The fisheye camera, with its large viewing angle, can acquire more spatial information in one shot and is widely used in many fields. However, a fisheye image contains large distortion, resulting in that many scholars have investigated its accuracy of orthorectification, i.e., generation of digital orthophoto map (DOM). This paper presents an orthorectification method, which first determines the transformation relationship between the fisheye image points and the perspective projection points according to the equidistant projection model, i.e., determines the spherical distortion of the fisheye image; then introduces the transformation relationship and the fisheye camera distortion model into the collinearity equation to derive the fisheye image orthorectification model. To verify the proposed method, high accuracy of the fisheye camera 3D calibration field is established to obtain the interior and exterior orientation parameters (IOPs/EOPs) and distortion parameters of the fisheye lens. Three experiments are used to verify the proposed orthorectification method. The root mean square errors (RMSEs) of the three DOMs are averagely 0.003 m, 0.29 m, and 0.61 m, respectively. The experimental results demonstrate that the proposed method is correct and effective.

The digital elevation model (DEM) acquired through photogrammetry or LiDAR usually exposes voids due to phenomena such as instrumentation artifact, ground occlusion, etc. For this reason, this paper proposes a multiattention generative adversarial network model to fill the voids. In this model, a multiscale feature fusion generation network is proposed to initially fill the voids, and then a multiattention filling network is proposed to recover the detailed features of the terrain surrounding the void area, and the channel-spatial cropping attention mechanism module is proposed as an enhancement of the network. Spectral normalization is added to each convolution layer in the discriminator network. Finally, the training of the model by a combined loss function, including reconstruction loss and adversarial loss, is optimized. Three groups of experiments with four different types of terrains, hillsides, valleys, ridges and hills, are conducted for validation of the proposed model. The experimental results show that (1) the structural similarity surrounding terrestrial voids in the three types of terrains (i.e., hillside, valley, and ridge) can reach 80–90%, which implies that the DEM accuracy can be improved by at least 10% relative to the traditional interpolation methods (i.e., Kriging, IDW, and Spline), and can reach 57.4%, while other deep learning models (i.e., CE, GL and CR) only reach 43.2%, 17.1% and 11.4% in the hilly areas, respectively. Therefore, it can be concluded that the structural similarity surrounding the terrestrial voids filled using the model proposed in this paper can reach 60–90% upon the types of terrain, such as hillside, valley, ridge, and hill.

Traditional bathymetry LiDAR (light detection and ranging) onboard manned and/or unmanned airborne systems cannot operate in the context of narrow rivers in urban areas with high buildings and in mountainous areas with high peaks. Therefore, this study presents a prototype of a lightweight bathymetry LiDAR onboard an unmanned shipborne vehicle (called “GQ-Cor 19”). The GQ-Cor 19 system primarily includes an emitting optical module, a receiving optical module, control module, detection module, high-speed A/D sampling module, and data processing system. Considering that the “GQ-Cor 19” is extremely close to the water surface, various new technical challenges are encountered, such as significant laser scattering energy from the surface of the water, which saturates signals received by the photomultiplier tube detector. Therefore, this study presents various new technical solutions, including (1) an improved Bresenham algorithm, (2) a small and lightweight receiving optical system with a split-field method, and (3) a data acquisition module with a high-speech A/D collector. Following a series of different experimental verifications, the results demonstrate that the new generation of single-band LiDAR onboard an unmanned shipborne vehicle can swiftly measure the underwater depth, and the maximum measurement depth is more than 25 m. The measurement accuracy is better than 30 cm and the weight is less than 12 kg.

Wildfires pose serious threats to ecosystems, human life, and climate stability, underscoring the urgent need for accurate monitoring. Traditional approaches based on either optical or thermal imagery often fail under challenging conditions such as lighting interference, varying data sources, or small-scale flames, as they do not account for the hierarchical nature of feature representations. To overcome these limitations, we propose a multimodal deep learning framework that integrates visible (RGB) and thermal infrared (TIR) imagery for accurate wildfire segmentation. The framework incorporates edge-guided supervision and multilevel fusion to capture fine fire boundaries while exploiting complementary information from both modalities. To assess its effectiveness, we constructed a multi-scale flame segmentation dataset and validated the method across diverse conditions, including different data sources, lighting environments, and five flame size categories ranging from small to large. Experimental results show that BFCNet achieves an IoU of 88.25% and an F1 score of 93.76%, outperforming both single-modality and existing multimodal approaches across all evaluation tasks. These results demonstrate the potential of multimodal deep learning to enhance wildfire monitoring, offering practical value for disaster management, ecological protection, and the deployment of autonomous aerial surveillance systems.

Transverse aeolian ridges (TARs) are the most widely distributed and enigmatic aeolian landforms on the surface of Mars, holding significant research value and implications for interpreting ancient wind fields and environments, searching for water and life, and selecting landing sites. However, accurately interpreting the morphological parameters of TARs, including their edge contours and ridge lines, remains a challenge. To tackle this issue, this paper proposes a Multi-branch Auxiliary Training Encoder-Decoder Network (MATED-Net) for detecting the edge contours and ridge lines of TARs on Mars. Built upon the Unet architecture, MATED-Net incorporates four auxiliary training losses to perceive features at different scales. Then, We introduce a lightweight attention mechanism to guide the fusion of multi-scale features. Finally, an edge tracing loss is introduced to enhance the distinction between edge pixels and surrounding confusing pixels, thereby accurately tracking the true positions of edges. To verify the effectiveness of the MATED-Net in detecting TARs’ contours and ridge lines, this paper constructs a dataset of TAR ridge lines based on HiRISE and HiRIC imagery. To facilitate subsequent training and testing, all images were clipped to a size of 512 × 512and converted to the VOC dataset format, resulting in a total of 1000 images and corresponding label data. The experimental results demonstrate a precision of 0.72, a recall of 0.67, and a mean Intersection over Union (MIoU) of 0.57 for edge extraction.

Building extraction from airborne LiDAR data is significant for the accurate modeling of urban structures and constitutes a pivotal step in urban 3D reconstruction. However, in complex scenarios with noise, uneven point density, and the coexistence of buildings with sparse vegetation or shrubs, existing geometric features (e.g., curvature) struggle to distinguish planar from non-planar regions, leading to extraction errors. To overcome this limitation, we propose the Adaptive Tensor-Weighted Planar Descriptor (ATPD), a novel planarity feature that employs second-order tensors and reconstructs the covariance matrix through a data-driven adaptive weighting scheme. This design significantly improves the separability of planar and non-planar regions under challenging conditions. Comparative experiments further confirm that ATPD consistently outperforms traditional geometric features such as roughness, omnivariance, planarity, and curvature. In addition, planar features alone are insufficient to capture regions with local variations (e.g., roof ridges). To address this, we integrate planar and mesh-based features and develop a multi-feature graph-cut algorithm for robust building extraction from large-scale point clouds. The proposed method is validated on three benchmark LiDAR datasets of Vaihingen provided by ISPRS, where it achieves the best performance with a quality of 90.3% and an F1-score of 94.9%, outperforming traditional methods. Compared with deep learning approaches, our method also yields higher correctness. Further experiments on large-scale datasets from Vaihingen, Germany, and Palmerston North, New Zealand confirm its robustness, successfully extracting all buildings larger than 50 m².