimates in two synthetic datasets was 0.3%, indicating the proposed method provides reasonable estimates and is a feasible method for estimating dimensions for the detected changes.
Sarlin et al. (2026) studied this question.