Spatial Weight Research Articles

CoSTA: Co-training spatial–temporal attention for blind video quality assessment

Self-attention-based Transformer has achieved great success in many computer vision tasks. However, its application to blind video quality assessment (VQA) is far from comprehensive. Evaluating the quality of in-the-wild videos is challenging due to the unknown of pristine reference and shooting distortion. This paper presents a Co-trained Space–Time Attention network for the blind VQA problem, termed CoSTA. Specifically, we first build CoSTA by alternately concatenating the divided space–time attention. Then, to facilitate the training of CoSTA, we design a vectorized regression loss by encoding the mean opinion score (MOS) to the probability vector and embedding a special token as the learnable variable of MOS, leading to the better fitting of the human rating process. Finally, to solve the data-hungry problem within Transformer, we propose to co-train the spatial and temporal attention weights using both images and videos. Various experiments are conducted on the de-facto in-the-wild video datasets, including LIVE-Qualcomm, LIVE-VQC, KoNViD-1k, YouTube-UGC, LSVQ, LSVQ-1080p, and DVL2021. Experimental results demonstrate the superiority of the proposed CoSTA over the state-of-the-art. The source code is publicly available at https://github.com/GZHU-DVL/CoSTA.

Spatial Nonlinear Effects of Street Vitality Constrained by Construction Intensity and Functional Diversity—A Case Study from the Streets of Shenzhen

As an important part of urban vitality, street vitality is an external manifestation of street economic prosperity and is affected by the built environment and the surrounding street vitality. However, existing research on the formation mechanism of street vitality focuses only on the built environment itself, ignoring the spatial spillover effect on street vitality. This study uses 5290 street segments in Shenzhen as examples. Utilizing geospatial and other multisource big data, this study creates spatial weight matrices at varying distances based on different living circle ranges. By combining the panel threshold model (PTM) and the spatial panel Durbin model (SPDM), this study constructs a spatial autoregressive threshold model to explore the spatial nonlinear effects of street vitality, considering various spatial weight matrices and thresholds of construction intensity and functional diversity. Our results show the following: (1) Street vitality exhibits significant spatial spillover effects, which gradually weaken as the living circle range expands (Moran indices are 0.178***, 0.160***, and 0.145*** for the 500 m, 1000 m, and 1500 m spatial weight matrices, respectively). (2) Construction intensity has a threshold, which is 0.1466 under spatial matrices of different distances. Functional diversity has two thresholds: 0.6832 and 2.2065 for the 500 m spatial weight matrix, and 0.6832 and 1.4325 for the 1000 m matrices, and 0.6832 and 1.2724 for 1500 m matrices. (3) As an international metropolis, street accessibility in Shenzhen has a significant and strong positive impact on its street vitality. This conclusion provides stakeholders with spatial patterns that influence street vitality, offering a theoretical foundation to further break down barriers to street vitality.

Enhanced Spatio-Temporal Modeling for Rainfall Forecasting: A High-Resolution Grid Analysis

Rainfall serves as a lifeline for crop cultivation in many agriculture-dependent countries including India. Being spatio-temporal data, the forecasting of rainfall becomes a more complex and tedious process. Application of conventional time series models and machine learning techniques will not be a suitable choice as they may not adequately account for the complex spatial and temporal dependencies integrated within the data. This demands some data-driven techniques that can handle the intrinsic patterns such as non-linearity, non-stationarity, and non-normality. Space–Time Autoregressive Moving Average (STARMA) models were highly known for its ability to capture both spatial and temporal dependencies, offering a comprehensive framework for analyzing complex datasets. Spatial Weight Matrix (SWM) developed by the STARMA model helps in integrating the spatial effects of the neighboring sites. The study employed a novel dataset consisting of annual rainfall measurements spanning over 50 (1970–2019) years from 119 different locations (grid of 0.25 × 0.25 degree resolution) of West Bengal, a state of India. These extensive datasets were split into testing and training groups that enable the better understanding of the rainfall patterns at a granular level. The study findings demonstrated a notable improvement in forecasting accuracy by the STARMA model that can exhibit promising implications for agricultural management and planning, particularly in regions vulnerable to climate variability.

