Form Of Constraints Research Articles

Research on refined UAV inspection method of wind/solar power stations based on YOLOv8

More and more research is focusing on the unmanned aerial vehicle (UAV) inspection of onshore wind/solar power stations; however, how to balance the contradiction between detection accuracy and efficiency is still a challenge for domestic and international researches. In this paper, for the sparse distribution of inspection targets in onshore wind/solar power stations, the YOLOv8 model is used to quickly extract information such as photovoltaic module contours and key points of wind turbine blades in the wide-angle image of the UAV. By connecting these key points in series, fine route planning of the UAV is realized, and the zoom lens is used to carry out efficient and fine inspection of photovoltaic (PV) modules and wind turbine blades. In order to accelerate the convergence of the PV module contour extraction model, this paper introduces a loss function for the difference in linear angular orientation, which improves the average IOU accuracy of the YOLOv8 model to 93%. In the extraction of key points such as the center and tip of wind turbine blades, this paper firstly adopts serpentine convolution to replace the traditional convolution operator in order to adapt the wind turbine blade features, and secondly incorporates the a priori information of the angle constraints between the blades into the loss function. Finally, multi-angle wind turbine imaging photos in desert, hill, rice field, and other scenes are quickly simulated by computer simulation software to improve the model generalization performance. The experimental results show that the improved model achieves 85.4% on the mAP50 metric, which is a 9.2% improvement over YOLOv8, and 6.2% on the more stringent mAP50:95 metric.

Dimension synthesis of linkage mechanism with approximate poses and velocity constraints

This paper mainly addresses the problem of dimension synthesis for planar open chain two-rod (2R) mechanisms with velocity constraints and position-pose information in agricultural transplanting mechanism (TM). The method proposed in this study extends the dimension synthesis problem of planar 2R mechanisms, which takes position-pose into account, to the dimension synthesis considering both position-pose information and velocity information (PPIAVA) at certain pose points in the field of TM. The geometric constraints of the open chain 2R mechanism are extracted from the prerequisite conditions obtained by a series of given position-pose points and velocity constraints at some pose points to solve the dimension synthesis problem of the planar 2R mechanism. First, the Taylor formula is used to connect the characteristic of the derivative value at a certain point as coefficient with PPIAVA at a certain point. The mathematical model of position-pose and velocity is established by combining the relationship of rod length constraint and velocity direction characteristic. Secondly, the nonlinear equations established according to the position-pose information and velocity constraints are combined with the Lagrange equation and the extremum of the multivariate function to transform the appropriate objective function and then solved the equations. Finally, a transplanting mechanism is designed and tested by combining the solution results with the trajectory characteristics of the TM. The results show that this method is an effective synthesis method for smoothly solving the dimension synthesis problem of planar open chain 2R linkage mechanism with velocity constraints, which can extract geometric constraints from a given series of position-pose and velocity constraints at some poses.

Invariant Representation Learning in Multimedia Recommendation with Modality Alignment and Model Fusion.

Multimedia recommendation systems aim to accurately predict user preferences from multimodal data. However, existing methods may learn a recommendation model from spurious features, i.e., appearing to be related to an outcome but actually having no causal relationship with the outcome, leading to poor generalization ability. While previous approaches have adopted invariant learning to address this issue, they simply concatenate multimodal data without proper alignment, resulting in information loss or redundancy. To overcome these challenges, we propose a framework called M3-InvRL, designed to enhance recommendation system performance through common and modality-specific representation learning, invariant learning, and model merging. Specifically, our approach begins by learning modality-specific representations along with a common representation for each modality. To achieve this, we introduce a novel contrastive loss that aligns representations and imposes mutual information constraints to extract modality-specific features, thereby preventing generalization issues within the same representation space. Next, we generate invariant masks based on the identification of heterogeneous environments to learn invariant representations. Finally, we integrate both invariant-specific and shared invariant representations for each modality to train models and fuse them in the output space, reducing uncertainty and enhancing generalization performance. Experiments on real-world datasets demonstrate the effectiveness of our approach.

