Airborne light detection and ranging (LiDAR) technology has been widely utilized for collecting three-dimensional (3D) point cloud data on forest scenes, enabling the generation of high-accuracy digital elevation models (DEMs) for the efficient investigation and management of forest resources. Point cloud filtering serves as the crucial initial step in DEM generation, directly influencing the accuracy of the resulting DEM. However, forest filtering presents challenges in dealing with sparse point clouds and selecting appropriate initial ground points. The introduction of full-waveform LiDAR data offers a potential solution to the problem of sparse point clouds. Additionally, advancements in multi-source data integration and machine learning algorithms have created new avenues that can address the issue of initial ground point selection. To tackle these challenges, this paper proposes a novel filtering method for forest scenes utilizing full-waveform LiDAR data and hyperspectral image data. The proposed method consists of two main steps. Firstly, we employ the improved dynamic graph convolutional neural network (IDGCNN) to extract initial ground points. In this step, we utilize three types of low-correlation features: LiDAR features, waveform features, and spectral features. To enhance its accuracy and adaptability, a self-attention module was incorporated into the DGCNN algorithm. Comparative experiments were conducted to evaluate the effectiveness of the algorithm, demonstrating that the IDGCNN algorithm achieves the highest classification accuracy with an overall accuracy (OA) value of 99.38% and a kappa coefficient of 95.95%. The second-best performer was the RandLA-net algorithm, achieving an OA value of 98.73% and a kappa coefficient of 91.68%. The second step involves refining the initial ground points using the cloth simulation filter (CSF) algorithm. By employing the CSF algorithm, non-ground points present in the initial ground points are effectively filtered out. To validate the efficacy of the proposed filtering method, we generated a DEM with a resolution of 0.5 using the ground points extracted in the first step, the refined ground points obtained with the combination of the first and second steps, and the ground points obtained directly using the CSF algorithm. A comparative analysis with 23 reference control points revealed the effectiveness of our proposed method, as evidenced by the median error of 0.41 m, maximum error of 0.75 m, and average error of 0.33 m.

The International Association of Geodesy (IAG) aims to establish the International Height Reference System (IHRS) as one of its primary scientific objectives. Central to the realization of the IHRS is global vertical datum unification, which requires the connection of existing local vertical height reference systems (VHS) robustly and consistently. However, conventional methods are not suitable for estimating the offsets between two distant local height systems. In this paper, we propose a framework for connecting two local VHSs using ultraprecise clock frequency signal links between satellites and ground stations, referred to as the satellite frequency signal transmission (SFST) approach. The SFST approach allows for the direct determination of the geopotential and height differences between two ground datum stations without any location restrictions between the two VHSs. The simulation results show that the VHSs of China and the US can be unified with an accuracy of several centimeters, provided that the stability of atomic clocks used on-board the satellite and at on-ground datum locations reaches $4.8\times {10}^{-17}{\tau }^{-1/2}$ for an averaging time $\tau$ (in seconds). We conclude that the SFST approach shows promise for achieving centimeter-level accuracy in unifying the global vertical height datum and represents a new paradigm for the realization of the IHRS.

We introduce an approach for the direct measurement of the gravitational potential (GP) along the trajectory of a satellite, with a specific focus on Low-Earth Orbit (LEO) satellites. A LEO satellite communicates with several Geosynchronous Equatorial Orbit (GEO) satellites via frequency signal links. The GP difference can be measured in real-time using the gravitational frequency shift approach by equipping both LEO and GEO satellites with precise atomic clocks. Since the GP at the high orbits of the GEO satellites can be precisely determined by the present gravitational field model EGM2008, the GP along the LEO satellite’s trajectory can be determined. In this study, simulation experiments were conducted, featuring a GRACE-type satellite as the LEO satellite in communication with three equidistant GEO satellites. The results indicated that the accuracy of the GP measurements along the LEO satellite’s trajectory primarily depends on the precision of the onboard atomic clocks. Supposing optical atomic clocks attain an instability level of $1\times {10}^{-17}{\tau }^{-1/2}$ ($\tau$ in seconds), we determined the GP distribution covered by the LEO satellite’s trajectories with 30-day observations. Then, we determined a gravitational field at the centimeter level based on the GP distribution. The GP data derived from the trajectory of a LEO satellite can be utilized to establish temporal gravitational fields, which have broad applications in different disciplines.