Prediction of Operational Noise Uncertainty in Automotive Micro-Motors Based on Multi-Branch Channel–Spatial Adaptive Weighting Strategy

The acoustic performance of automotive micro-motors directly impacts the comfort and driving experience of both drivers and passengers. However, various motor production and testing uncertainties can lead to noise fluctuations during operation. Thus, predicting the operational noise range of motors on the production line in advance becomes crucial for timely adjustments to production parameters and process optimization. This paper introduces a prediction model based on a Multi-Branch Channel–Spatial Adaptive Weighting Strategy (MCSAWS). The model includes a multi-branch feature extraction (MFE) network and a channel–spatial attention module (CSAM). It uses the vibration and noise data from micro-motors’ idle operations on the production line as input to efficiently predict the operational noise uncertainty interval of automotive micro-motors. The model employs the VAE-GAN approach for data augmentation (DA) and uses Gammatone filters to emphasize the noise at the commutation frequency of the motor. The model was compared with Convolutional Neural Networks (CNNs) and Multilayer Perceptrons (MLPs). Experimental results demonstrate that the MCSAWS method is superior to conventional methods in prediction accuracy and reliability, confirming the feasibility of the proposed approach. This research can help control noise uncertainty in micro-motors’ production and manufacturing processes in advance.

Identifying obesogenic environment through spatial clustering of body mass index among adults

BackgroundThe escalating trend of obesity in Malaysia is surmounting, and the lack of evidence on the environmental influence on obesity is untenable. Obesogenic environmental factors often emerge as a result of shared environmental, demographic, or cultural effects among neighbouring regions that impact lifestyle. Employing spatial clustering can effectively elucidate the geographical distribution of obesity and pinpoint regions with potential obesogenic environments, thereby informing public health interventions and further exploration on the local environments. This study aimed to determine the spatial clustering of body mass index (BMI) among adults in Malaysia.MethodThis study utilized information of respondents aged 18 to 59 years old from the National Health and Morbidity Survey (NHMS) 2014 and 2015 at Peninsular Malaysia and East Malaysia. Fast food restaurant proximity, district population density, and district median household income were determined from other sources. The analysis was conducted for total respondents and stratified by sex. Multilevel regression was used to produce the BMI estimates on a set of variables, adjusted for data clustering at enumeration blocks. Global Moran’s I and Local Indicator of Spatial Association statistics were applied to assess the general clustering and location of spatial clusters of BMI, respectively using point locations of respondents and spatial weights of 8 km Euclidean radius or 5 nearest neighbours.ResultsSpatial clustering of BMI independent of individual sociodemographic was significant (p < 0.001) in Peninsular and East Malaysia with Global Moran’s index of 0.12 and 0.15, respectively. High-BMI clusters (hotspots) were in suburban districts, whilst the urban districts were low-BMI clusters (cold spots). Spatial clustering was greater among males with hotspots located closer to urban areas, whereas hotspots for females were in less urbanized areas.ConclusionObesogenic environment was identified in suburban districts, where spatial clusters differ between males and females in certain districts. Future studies and interventions on creating a healthier environment should be geographically targeted and consider gender differences.

Computational assessment of the aesthetic quality of map colors

AbstractCrafting beautiful map colors is challenging for not only experts but also novices, suggesting the need for an aesthetic quality assessment of map colors for the effective design of maps and visualizations. To fill this gap, we present a computational method to assess the aesthetic quality of map colors. First, we couple the idea of computational aesthetics with map aesthetic principles and identify four and two basic types of aesthetic features in terms of order and complexity, respectively. Then, we collect 2000 map samples and derive 149 aesthetic metrics by instantiating the above aesthetic features while considering the spatial weights and figure–ground organization of the map samples. We also recruit participants (N = 438) to rate the aesthetic quality of the map colors. Finally, we train an aesthetic predictor by fitting those aesthetic metrics with user ratings. The experimental results show that the proposed method can assess aesthetic quality of map colors with high accuracy (R2 = 0.73 on the training set and R2 = 0.65 on the validation set). We also explore the dominant aesthetic metrics that positively and negatively influence aesthetic appreciation, as well as those metrics that have no significant influence. This work offers a portable and flexible aesthetic quality assessment approach for map colors and can be further improved by considering complex symbols, spatial structures, and color–semantic and color–emotion associations.