The Gaia-ESO survey: New spectroscopic binaries in the Milky Way

Context. The Gaia-ESO survey (GES) is a large public spectroscopic survey that acquired spectra for more than 100 000 stars across all major components of the Milky Way. In addition to atmospheric parameters and stellar abundances that have been derived in previous papers of this series, the GES spectra allow us to detect spectroscopic binaries with one (SB1), two (SB2), or more (SBn ≥ 3) components. Aims. The present paper discusses the statistics of GES SBn ≥ 2 after analysing 160 727 GIRAFFE HR10 and HR21 spectra, amounting to 37 565 unique Milky Way field targets. Methods. Cross-correlation functions (CCFs) have been re-computed thanks to a dozen spectral masks probing a range of effective temperatures (3900 K < Teff < 8000 K), surface gravities (1.0 < log g < 4.7), and metallicities (−2.6 < [Fe/H] < 0.3). By optimising the mask choice for a given spectrum, the newly computed, so-called NACRE (NArrow CRoss-correlation Experiment) CCFs are narrower and allow more stellar components to be unblended than standard masks. The DOE (Detection Of Extrema) extremum-finding code then selects the individual components and provides their radial velocities. Results. From the sample of HR10 and HR21 spectra corresponding to 37 565 objects, the present study leads to the detection of 322 SB2, ten SB3 (three of them being tentative), and two tentative SB4. In particular, compared to our previous study, the NACRE CCFs allowed us to multiply the number of SB2 candidates by ≈1.5. The colour-magnitude diagram reveals, as expected, the shifted location of the SB2 main sequence. A comparison between the SB identified in Gaia DR3 and the ones detected in the present work was performed and the complementarity of the two censuses is discussed. An application to the mass-ratio determination is presented, and the mass-ratio distribution of the GES SB2 is discussed. When accounting for the SB2 detection rate, an SB2 frequency of ≈1.4 % is derived within the present stellar sample of mainly FGK-type stars. Conclusions. As primary outliers identified within the GES data, SBn spectra produce a wealth of information and useful constraints for the binary population synthesis studies.

Grouping-based Cyclic Scheduling under Age of Correlated Information Constraints

Grouping-based Cyclic Scheduling under Age of Correlated Information Constraints

SOLVING VEHICLE ROUTING PROBLEMS IN THE FRAMEWORK OF THE CONSTRAINT PROGRAMMING PARADIGM

The article is devoted to an analytical survey of the possibilities of solving vehicle routing problems within the framework of the constraint programming paradigm. First, the article provides an overview of the various formulations of the vehicle routing problem. Then, to facilitate understanding, these statements are formulated in the language of integer linear programming. After that, a description of the constraint programming technology is given and typical global constraints are given that are useful in various formulations of the vehicle routing problem. It is concluded that flexibility in the formulation of constraints, the possibility of integration with other optimization methods, as well as the ability to take into account uncertainties make programming in constraints one of the most popular tools for solving vehicle routing problems.

Extraction of agricultural plastic greenhouses based on a U-Net convolutional neural network coupled with edge expansion and loss function improvement

ABSTRACT Agricultural plastic greenhouses (APGs) are crucial for sustainable agricultural planting, and accurate spatial distribution information acquisition is crucial. Deep learning network models can extract target features from remote sensing images more effectively than traditional interpretation methods, which face challenges like high workloads and poor repeatability. In this study, we aim to enhance the inventorying of Agricultural Plastic Greenhouses (APGs) by improving the extraction accuracy of their locations and numbers through remote sensing techniques. Utilizing GF-7 satellite imagery, we propose an enhanced U-Net convolutional neural network (CNN) model that incorporates edge information expansion and a joint loss function to optimize performance. The primary objective is to provide a rapid and accurate method for mapping APGs, which is crucial for effective agricultural management and environmental monitoring. The U-Net network’s accuracy was enhanced by 1.1% after expanding 3 × 3 sample edge information, and further by 1.9% by combining edge extension and loss function constraints. Our results demonstrate that the modified U-Net model significantly improves extraction accuracy compared to traditional methods, thereby facilitating better inventory management and planning for agricultural cash crops. This advancement not only supports farmers in optimizing resources but also contributes to sustainable agricultural practices by enabling precise monitoring of APG distribution. Implications: Compared to traditional interpretation methods, which suffer problems such as heavy workloads, small adaptation ranges and poor repeatability, deep learning network models can better extract target features from remote sensing images. In this study, we used GF-7 image data to improve the traditional U-Net convolutional neural network (CNN) model. The Canny operator and Gaussian kernel (GK) function were used for sample edge expansion, and the binary cross-entropy and GK functions were used to jointly constrain the loss. Finally, APGs were accurately extracted and compared to those obtained with the original model. The results indicated that the APG extraction accuracy of the U-Net network was improved through the expansion of sample edge information and adoption of joint loss function constraints.

