A novel reconstruction method for aerial LiDAR point clouds building models is proposed in this paper to obtain valid roof building models. There are two problems in the reconstruction. Firstly, in the process of primitive segmentation, due to the uneven density of the point cloud, there is the problem of over-segmentation or under-segmentation of the plane, resulting in the inability to extract concise and suitable building planes, which in turn affects the reconstruction topology. Secondly, in the model construction process, due to the variety and complexity of building structures, obtaining regular, compact and topologically correct surface models from sparse and noisy point clouds is still a challenge. To address the first problem, the initial primitives are first obtained using an improved multi-resolution supervoxel based region growing segmentation algorithm. Then, a new progressive primitive fusion algorithm BPPM (belief propagation primitive merge) is proposed to optimize the fragmented primitives. For the second problem, the CKL (Corner KLine) regularization algorithm is first proposed to obtain the building footprints, and then the height map is constructed from the point cloud to extract the polyline of the building boundaries and deduce the vertical planes. Finally, a new energy function is proposed to encourage the selection of the recommended combination of model planes to obtain a compact and valid roof reconstruction model. Experiments are performed on different roof point clouds to quantitatively evaluate the proposed method and qualitative experiments are conducted with comparative experiments to confirm the effectiveness of the algorithm.

Accurate roof plane extraction is a crucial step in constructing a three-dimensional model for buildings. Due to the significant differences in size and shape of building roofs in airborne light detection and ranging point clouds, many existing plane extraction methods are struggling to achieve good performance. To solve the above problem, a bottom–up method for roof plane extraction is proposed in this paper. Starting with the division of the roof point cloud into voxels, the initial planes are obtained in the voxels. The initial planes are then expanded by a parameter-adaptive region growing algorithm. Then, the grown planes are merged according to predefined constraints. Finally, an energy minimization-based method is applied to optimize the results of roof plane extraction. The performance of our proposed method is evaluated on the Vaihingen dataset and the DALES dataset. Experiments demonstrate that our proposed method achieves a superior roof plane extraction result.

Remote-sensing images play a crucial role in a wide range of applications and have been receiving significant attention. In recent years, great efforts have been made in developing various methods for intelligent interpretation of remote-sensing images. Generally speaking, machine learning-based methods of remote-sensing image interpretation require a large number of labeled samples and there are still not enough annotated datasets in the field of remote sensing. However, manual annotation of remote-sensing images is usually labor-intensive and requires expert knowledge and the accuracy of annotation results is relatively low. The goal of this paper is to propose a novel tile-level annotation method of remote-sensing images to obtain remote-sensing datasets which are well-labeled and contain accurate semantic concepts. Firstly, we use a set of images with defined semantic concepts to represent the training set and divide them into several nonoverlapping regions. Secondly, the color features, texture features, and spatial features of each region are extracted, and discriminative features are obtained by the weight optimization feature fusion method. Then, the features are quantized into visual words by applying a density-based clustering center selection method and an isolated feature point elimination method. And the remote-sensing images can be represented by a series of visual words. Finally, the LDA model is used to calculate the probabilities of semantic categories for each region. The experiments are conducted on remote-sensing images which demonstrate that our proposed method can achieve good performance on remote-sensing image tile-level annotation. The implications of our research can obtain annotated datasets with accurate semantic concepts for intelligent interpretation of remote-sensing images.