The spatial spillover effect of financial growth on high-quality development: Evidence from Yellow River Basin in China

River basin cities are areas with remarkable conflicts between the human activity and the ecological environment. They are also important targets for policy implementation of sustainable and high-quality development (HD) in various countries around the world. This article exploits the panel data of 99 cities located in the Yellow River Basin (YRB) from 2006 to 2019 to empirically analyze the spatial effect of financial growth on HD. Spatial weights participated econometric models are utilized to analyze this spatial effect. Empirical results reveal that: (1) the HD in the YRB shows a strong positive spatial autocorrelation. (2) Financial growth exerts an N-shaped curve effect on the HD from a long-term perspective. When this influence spills out to the surroundings, it exhibits an inverted U-shaped characteristic. (3) Green innovation can be an important intermediary factor in the influence of financial growth on HD. (4) The influence of financial growth on HD appears stronger in regions with higher economic levels, where N-shaped effects can be transmitted to the surrounding regions. However, the backward economic development in low-economy regions prevents the spatial spillover of N-shaped effects. This study can be instrumental for countries to formulate financial policies that aim to promote HD in river basin cities.

Novel definition of local baseline values for potentially toxic elements in Czech farmland using adaptive spatial weighting

Exploratory data analysis is commonly used in geoscientific research to identify various data populations within datasets. Frequently, the distribution-wise metrics for the statistical centre and spread are combined to define normal (background) ranges of topsoil contents of potentially toxic elements. When decreasing the geographical scale, the survey areas turn too heterogeneous for the statistical definition of a single background range. The traditional solution is the domain approach wherein various data populations (and their statistical parameters) can be attributed to contextual (geological, ecological) information. Nevertheless, summarising the entire data set from large areas as a single statistical entity would provide too much data reduction which would decrease sensitivity of detecting localised anthropogenic contamination and work wrong in areas of geogenic anomalies spatially larger than contamination. In this paper, we tested a novel numerical solution for deriving local distribution-wise baseline values via spatially limited sliding window combined with geographical weighting. Considering environmental variables (soil and topographical properties) at every analysed soil sample point, we extended the geographical kernel weighting approach which considers only spatial dimension (given by geographical coordinates). The advanced version combines two similarity modes to assign highest weights to the nearest points expected to share similar environmental contexts within the user-defined moving kernel. The method was implemented for data-mining in the Czech high-density monitoring data for agricultural soils which had to be firstly regressed to achieve analytical harmony between two distinct extraction methods employed in that monitoring, in particular cold diluted nitric acid and hot aqua regia. After the reliable harmonisation, local baseline values were delivered as the localised outer limits of variation using the proposed double-weighted kernel approach. We compared the estimated localised background ranges for 10 potentially toxic elements (As, Be, Cd, Co, Cr, Cu, Ni, Pb, V, and Zn) with those based on the conventional substrate-wise domain approach and single nationwide legislation thresholds. The comparison was also efficient in identifying an inappropriate aggregation of some geological units. Finally, the kernel approach delivered regional outer limits of variability sensitive to subtle regional variations of topsoil geochemistry.

Bone feature quantization and systematized attention gate UNet-based deep learning framework for bone fracture classification