How Digital Twins Impact Responsiveness: A Dynamic Fit Approach to Information Processing for High-Involvement Product Demand Management

In an era of rapidly evolving customer demands and technological advancements, organisations face mounting pressure to enhance their responsiveness. Digital Twins offer novel capabilities by providing comprehensive transparency on customer demands through dynamic and interactive links between real-world entities and their digital counterparts. As a result, Digital Twins have the potential to enable organisations to respond more swiftly and effectively to fluctuating customer demands across volume, process, and product dimensions. However, Digital Twins still lack empirical explorations and theoretical embedment, leading to conceptual ambiguity. Through a case study approach, the study integrates Digital Twins into the framework of Organisational Information Processing Theory (OIPT). Empirical evidence proposes that Digital Twins offer a spectrum of capabilities rather than a one-size-fits-all solution. The findings further emphasise the need for aligning Digital Twin sophistication levels with Information Processing Needs and Constraints, highlighting that the dynamic interplay between Digital Twin advancements, Information Processing Constraints, and Information Processing Capabilities creates a dynamic cycle of adaptation and optimisation. On the whole, the study presents four key implications: (1) Awareness about the versatile application areas of Digital Twins is created. (2) The positive impact of Digital Twins on volume, process, and product responsiveness is outlined. (3) The concept of Information Processing Constraints is introduced to OIPT. And (4) a dynamic cycle of aligning Digital Twin sophistication with Information Processing Capabilities and Constraints is suggested. Therefore, the study contributes not only to a deeper understanding of Digital Twins in demand management but also proposes a new dynamic fit perspective on OIPT.

Informative Trajectory Planning Using Reinforcement Learning for Minimum-Time Exploration of Spatiotemporal Fields.

This article studies the informative trajectory planning problem of an autonomous vehicle for field exploration. In contrast to existing works concerned with maximizing the amount of information about spatial fields, this work considers efficient exploration of spatiotemporal fields with unknown distributions and seeks minimum-time trajectories of the vehicle while respecting a cumulative information constraint. In this work, upon adopting the observability constant as an information measure for expressing the cumulative information constraint, the existence of a minimum-time trajectory is proven under mild conditions. Given the spatiotemporal nature, the problem is modeled as a Markov decision process (MDP), for which a reinforcement learning (RL) algorithm is proposed to learn a continuous planning policy. To accelerate the policy learning, we design a new reward function by leveraging field approximations, which is demonstrated to yield dense rewards. Simulations show that the learned policy can steer the vehicle to achieve an efficient exploration, and it outperforms the commonly-used coverage planning method in terms of exploration time for sufficient cumulative information.

A generalized maximum correntropy based constraint adaptive filtering: Constraint-forcing and performance analyses.

The quadratic cost functions, exemplified by mean-square-error, often exhibit limited robustness and flexibility when confronted with impulsive noise contamination. In contrast, the generalized maximum correntropy (GMC) criterion, serving as a robust nonlinear similarity measure, offers superior performance in such scenarios. In this paper, we develop a recursive constrained adaptive filtering algorithm named recursive generalized maximum correntropy with a forgetting factor (FF-RCGMC). This algorithm integrates the exponential weighted GMC criterion with a linear constraint framework based on least-squares. However, the lack of constraint information during the learning process may lead to divergence or malfunctioning of FF-RCGMC after a certain number of iterations because of round-off errors. To rectify this deficiency, we introduce a constraint-forcing strategy into FF-RCGMC, resulting in a more stable variant termed robust type constraint-forcing FF-RCGMC (CFFF-RCGMC). In the context of CFFF-RCGMC, we embark on a thorough examination of its computational burden, encompassing both mean and mean-square stability analyses, along with an in-depth exploration of its transient and steady-state filtering characteristics under a set of plausible assumptions. Our simulation-based evaluations, specifically tailored for system identification tasks within non-Gaussian noisy environments, unequivocally underscore the excellent performance of CFFF-RCGMC when against its relevant algorithmic counterparts.

