The photovoltaic (PV) industry boom has accelerated the need for accurately understanding the spatial distribution of PV energy systems. The synergy of remote sensing and artificial intelligence presents significant prospects for PV energy monitoring. Currently, numerous studies have focused on extracting rooftop PV systems from airborne or satellite imagery, but their small-scale and size-varying characteristics make the segmentation results suffer from PV internal incompleteness and small PV omission. To address these issues, this study proposed a size-aware deep learning network called Rooftop PV Segmenter (RPS) for segmenting small-scale rooftop PV systems from high-resolution imagery. In detail, the RPS network introduced a Semantic Refinement Module (SRM) to sense size variations of PV panels and reconstruct high-resolution deep semantic features. Moreover, a Feature Aggregation Module (FAM) enhanced the representation of robust features by continuously aggregating deeper features into shallower ones. In the output stage, a Deep Supervised Fusion Module (DSFM) was employed to constrain and fuse the outputs at different scales to achieve more refined segmentation. The proposed RPS network was tested and shown to outperform other models in producing segmentation results closer to the ground truth, with the F1 score and IoU reaching 0.9186 and 0.8495 on the publicly available California Distributed Solar PV Array Dataset (C-DSPV Dataset), and 0.9608 and 0.9246 on the self-annotated Heilbronn Rooftop PV System Dataset (H-RPVS Dataset). This study has provided an effective solution for obtaining a refined small-scale energy distribution database.

Geospatial code generation is emerging as a key direction in the integration of artificial intelligence and geoscientific analysis. However, there remains a lack of standardized tools for automatic evaluation in this domain. To address this gap, we propose AutoGEEval, the first multimodal, unit-level automated evaluation framework for geospatial code generation tasks on the Google Earth Engine (GEE) platform powered by large language models (LLMs). Built upon the GEE Python API, AutoGEEval establishes a benchmark suite (AutoGEEval-Bench) comprising 1325 test cases that span 26 GEE data types. The framework integrates both question generation and answer verification components to enable an end-to-end automated evaluation pipeline—from function invocation to execution validation. AutoGEEval supports multidimensional quantitative analysis of model outputs in terms of accuracy, resource consumption, execution efficiency, and error types. We evaluate 18 state-of-the-art LLMs—including general-purpose, reasoning-augmented, code-centric, and geoscience-specialized models—revealing their performance characteristics and potential optimization pathways in GEE code generation. This work provides a unified protocol and foundational resource for the development and assessment of geospatial code generation models, advancing the frontier of automated natural language to domain-specific code translation.

It is crucial to simulate land use change and assess the corresponding impact on ecosystem services to develop informed land management policies and conservation strategies. To comprehensively simulate the patterns of land use change under different management policies and evaluate the corresponding ecological service values (ESV), a method for coupling the Generalized Multi-Objective Programming (GMOP) model and Patch-generating Land Use Simulation (PLUS) model is proposed in this study. First, the GMOP model is used to obtain optimized land use solutions under different scenarios. Then, the PLUS model is used to analyze the mechanism driving land expansion, explore land conversion patterns, and, ultimately, achieve spatial expression of land use quantity changes. The uncertain parameters in the coupled model are processed by intuitionistic fuzzy numbers. The coupled model successfully integrates the outstanding spatiotemporal dynamic simulation capability of the PLUS model and the multiobjective optimization advantages of the GMOP model, effectively overcoming the limitations of applying a single model in land use analysis. Finally, four different scenarios are established for land use change, namely, business as usual (BAU), economic efficiency priority (RED), ecological protection priority (ELP), and coordinated economic and ecological development (EEB), to predict land use change trends and ecological service values. A case study of the Ningxia Hui Autonomous Region demonstrates that the area of agricultural land exhibits a stable growth trend in the four different scenarios, with the majority of the expansion occurring through the conversion of grassland. Concurrently, the rate of expansion of construction land is highest in the BAU scenario at 31.72%, compared with the area in 2020. This is notably higher than the rates observed in the RED (10.10%) and EEB (9.47%) cases. With the expansion of construction land, the ESV decreased by 3.485 billion, 1.514 billion, and 1.658 billion yuan in the BAU, RED, and ELP scenarios, representing 41.72%, 24.96%, and 34.05% decreases in ESV, respectively. The proposed integrated methodology accounts for various spatial constraints and land conversion behaviors, thereby ensuring a true and accurate reflection of land use dynamics. This methodology supports the quantification of ESV under different land management strategies, thereby providing policymakers with effective support for data-driven sustainable land use planning and conservation.