Bones collaborate with muscles and joints to sustain and maintain our freedom of mobility. The proper musculoskeletal activity of bone protects and strengthens the brain, heart, and lung function. When a bone is subjected to a force greater than its structural capacity, it fractures. Bone fractures should be detected with the appropriate type and should be treated early to avoid acute neurovascular complications. The manual detection of bone fracture may lead to highly delayed complications like malunion, Joint stiffness, Contractures, Myositis ossificans, and Avascular necrosis. A proper classification system must be integrated with deep learning technology to classify bone fractures accurately. This motivates me to propose a Systematized Attention Gate UNet (SAG-UNet) that classifies the type of bone fracture with high accuracy. The main contribution of this research is two-fold. The first contribution focuses on dataset preprocessing through feature extraction using unsupervised learning by adapting the Growing Neural Gas (GNG) method. The second contribution deals with refining the supervised learning Attention UNet model that classifies the ten types of bone fracture. The attention gate of the Attention UNet model is refined and applied to the upsampling decoding layer of Attention UNet. The KAGGLE Bone Break Classification dataset was processed to extract only the essential features using GNG extraction. The quantized significant feature RGB X-ray image was divided into 900 training and 230 testing images in the ratio of 80:20. The training images are fitted with the existing CNN models like DenseNet, VGG, AlexNet, MobileNet, EfficientNet, Inception, Xception, UNet and Attention UNet to choose the best CNN model. Experiment results portray that Attention UNet offers the classification of bone fractures with an accuracy of 89% when testing bone break images. Now, the Attention UNet was chosen to refine the Attention gate of the Decoding upsampling layer that occurs after the encoding layer. The Attention Gate of the proposed SAG-UNet forms the gating coefficient from the input feature map and gate signal. The gating coefficient is then processed with batch normalization that centers the aligned features in the active region, thereby leaving the focus on the unaligned weights of feature maps. Then, the ReLU activation function is applied to introduce the nonlinearity in the aligned features, thereby learning the complex representation in the feature vector. Then, dropout is used to exclude the error noise in the aligned weights of the feature map. Then, 1 × 1 linear convolution transformation was done to form the vector concatenation-based attention feature map. This vector has been applied to the sigmoid activation to create the attention coefficient feature map with weights assigned as ‘1’ for the aligned features. The attention coefficient feature map was grid resampled using trilinear interpolation to form the spatial attention weight map, which is passed to the skip connection of the next decoding layer. The implementation results reveal that the proposed SAG-UNet deep learning model classifies the bone fracture types with a high accuracy of 98.78% compared to the existing deep learning models.

HMT: Hybrid mechanistic Transformer for bio-fabrication prediction under complex environmental conditions

In the field of bio-manufacturing, hybrid models that combine mechanistic models with data-driven models have not been widely adopted. This is primarily due to the complexity of the manufacturing process itself and the limited quality of available bio-manufacturing data. To address these limitations, this study introduces a novel hybrid model that combines a mechanistic-feature network with a Transformer-based data-driven model (HMT), effectively overcoming these challenges. The HMT integrates mechanistic and data-driven modeling to accurately predict target variables in bio-manufacturing processes. Initially, the HMT utilizes a mechanistic model to explore the feature relationships of the target sequence variables. It then compresses temporal domain variables into a univariate time series and applies spatial fusion weights for spatial domain variables, achieving variable fusion independently in each domain. Subsequently, a Transformer-based data-driven model is employed to model and predict the fused target variables, utilizing attention mechanisms to capture feature relationships within and between sequences. Upon comparison with various data-driven models, particularly those in the bio-manufacturing sector that employ artificial neural network approaches, the hybrid model demonstrates superior performance in mean squared error, mean absolute error, mean absolute percentage error, robustness, and Theil’s U statistic. The state-of-the-art performance of the HMT in long sequence prediction and short sequence prediction confirms its feasibility and potential for industrial applications.11The code is under the address: https://github.com/Scholar-Song/HMT.

Assessing the applicability of rainfall index spatial interpolation in predicting landslide susceptibility: A study in Brunei Darussalam

Abstract Rainfall is crucial in meteorology and hydrology, significantly impacting landslide risk assessment. This study focused on evaluating rainfall’s role in determining landslide susceptibility in Brunei Darussalam’s Jalan Kota Batu-Subok and Jalan Jangsak-Tutong regions. Using monthly rainfall data (2008 – 2018) from four weather stations, three spatial interpolation methods inverse distance weighting (IDW), radial basis function (RBF), and global polynomial interpolation (GPI) were assessed. The RBF proved superior in predicting rainfall distribution, evidenced by lower error metrics and higher correlation coefficients. The landslide susceptibility index (LSI) derived from the RBF’s rainfall interpolation showed high accuracy in identifying landslide-prone areas, with success rates between 89 % and 94.3 %, and prediction rates from 85.2 % to 95.9 % across the two areas studied. These findings suggest that the RBF-derived LSI is a reliable tool for landslide risk assessment. However, the LSI’s stability, irrespective of the rainfall data or interpolation method used, indicates that factors like terrain and human activities might have a more significant impact on landslide risks than rainfall alone. This research highlights the importance of considering various factors in landslide risk management and land-use planning, offering valuable insights for policymakers and local authorities.

