In view of the defects of total station and laser tracker in the construction of three-dimensional control network, a construction method of three-dimensional control network is proposed by combining total station and laser tracker. Total stations and laser trackers are utilized separately to observe the control points according to the free-stationing method. In data processing, the weights of different types of observations of total station and laser tracker are reasonably determined based on the Helmert variance component estimation. The high-precision horizontal direction observations of total station, which are better than 0 . 5 ″ , and the ultra-high-precision ranging observations of laser tracker, which are better than 10 μm, are effectively combined in order to obtain more precise adjustment results. The experimental results show that the positional precision of the combined adjustment is better than that of the adjustment results of a single measurement system, and the precision of three-dimensional control network can be further improved by employing Helmert variance component estimation to determine the weight reasonably.

Visual foundation models exhibit significant potential for extracting generalized characteristics from imagery, which is crucial for remote sensing image interpretation tasks. However, current approaches underutilize these capabilities for hyperspectral image semantic segmentation. To bridge this gap, this study establishes SAM-derived generalized feature fusion with a hierarchical network for hyperspectral image semantic segmentation (SAM-GFNet), which integrates generalized features with hierarchical task-specific spatial–spectral characteristics for the segmentation task. Specifically, hyperspectral images are partitioned into multiple spectral groups, and SAM-induced generalized features are extracted within each group’s corresponding spectral range. To leverage multi-scale contextual information, the hierarchical hyperspectral image semantic segmentation architecture fuses multi-scale spectral–spatial features with SAM-induced generalized visual characteristics through a hierarchical fusion pattern. Furthermore, an efficient feature fusion module is further developed to adaptively integrate encoded features with decoder-generated characteristics, further enhancing semantic segmentation performance. The proposed semantic segmentation scheme is evaluated on WHU-OHS and WHU-H2SR benchmark datasets, achieving overall accuracies of 79.60% and 86.92% on two datasets, surpassing representative baseline methods by over 2.94% and 1.56%, respectively.

To address the difficulty in reasonably determining the datum in the combined adjustment of a two-order horizontal control network, a new weighted quasi-stable adjustment method is proposed that uses posteriori variance to determine the weight of the datum. First, the most suitable classic datum is selected from the high-order control points to fix the coordinate system, and the posterior variance matrix of the high-order control points is obtained by the classical free network adjustment. Then, keeping the coordinates of the high-order control points unchanged and taking them as the quasi-stable datum, the datum weight matrix is determined according to the posterior variance matrix, and weighted quasi-stable combined adjustment is performed to effectively use all the high-order control points to restrain the error accumulation in the low-order network. Finally, to achieve better conformity between the coordinates of the high-order control points and the low-order observations, the weighted quasi-stable combined adjustment is carried out iteratively until the posteriori variance matrix remains unchanged before and after the adjustment. The proposed method is applied to the combined adjustment of the route plane control network and track control network of high-speed railway as well as the ground and tunnel control networks of a particle accelerator. The adjustment result of this method is compared with the adjustment results of classical free network and annexed network. Experimental results show that the root-mean-square of positional error of the proposed method is close to the result of the annexed network adjustment and better than the result of the classical free network adjustment; the error accumulation is also effectively restrained. The relative positional error of this method is smaller than that in the annexed network, which shows that the proposed method can effectively reduce the influence of high-order control points with initial error on relative positional error.

Digital industrial photogrammetry is used for acceptance of large precision assemblies, but separating the time-invariant assembly form error from the epoch-varying measurement perturbation remains challenging. Registering each repeat independently to CAD with a six-degree-of-freedom (6-DoF) datum from only a few control points absorbs part of the form error, biasing the assembly RMS. This article proposes joint datum–form decomposition (JDFD), a generalised Procrustes analysis that shares one CAD-anchored common rigid across repeats and removes each epoch’s residual whole-body similarity. JDFD offers two advantages. First, its assembly RMS is independent of the sparse-control budget: an independent per-epoch 6-DoF overestimates the assembly RMS by ≈43% in Monte Carlo (${\sigma }_M=6.0$ μm) and by ≈79% on real data at $n_c=4$ (median ≈ 51% over random layouts), whereas JDFD returns the full-set reference $\mathrm{RMS}(\widehat{I})=11.92$ μm at every $n_c$ tested. Second, by removing residual similarity, JDFD yields an uncontaminated within-session perturbation $\mathrm{RMS}(\widehat{M})=3.36$ μm, compared with 5.57 μm without peeling and 3.91 μm against the full-set independent-datum baseline. The resulting assembly RMS exceeds the 10 μm specification, so acceptance should rest on the $n_c$-invariant $\mathrm{RMS}(\widehat{I})$.