A robust GNSS velocity estimation method combining Doppler and carrier phase observations in complex urban environments

Accurate and reliable velocity information is crucial for kinematic positioning, as it not only characterizes the carrier state but also serves as constraint information for positioning. However, due to the influence of low-cost Global Navigation Satellite System (GNSS) chips and complex environments, GNSS signals are prone to attenuation and interruptions. Consequently, frequent gross errors and cycle slips occur in observations, significantly degrading the velocity precision and reliability. To address this challenge, we have proposed a method combining Doppler and Time-Differenced Carrier Phase Velocity Estimation (D-TDCPVE). First, we assess the accuracy of both Doppler and carrier phase observations to determine the appropriate weight ratio for their combination. Second, we analyze the correlation between the two observation types and propose an adaptive parameter estimation strategy for estimating one or two clock bias variations to address their potential differences. Finally, we incorporate a combination of pre- and post-detection quality control measures along with additional cycle slip parameters to mitigate the impact of gross errors and cycle slips in complex environments. Experimental results conducted with three low-cost devices in urban environments demonstrate the superior performance of the D-TDCPVE method over both Doppler velocity estimation (DVE) and time-differenced carrier phase velocity estimation (TDCPVE). It enhances the success rate of velocity estimation by 3.3% to 20.1% compared to TDCPVE and improves velocity estimation accuracy by 16.1% to 60.9% compared to DVE. Moreover, the method does not necessitate the combination of dual-frequency observations, making it particularly valuable for cycle slip detection in scenarios involving mixed single- and dual-frequency observations, which is particularly useful in urban environments.

Efficient handling of radiality constraints for large-scaled power distribution networks

Power distribution networks are usually characterized by a radial topology and therefore the related optimization problems require radiality constraints in their formulations. However, practical instances may have very large size, and the number of radiality constraints may grow faster than the size of the instance. Hence, the arising optimization models may become computationally intractable due to their huge dimension. This work proposes a combinatorial optimization approach to overcome this issue. Our approach is based on the combinatorial formulation of the radiality constraints and their delayed and efficient generation in the model, following a separation-optimization scheme. We find an optimal acyclic spanning subgraph subject to both technical side constraints and topological radiality requirements on a graph representing the distribution network. This can be done without considering all the exponentially many radiality constrains from the beginning of the model solution, but introducing only the ones that are needed to make the solution feasible. It turns out that the number of required constraints is much smaller than the number of all possible radiality constraints, also because some of the electric constraints already favor the elimination of infeasible configurations. Computational experiments on reconfiguration benchmarks from the literature show the effectiveness of the proposed approach.

A novel bearing weak fault diagnosis method based on rank constrained low-rank and sparse decomposition

Abstract Addressing the challenge of diagnosing incipient bearing faults amidst significant noise, a novel diagnostic approach is introduced, leveraging a Rank Constrained Low-Rank and Sparse Decomposition (RCLRSD) model tailored for weak fault detection in bearings. Initially, we raised the Autocorrelation Function of the Square Envelope in Frequency Domain as an innovative method for the estimation of fault frequencies. Subsequently, we constructed a two-dimensional observation matrix, which is formulated independently of predefined assumptions. Then, we examine the configuration and distribution patterns of bearing fault signals, uncovering the low-rank nature of fault characteristics within a designated two-dimensional transform domain. Moreover, we found the noise signal to exhibit sparsity and an approximate Gaussian distribution. Based on this, a rank-constrained low-rank sparse decomposition model is established, and rank-constrained low-rank regular constraints for feature information and sparse regular constraints and Gaussian constraints for interference signals are constructed respectively. Ultimately, we employed the Gray Wolf Optimization algorithm to refine the model parameters, and we deduced the model’s solver via the Alternating Direction Method of Multipliers. The proposed RCLRSD model decomposes bearing fault data into three constituents: low-rank, sparse, and Gaussian components, effectively addressing the challenge of extracting weak fault signatures from bearings. The weak feature extraction capability of the RCLRSD model is verified using a multi-interference simulation model and experimental data of bearing failures under strong noise conditions; the generalization of the model is verified by the classification effect of the Support Vector Machine under two bearing failure datasets. Comparison with various algorithms confirms the superiority of the proposed algorithm.

Joint Sampling and Transmission Policies for Minimizing Cost Under Age of Information Constraints.