Public debt, investment and economic growth dynamics: Do geographical proximity and spatial spillover effects matter?

This study employed an innovative spatial econometrics approach to analyze the spillover effects of foreign public debt and foreign direct investment (FDI) on regional economic growth dynamics of Eastern Africa. Specifically, we utilize Spatial Durbin Fixed Effect (SDM-FE) Model, drawing on balanced panel data from the World Development Indicators (WDI) for the period 1992–2019. This study contributes by utilizing a spatial weight matrix and investigating its interactions with economic growth and various other macro-regional factors, thereby unveiling the presence of spatial spillover effects and the impact of geographical proximity of countries within the region. Our results indicate that foreign public debt, FDI, gross fixed capital formation, human capital development, inflation, and government expenditure, significantly have spatial spillover effects on regional economic growth. Additionally, the results underscore the presence of regional growth divergence and provide further evidence of substantial regional spatial dependence in economic growth. Based on these outcomes, it is evident that foreign public debt has a negative impact on economic growth, which is consistent with the debt overhang theory and the crowding-out hypothesis. Moreover, the foreign public debt and regional growth policies should be devised considering the geographical interconnectedness between countries in the region, as the research output has revealed the impacts of growth and debt spillovers among the countries studied. Policies that prioritize these objectives would promote long-term regional economic growth and prevent the effects of excessive foreign public debt by creating a favorable macroeconomic climate.

An ensemble spatial prediction method considering geospatial heterogeneity

Ensemble learning synthesizes the advantages of different models and has been widely applied in the field of spatial prediction. However, the nonlinear constraints of spatial heterogeneity on the model ensemble process make it difficult to adaptively determine the ensemble weights, greatly limiting the predictive ability of the ensemble learning model. This paper therefore proposes a novel geographical spatial heterogeneous ensemble learning method (GSH-EL). Firstly, the geographically weighted regression model, geographically optimal similarity model, and random forest model are used as three base learners to express local spatial heterogeneity, global feature correlation, and nonlinear relationship of geographic elements, respectively. Then, a spatially weighted ensemble neural network module (SWENN) of GSH-EL is proposed to express spatial heterogeneity by exploring the complex nonlinear relationship between the spatial proximity and ensemble weights. Finally, the outputs of the three base learners are combined with the spatial heterogeneous ensemble weights from SWENN to obtain the spatial prediction results. The proposed method is validated on the PM2.5 air quality and landslide dataset in China, both of which obtain more accurate prediction results than the existing ensemble learning strategies. The results confirm the need to accurately express spatial heterogeneity in the model ensemble process.

Uncertainty estimation- and attention-based semi-supervised models for automatically delineate clinical target volume in CBCT images of breast cancer