The urgency of global climate change has driven the rapid expansion of photovoltaic (PV) energy systems. However, accurately identifying PV panels remains a major challenge due to complex environmental backgrounds, spectral confusion, and the lack of high-quality annotated datasets. These factors significantly impact the generalization ability of deep learning models in large-scale high-resolution remote sensing applications, thereby limiting the effective monitoring and planning of PV power stations. To address these challenges, this article proposes a highly adaptable PV panel segmentation network, detailed PV monitoring (DPVM), specifically designed to enhance PV panel recognition in high-resolution imagery. DPVM integrates an adaptive context-aware module (ACAM) and a deep feature reconstruction decoder (DFRD). ACAM improves segmentation accuracy by leveraging multiscale feature fusion and spatial attention mechanisms. DFRD employs multistage decoding and feature synthesis to achieve high-quality image reconstruction. We trained DPVM on our self-constructed Northwest China PV dataset to ensure comprehensive learning of PV panel characteristics. Subsequently, we conducted generalization tests on other publicly available datasets, including AIR-PV and PVP. Experimental results demonstrate that DPVM exhibits outstanding robustness and broad adaptability, ensuring stable performance across diverse scenarios. Specifically, DPVM excels in complex backgrounds, significantly reducing PV panel missed detections, improving edge delineation, and outperforming classical and state-of-the-art segmentation models in key metrics.

Monocular 3D Human Pose Estimation (3D HPE) typically adopts a two-stage approach: estimating 2D joint positions from images and then lifting them to 3D coordinates, effectively reducing dataset bias inherent in direct methods. However, current lifting techniques face two key challenges: many Transformer-based methods rely on attention-based or staged spatial–temporal modeling, which can limit efficient long-range frame-joint reasoning, while diffusion models support probabilistic modeling of pose uncertainty but remain sensitive to joint-coordinate noise. We propose DiMMPose, a diffusion-based framework enhanced by Mamba’s state-space model for robust and efficient 3D pose estimation. Its denoising process consists of two coordinated modules. The Spatiotemporal Mamba Block (STMB) serves as the core feature extraction module, employing internal Pose Mamba components with bidirectional state propagation and linear complexity to efficiently model long-range frame-joint dependencies. STMB further refines these features through Spatiotemporal Scan and Merge, which traverses the same skeleton tokens in complementary frame-joint orders and fuses the resulting representations. The Multi-Prompt Mamba Denoiser (MPMD) combines structured prompts encoded by LongCLIP with learnable prompt representations to provide anatomical and motion-related guidance during denoising. DiMMPose achieves an average MPJPE of 28.9 mm on Human3.6M under the DET setting, with action-specific errors of 21.2 mm for Walking and 22.0 mm for WalkTogether. It improves over FinePOSE by 3.0 mm, reduces inference latency by 57.1%, and achieves 23.0 mm MPJPE on MPI-INF-3DHP (N = 243).

Pixel-level slum mapping remains constrained by limited cross-city generalisation, the lack of continuous estimates of intra-pixel slum coverage, and weak comparability across geographic contexts. AlphaEarth Foundations (AEF), which provides globally consistent 64-dimensional annual surface embeddings at 10 m resolution, offers an analysis-ready representation for scalable and lightweight slum monitoring. However, its applicability to slum mapping remains unclear because slum conditions are only indirectly observable from Earth observation data and are jointly shaped by built-form and socioeconomic processes. We systematically evaluate AEF for binary slum classification and sub-pixel slum cover fraction estimation across 12 cities and 69 city–year pairs from 2017 to 2024, using high-resolution GRAM-derived pseudo-masks as supervisory labels. The evaluation spans four training strategies, two protocols (random split and 3 × 3 spatial block cross-validation), six auxiliary feature configurations, and five baseline models, complemented by representation-level analyses (PCA, SHAP) and full-AOI mapping. Five findings emerge. (1) Same-city cross-year training is optimal under both protocols (median spatial F1 = 0.616, R² = 0.466); temporal expansion outperforms cross-city transfer, indicating city-scale representational drift. (2) Regression R² is driven primarily by zero/non-zero boundary discrimination: positive-pixel R² is consistently negative across all cities, revealing limited capacity to model intra-pixel variation in sub-pixel slum cover fraction at 10 m. (3) PC36 is consistently top-ranked across tasks; classification saturates at k = 32 while regression remains unsaturated at k = 64. (4) POI features yield the largest density gain (ΔR² = +0.064). (5) For six cities meeting dual-task usability thresholds, full-AOI inference across 2017–2024 preserves slum cluster structure (mean SSIM = 0.926). The study delineates the capabilities and complementarity needs of foundation-model embeddings for slum monitoring.