Semantic segmentation is a fundamental task in remote sensing image understanding. Recently, Deep Convolutional Neural Networks (DCNNs) have considerably improved the performance of the semantic segmentation of natural scenes. However, it is still challenging for Very High Resolution (VHR) remote sensing images. Due to the large and complex scenes as well as the influence of illumination and imaging angle, it is particularly difficult for the existing methods to accurately obtain the category of pixels at object boundaries---the so-called boundary blur. We propose a framework called Boundary-Aware Semi-Supervised Semantic Segmentation Network ( $\mathrm{BAS}^4$ Net), which obtains more accurate segmentation results without additional annotation workload, especially at the object boundaries. The Channel-weighted Multi-scale Feature (CMF) module balances semantic and spatial information and the Boundary Attention Module (BAM) weights the features with rich semantic boundary information to alleviate the boundary blur. Additionally, to decrease the amount of difficult and tedious manual labeling of remote sensing images, a discriminator network infers pseudo-labels from unlabeled images to assist semi-supervised learning and further improves the performance of the segmentation network. To validate the effectiveness of the proposed framework, extensive experiments have been performed on both the ISPRS Vaihingen dataset and the novel remote sensing dataset AIR-SEG with more categories and complex boundaries. The results demonstrate a significant improvement of accuracy especially on boundaries and for small objects.

Recent advances in the availability of very high-resolution (VHR) satellite data together with efficient data acquisition and large area coverage have led to an upward trend in their applications for automatic 3-D building model reconstruction which require large-scale and frequent updates, such as disaster monitoring and urban management. Digital Surface Models (DSMs) generated from stereo satellite imagery suffer from mismatches, missing values, or blunders, resulting in rough building shape representations. To handle 3-D building model reconstruction using such low-quality DSMs, we propose a novel automatic multistage hybrid method using DSMs together with orthorectified panchromatic (PAN) and pansharpened data (PS) of multispectral (MS) satellite imagery. The algorithm consists of multiple steps including building boundary extraction and decomposition, image-based roof type classification, and initial roof parameter computation which are prior knowledge for the 3-D model fitting step. To fit 3-D models to the normalized DSM (nDSM) and to select the best one, a parameter optimization method based on exhaustive search is used sequentially in 2-D and 3-D. Finally, the neighboring building models in a building block are intersected to reconstruct the 3-D model of connecting roofs. All corresponding experiments are conducted on a dataset including four different areas of Munich city containing 208 buildings with different degrees of complexity. The results are evaluated both qualitatively and quantitatively. According to the results, the proposed approach can reliably reconstruct 3-D building models, even the complex ones with several inner yards and multiple orientations. Furthermore, the proposed approach provides a high level of automation by limiting the number of primitive roof types and by performing automatic parameter initialization.

In recent years, Fully Convolutional Networks (FCN) have led to a great improvement of semantic labeling for various applications including multi-modal remote sensing data. Although different fusion strategies have been reported for multi-modal data, there is no in-depth study of the reasons of performance limits. For example, it is unclear, why an early fusion of multi-modal data in FCN does not lead to a satisfying result. In this paper, we investigate the contribution of individual layers inside FCN and propose an effective fusion strategy for the semantic labeling of color or infrared imagery together with elevation (e.g., Digital Surface Models). The sensitivity and contribution of layers concerning classes and multi-modal data are quantified by recall and descent rate of recall in a multi-resolution model. The contribution of different modalities to the pixel-wise prediction is analyzed explaining the reason of the poor performance caused by the plain concatenation of different modalities. Finally, based on the analysis an optimized scheme for the fusion of layers with image and elevation information into a single FCN model is derived. Experiments are performed on the ISPRS Vaihingen 2D Semantic Labeling dataset (infrared and RGB imagery as well as elevation) and the Potsdam dataset (RGB imagery and elevation). Comprehensive evaluations demonstrate the potential of the proposed approach.

Datasets play a key role in developing superior building detection approaches. However, most of the previous work focuses on accurate building masks and scale expansion, while the categories are always missing, which hinders the further analysis of urban development and cultures. Therefore, we propose a benchmark for building detection and fine-grained classification from very high-resolution (VHR) satellite imagery. An extensive annotation is performed for about 0.5 million building instances with 12 fine-grained roof types and individual polygons. The annotation of building functions of two cities in the previous version (UBCv1) [1] is also integrated. To ensure the building variety, it consists of VHR optical images of 20 unique cities worldwide with various landforms and styles of architecture. Its variety and fine-grained categories pose great challenges and meanwhile provide a foundation for the building extraction and fine-grained classification on a global scale. Besides, 17 cities are provided with finely aligned Synthetic Aperture Radar (SAR) images, which can be employed for the development and evaluation of approaches optionally based on optical, SAR, or multi-modal images. Significantly, the proposed benchmark is used as the base of the 2023 IEEE GRSS Data Fusion Contest [2]. The dataset and codes of the baseline methods are available at: https://github.com/AICyberTeam/UBC-dataset/tree/UBCv2.

We propose a pipeline for the detection as well as modeling of individual buildings based on multi-source images. It allows to consistently reconstruct whole buildings at Level of Detail 3 (LoD3): the roof from airborne images and the facades including elements such as windows and doors mainly from terrestrial images. We employ a parametrized top-down model – the “shell model” – with the roof as well as the facades semantically and geometrically integrated. This generative model fosters stability for building detection by enabling the use of multi-source data and offers flexibility in modeling by means of a fully CAD-compatible integration of building components. Experiments performed on imagery from different terrestrial and airborne (Unmanned Aerial Vehicle – UAV) cameras demonstrate the potential of the approach.