In this work, we consider the problem of jointly minimizing the average cost of sampling and transmitting status updates by users over a wireless channel subject to average Age of Information (AoI) constraints. Errors in the transmission may occur and a policy has to decide if the users sample a new packet or attempt to retransmission the packet sampled previously. The cost consists of both sampling and transmission costs. The sampling of a new packet after a failure imposes an additional cost on the system. We formulate a stochastic optimization problem with the average cost in the objective under average AoI constraints. To solve this problem, we propose three scheduling policies: (a) a dynamic policy, which is centralized and requires full knowledge of the state of the system and (b) two stationary randomized policies that require no knowledge of the state of the system. We utilize tools from Lyapunov optimization theory and Discrete-Time Markov Chain (DTMC) to provide the dynamic policy and the randomized ones, respectively. Simulation results show the importance of providing the option to transmit an old packet in order to minimize the total average cost.

A framework of multi-view machine learning for biological spectral unmixing of fluorophores with overlapping excitation and emission spectra.

The accuracy of assigning fluorophore identity and abundance, known as spectral unmixing, in biological fluorescence microscopy images remains a significant challenge due to the substantial overlap in emission spectra among fluorophores. In traditional laser scanning confocal spectral microscopy, fluorophore information is acquired by recording emission spectra with a single combination of discrete excitation wavelengths. However, organic fluorophores possess characteristic excitation spectra in addition to their unique emission spectral signatures. In this paper, we propose a generalized multi-view machine learning approach that leverages both excitation and emission spectra to significantly improve the accuracy in differentiating multiple highly overlapping fluorophores in a single image. By recording emission spectra of the same field with multiple combinations of excitation wavelengths, we obtain data representing different views of the underlying fluorophore distribution in the sample. We then propose a multi-view machine learning framework that allows for the flexible incorporation of noise information and abundance constraints, enabling the extraction of spectral signatures from reference images and efficient recovery of corresponding abundances in unknown mixed images. Numerical experiments on simulated image data demonstrate the method's efficacy in improving accuracy, allowing for the discrimination of 100 fluorophores with highly overlapping spectra. Furthermore, validation on images of mixtures of fluorescently labeled Escherichia coli highlights the power of the proposed multi-view strategy in discriminating fluorophores with spectral overlap in real biological images.

Improving Hydrological Simulations with a Dynamic Vegetation Parameter Framework

Many hydrological models incorporate vegetation-related parameters to describe hydrological processes more precisely. These parameters should adjust dynamically in response to seasonal changes in vegetation. However, due to limited information or methodological constraints, vegetation-related parameters in hydrological models are often treated as fixed values, which restricts model performance and hinders the accurate representation of hydrological responses to vegetation changes. To address this issue, a vegetation-related dynamic-parameter framework is applied on the Xinanjiang (XAJ) model, which is noted as Eco-XAJ. The dynamic-parameter framework establishes the regression between the Normalized Difference Vegetation Index (NDVI) and the evapotranspiration parameter K. Two routing methods are used in the models, i.e., the unit hydrograph (XAJ-UH and Eco-XAJ-UH) and the Linear Reservoir (XAJ-LR and Eco-XAJ-LR). The original XAJ model and the modified Eco-XAJ model are applied to the Ou River Basin, with detailed comparisons and analyses conducted under various scenarios. The results indicate that the Eco-XAJ model outperforms the original model in long-term discharge simulations, with the NSE increasing from 0.635 of XAJ-UH to 0.647 of Eco-XAJ-UH. The Eco-XAJ model also reduces overestimation and incorrect peak flow simulations during dry seasons, especially in the year 1991. In drought events, the modified model significantly enhances water balance performance. The Eco-XAJ-UH outperforms the XAJ-UH in 9 out of 16 drought events, while the Eco-XAJ-LR outperforms the XAJ-LR in 14 out of 16 drought events. The results demonstrate that the dynamic-parameter model, in regard to vegetation changes, offers more accurate simulations of hydrological processes across different scenarios, and its parameters have reasonable physical interpretations.

On the feasibility of solutions to the split delivery vehicle routing problem represented as edge variables

Mixed integer programming formulations for the Split Delivery Vehicle Routing Problem (SDVRP) typically use edge decision variables. It was believed that feasibility couldn't be verified in polynomial time. We show that this recognition problem depends on the formulation's constraints and prove that it's strongly NP-hard for a recent formulation. With subtour elimination constraints, we provide an O(nlog⁡n)-time recognition algorithm. This challenges assumptions about edge-based formulations and provides new insights into SDVRP solution verification complexity.