Airborne LiDAR (Light Detection and Ranging) is an active Earth observing system, which can directly acquire high-accuracy and dense building roof data. Thus, airborne LiDAR has become one of the mainstream source data for building detection and reconstruction. The emphasis for building reconstruction focuses on the accurate extraction of feature lines. Building roof feature lines generally include the internal and external feature lines. Efficient extraction of these feature lines can provide reliable and accurate information for constructing three-dimensional building models. Most related algorithms adopt intersecting the extracted planes fitted by the corresponding points. However, in these methods, the accuracy of feature lines mostly depends on the results of plane extraction. With the development of airborne LiDAR hardware, the point density is enough for accurate extraction of roof feature lines. Thus, after acquiring the results of building detection, this paper proposed a feature lines extraction strategy based on the geometric characteristics of the original airborne LiDAR data, tracking roof outlines, normal ridge lines, oblique ridge lines and valley lines successively. The final refined feature lines can be obtained by normalization. The experimental results showed that our methods can achieve several promising and reliable results with an accuracy of 0.291 m in the X direction, 0.295 m in the Y direction and 0.091 m in the H direction for outlines extraction. Further, the internal feature lines can be extracted with reliable visual effects using our method.

The inter-system-like bias between the regional (BDS-2) and global (BDS-3) constellation of the BeiDou Navigation Satellite System (BDS) has been identified on common B1I pseudo-range observations. In this study, its characteristics are investigated with tracking data from the International GNSS Service (IGS) and International GNSS Monitoring and Assessment System (iGMAS) network. Firstly, the satellite-specific inter-system-like bias is calculated and the dependency on satellite is observed. Clearly noticeable discrepancies on BDS-2 and BDS-3 can be identified. Hence, the constellation-specific inter-system-like bias is estimated. Biases for all receivers are quite stable, with standard derivation (STDev) less than 0.2 m in average. The bias shows clear dependence on the receiver, while the firmware and antenna have limited but not negligible impacts, particularly for Trimble NetR9 and Alloy receivers. The Trimble NetR9 with TRM59800.00 antenna shows noticeable discrepancy up to about 1.5 m with different antenna, and the bias of the Trimble Alloy 5.37 jumps about 2.4 m with respect to later firmware. In addition, clear annual variations are observed for stations ABPO and MIZU with Septentrio POLARX5 5.3.2 and ASTERX4 4.4.2 receivers, respectively. Furthermore, the impacts of the biases on the BDS orbit and clock solutions are analyzed. Once BDS-2 and BDS-3 are treated as two independent systems, the root mean square (RMS) of code and carrier phase residuals can be reduced by around 9.3 cm and 0.23 mm, respectively, while the three-dimensional orbit consistency is improved by 6.8%, mainly in the tracking direction. Satellite laser ranging (SLR) shows marginal impacts on IGSO and MEO satellites. However, the SLR residual of C01 shifts −13.2 cm, resulting in a smaller RMS value. In addition, the RMS of linear clock fitting is reduced from 0.050 ns to 0.038 ns for BDS-3 MEO satellites in average.

The existing deep learning-based Airborne Laser Scanning (ALS) point cloud semantic segmentation methods require a large amount of labeled data for training, which is not always feasible in practice. Insufficient training data may lead to over-fitting. To address this issue, we propose a novel Multi-branch Feature Extractor (MFE) and a three-stage transfer learning strategy that conducts pre-training on multi-source ALS data and transfers the model to another dataset with few samples, thereby improving the model’s generalization ability and reducing the need for manual annotation. The proposed MFE is based on a novel multi-branch architecture integrating Neighborhood Embedding Block (NEB) and Point Transformer Block (PTB); it aims to extract heterogeneous features (e.g., geometric features, reflectance features, and internal structural features) by leveraging the parameters contained in ALS point clouds. To address model transfer, a three-stage strategy was developed: (1) A pre-training subtask was employed to pre-train the proposed MFE if the source domain consisted of multi-source ALS data, overcoming parameter differences. (2) A domain adaptation subtask was employed to align cross-domain feature distributions between source and target domains. (3) An incremental learning subtask was proposed for continuous learning of novel categories in the target domain, avoiding catastrophic forgetting. Experiments conducted on the source domain consisted of DALES and Dublin datasets and the target domain consists of ISPRS benchmark dataset. The experimental results show that the proposed method achieved the highest OA of 85.5% and an average F1 score of 74.0% using only 10% training samples, which means the proposed framework can reduce manual annotation by 90% while keeping competitive classification accuracy.