Deep learning methods, especially Convolutional Neural Networks (CNN) and Vision Transformer (ViT), are frequently employed to perform semantic segmentation of high-resolution remotely sensed images. However, CNNs are constrained by their restricted receptive fields, while ViTs face challenges due to their quadratic complexity. Recently, the Mamba model, featuring linear complexity and a global receptive field, has gained extensive attention for vision tasks. In such tasks, images need to be serialized to form sequences compatible with the Mamba model. Numerous research efforts have explored scanning strategies to serialize images, aiming to enhance the Mamba model's understanding of images. However, the effectiveness of these scanning strategies remains uncertain. In this research, we conduct a comprehensive experimental investigation on the impact of mainstream scanning directions and their combinations on semantic segmentation of remotely sensed images. Through extensive experiments on the LoveDA, ISPRS Potsdam, ISPRS Vaihingen, and UAVid datasets, we demonstrate that no single scanning strategy outperforms others, regardless of their complexity or the number of scanning directions involved. A simple, single scanning direction is deemed sufficient for semantic segmentation of high-resolution remotely sensed images. Relevant directions for future research are also recommended.

Over the last decade, a 3D reconstruction technique has been developed to present the latest as-is information for various objects and build the city information models. Meanwhile, deep learning based approaches are employed to add semantic information to the models. Studies have proved that the accuracy of the model could be improved by combining multiple data channels (e.g., XYZ, Intensity, D, and RGB). Nevertheless, the redundant data channels in large-scale datasets may cause high computation cost and time during data processing. Few researchers have addressed the question of which combination of channels is optimal in terms of overall accuracy (OA) and mean intersection over union (mIoU). Therefore, a framework is proposed to explore an efficient data fusion approach for semantic segmentation by selecting an optimal combination of data channels. In the framework, a total of 13 channel combinations are investigated to pre-process data and the encoder-to-decoder structure is utilized for network permutations. A case study is carried out to investigate the efficiency of the proposed approach by adopting a city-level benchmark dataset and applying nine networks. It is found that the combination of IRGB channels provide the best OA performance, while IRGBD channels provide the best mIoU performance.

Semantic segmentation of Very High Resolution (VHR) remote sensing images is a fundamental task for many applications. However, large variations in the scales of objects in those VHR images pose a challenge for performing accurate semantic segmentation. Existing semantic segmentation networks are able to analyse an input image at up to four resizing scales, but this may be insufficient given the diversity of object scales. Therefore, Multi Scale (MS) test-time data augmentation is often used in practice to obtain more accurate segmentation results, which makes equal use of the segmentation results obtained at the different resizing scales. However, it was found in this study that different classes of objects had their preferred resizing scale for more accurate semantic segmentation. Based on this behaviour, a Stacking-Based Semantic Segmentation (SBSS) framework is proposed to improve the segmentation results by learning this behaviour, which contains a learnable Error Correction Module (ECM) for segmentation result fusion and an Error Correction Scheme (ECS) for computational complexity control. Two ECS, i.e., ECS-MS and ECS-SS, are proposed and investigated in this study. The Floating-point operations (Flops) required for ECS-MS and ECS-SS are similar to the commonly used MS test and the Single-Scale (SS) test, respectively. Extensive experiments on four datasets (i.e., Cityscapes, UAVid, LoveDA and Potsdam) show that SBSS is an effective and flexible framework. It achieved higher accuracy than MS when using ECS-MS, and similar accuracy as SS with a quarter of the memory footprint when using ECS-SS.

Point cloud data acquired using terrestrial laser scanning (TLS) often need to be semantically segmented to support many applications. To this end, various point-, voxel-, and image-based methods have been developed. For large-scale point cloud data, the former two types of methods often require extensive computational effort. In contrast, image-based methods are favorable from the perspective of computational efficiency. However, existing image-based methods are highly dependent on RGB information and do not provide an effective means of representing and utilizing the local geometric characteristics of point cloud data in images. This not only limits the overall segmentation accuracy but also prohibits their application to situations where the RGB information is absent. To overcome such issues, this research proposes a novel image enhancement method to reveal the local geometric characteristics in images derived by the projection of the point cloud coordinates. Based on this method, various feature channel combinations were investigated experimentally. It was found that the new combination $IZ_{e}D_{e}$ (i.e., intensity, enhanced $Z$ -coordinate, and enhanced range images) outperformed the conventional $I$ RGB and $I$ RGB $D$ channel combinations. As such, the approach can be used to replace the RGB channels for semantic segmentation. Using this new combination and the pretrained HR-EHNet considered, a mean intersection over union (mIoU) of 74.2% and an overall accuracy (OA) of 92.1% were achieved on the Semantic3D benchmark, which sets a new state of the art (SOTA) for the semantic segmentation accuracy of image-based methods.

Recent years have witnessed an increasing use of 3D models in general and 3D geometric models specifically of built environment for various applications, owing to the advancement of mapping techniques for accurate 3D information. Depending on the application scenarios, there exist various types of approaches to automate the construction of 3D building geometry. However, in those studies, less attention has been paid to watertight geometries derived from point cloud data, which are of use to the management and the simulations of building energy. To this end, an efficient reconstruction approach was introduced in this study and involves the following key steps. The point cloud data are first voxelised for the ray-casting analysis to obtain the 3D indoor space. By projecting it onto a horizontal plane, an image representing the indoor area is obtained and is used for the room segmentation. The 2D boundary of each room candidate is extracted using new grammar rules and is extruded using the room height to generate 3D models of individual room candidates. The room connection analyses are applied to the individual models obtained to determine the locations of doors and the topological relations between adjacent room candidates for forming an integrated and watertight geometric model. The approach proposed was tested using the point cloud data representing six building sites of distinct spatial confirmations of rooms, corridors and openings. The experimental results showed that accurate watertight building geometries were successfully created. The average differences between the point cloud data and the geometric models obtained were found to range from 12 to 21 mm. The maximum computation time taken was less than 5 min for the point cloud of approximately 469 million data points, more efficient than the typical reconstruction methods in the literature.

Pansharpening is a process of fusing a high spatial resolution panchromatic image and a low spatial resolution multispectral image to create a high-resolution multispectral image. A novel single-branch, single-scale lightweight convolutional neural network, named SDRCNN, is developed in this study. By using a novel dense residual connected structure and convolution block, SDRCNN achieved a better trade-off between accuracy and efficiency. The performance of SDRCNN was tested using four datasets from the WorldView-3, WorldView-2 and QuickBird satellites. The compared methods include eight traditional methods (i.e., GS, GSA, PRACS, BDSD, SFIM, GLP-CBD, CDIF and LRTCFPan) and five lightweight deep learning methods (i.e., PNN, PanNet, BayesianNet, DMDNet and FusionNet). Based on a visual inspection of the pansharpened images created and the associated absolute residual maps, SDRCNN exhibited least spatial detail blurring and spectral distortion, amongst all the methods considered. The values of the quantitative evaluation metrics were closest to their ideal values when SDRCNN was used. The processing time of SDRCNN was also the shortest among all methods tested. Finally, the effectiveness of each component in the SDRCNN was demonstrated in ablation experiments. All of these confirmed the superiority of SDRCNN.