Realising the potential of information acquisition in promoting sustainable agriculture: A systematic review

A growing body of literature identifies information constraints as a potential barrier to farmers’ adoption of sustainable agricultural practices (SAPs) and information acquisition (IA) as the key to breaking the dilemma. Due to the heterogeneity of information, channels and farmers, IA has different effects on farmers’ sustainable behaviour in different SAP scenarios. This study systematically reviews the mechanisms, empirical cases, models and methods of the impact of IA on farmers’ adoption of SAPs. Results show that IA plays a crucial role in promoting SAPs, and exposure to different information sources/channels usually has differential effects on farmers’ intention or behaviour to adopt SAPs. It is worth noting that the positive effect of the use of modern information channels (especially the Internet and smartphone) on SAPs is gradually being demonstrated in empirical studies across countries, and the effect of Internet or smartphone use on farmers with different resource endowment characteristics may also be heterogeneous. According to further analysis, traditional and modern information channels should complement each other and be fully integrated into SAPs extensions. This study also discusses the gaps that need to be bridged in IA-to-SAPs and describes possible future research directions. These findings will contribute to the scientific design of information interventions and environmental policies in SAPs extension services.

Input Pose is Key to Performance of Free Energy Perturbation: Benchmarking with Monoacylglycerol Lipase.

Free energy perturbation (FEP) methodologies have become commonplace methods for modeling potency in hit-to-lead and lead optimization stages of drug discovery. The conformational states of the initial poses of compounds for FEP+ calculations are often set up by alignment to a cocrystal structure ligand, but it is not clear if this method provides the best result for all proteins or all ligands. Not only are ligand conformational states potential variables in modeling compound potency in FEP but also the selection of crystallographic water molecules for inclusion in the FEP input structures can impact FEP models. Here, we report the results of FEP calculations using FEP+ from Schrödinger and starting from maximum common substructure alignment and docked poses generated with an array of docking methodologies. As a benchmark data set, we use monoacylglycerol lipase (MAGL), an important clinical drug target in cancer malignancy, neurological diseases, and metabolic disorders, and a set of 17 MAGL inhibitors. We found a large variation among FEP+ correlations to experimental IC50 values depending on the method used to generate the input pose and that the inclusion of ligand-based information in the docking process, with some methods, increases the correlation between FEP+ free energies and IC50 values. Upon analysis of the initial poses, we found that the differences in FEP+ correlations stemmed from rotation around a tertiary amide bond as well as translation of the compound toward the more hydrophobic side of the MAGL pocket. FEP+ estimation improved across all pose modeling methods when hydrogen bond constraint information was added. However, simple maximum common substructure alignment in the presence of all crystallographic water molecules outperformed all other methods in correlation between estimated and experimental IC50 values. Taken together, these findings suggest that pose selection and crystallographic water inclusion greatly impact how well FEP+ estimated IC50 values align with experimental IC50 values and that modelers should benchmark a few different pose generation methodologies and different water inclusion strategies for their hit-to-lead and lead optimization drug discovery projects.

Canopy structure dynamics constraints and time sequence alignment for improving retrieval of rice leaf area index from multi-temporal Sentinel-1 imagery

Due to the limited availability of in-situ observation data, most existing leaf area index (LAI) inversion models do not fully leverage temporal information. Furthermore, the phenological evolution of crops can result in unstable and inaccurate retrieval outcomes. To address these challenges, this study proposes a novel framework for LAI inversion based on Sentinel-1. First, the constrained canopy structure dynamic hierarchical linear model (CSDHLM) is constructed, which integrates canopy dynamics information and temporal constraints. Second, the microwave scattering characteristics at various crop growth stages used to develop the phenological segment dynamic time warping (PSDTW). The PSDTW aims to address the challenges posed by inconsistent phenological dynamics across different plots. The quantitative evaluation results indicate that CSDHLM more accurately captures the temporal changes of LAI (R2 = 0.7688, RMSE = 0.8742) compared to hierarchical linear model (R2 = 0.7234, RMSE = 0.9561) and gaussian process regression (R2 = 0.7143, RMSE = 0.9717). Additionally, the LAI inversion results obtained by combining CSDHLM and PSDTW have greater robustness (R2 = 0.7332, RMSE = 1.4032) across diverse agricultural scenarios. This study emphasizes the importance of phenological information in estimating rice LAI, and the proposed framework is capable of generating long-term rice LAI maps with high resolution, demonstrating significant potential for agricultural applications at the regional scale.