ObjectivesAccurate segmentation of the clinical target volume (CTV) of CBCT images can observe the changes of CTV during patients' radiotherapy, and lay a foundation for the subsequent implementation of adaptive radiotherapy (ART). However, segmentation is challenging due to the poor quality of CBCT images and difficulty in obtaining target volumes. An uncertainty estimation- and attention-based semi-supervised model called residual convolutional block attention-uncertainty aware mean teacher (RCBA-UAMT) was proposed to delineate the CTV in cone-beam computed tomography (CBCT) images of breast cancer automatically.MethodsA total of 60 patients who undergone radiotherapy after breast-conserving surgery were enrolled in this study, which involved 60 planning CTs and 380 CBCTs. RCBA-UAMT was proposed by integrating residual and attention modules in the backbone network 3D UNet. The attention module can adjust channel and spatial weights of the extracted image features. The proposed design can train the model and segment CBCT images with a small amount of labeled data (5%, 10%, and 20%) and a large amount of unlabeled data. Four types of evaluation metrics, namely, dice similarity coefficient (DSC), Jaccard, average surface distance (ASD), and 95% Hausdorff distance (95HD), are used to assess the model segmentation performance quantitatively.ResultsThe proposed method achieved average DSC, Jaccard, 95HD, and ASD of 82%, 70%, 8.93, and 1.49 mm for CTV delineation on CBCT images of breast cancer, respectively. Compared with the three classical methods of mean teacher, uncertainty-aware mean-teacher and uncertainty rectified pyramid consistency, DSC and Jaccard increased by 7.89–9.33% and 14.75–16.67%, respectively, while 95HD and ASD decreased by 33.16–67.81% and 36.05–75.57%, respectively. The comparative experiment results of the labeled data with different proportions (5%, 10% and 20%) showed significant differences in the DSC, Jaccard, and 95HD evaluation indexes in the labeled data with 5% versus 10% and 5% versus 20%. Moreover, no significant differences were observed in the labeled data with 10% versus 20% among all evaluation indexes. Therefore, we can use only 10% labeled data to achieve the experimental objective.ConclusionsUsing the proposed RCBA-UAMT, the CTV of breast cancer CBCT images can be delineated reliably with a small amount of labeled data. These delineated images can be used to observe the changes in CTV and lay the foundation for the follow-up implementation of ART.

Bayesian Spatio-Temporal Multilevel Modelling of Patient-Reported Quality of Life following Prostate Cancer Surgery.

Globally, prostate cancer is the second leading cause of cancer deaths among males. It is the most commonly diagnosed cancer in Australia. The quality of life of prostate cancer patients is poorer when compared to the general population due to the disease itself and its related complications. However, there is limited research on the geographic pattern of quality of life and its risk factors in Victoria. Therefore, an examination of the spatio-temporal pattern and risk factors of poor quality of life, along with the impact of spatial weight matrices on estimates and model performance, was conducted. A retrospective study was undertaken based on the Prostate Cancer Outcome Registry-Victoria data. Patient data (n = 5238) were extracted from the Prostate Cancer Outcome Registry, a population-based clinical quality outcome assessment from 2015 to 2021. A Bayesian spatio-temporal multilevel model was fitted to identify risk factors for poor quality of life. This study also evaluated the impact of distance- and adjacency-based spatial weight matrices. Model convergence was assessed using Gelman-Rubin statistical plots, and model comparison was based on the Watanabe-Akaike Information Criterion. A total of 1906 (36.38%) prostate cancer patients who had undergone surgery experienced poor quality of life in our study. Belonging to the age group between 76 and 85 years (adjusted odds ratio (AOR) = 2.90, 95% credible interval (CrI): 1.39, 2.08), having a prostate-specific antigen level between 10.1 and 20.0 (AOR = 1.33, 95% CrI: 1.12, 1.58), and being treated in a public hospital (AOR = 1.35, 95% CrI: 1.17, 1.53) were significantly associated with higher odds of poor quality of life. Conversely, residing in highly accessible areas (AOR = 0.60, 95% CrI: 0.38, 0.94) was significantly associated with lower odds of poor prostate-specific antigen levels. Variations in estimates and model performance were observed depending on the choice of spatial weight matrices. Belonging to an older age group, having a high prostate-specific antigen level, receiving treatment in public hospitals, and remoteness were statistically significant factors linked to poor quality of life. Substantial spatio-temporal variations in poor quality of life were observed in Victoria across local government areas. The distance-based weight matrix performed better than the adjacency-based matrix. This research finding highlights the need to reduce geographical disparities in quality of life. The statistical methods developed in this study may also be useful to apply to other population-based clinical registry settings.

GS-DeepLabV3+: A mountain tea disease segmentation network based on improved shuffle attention and gated multidimensional feature extraction

The prevention and control of Camellia Oleifera diseases can produce significantly comprehensive economic benefits. And precisely identifying the types of diseases on oil tea leaves can enhance the effectiveness of pest and disease control. Currently, the most effective deep learning methods for segmentation are suffering from the challenges such as indistinct disease spots on the actual oil tea leaves that are difficult to differentiate from the background, the high similarities in characteristics between different diseases, and the small target area of disease regions that are easily ignored. In this paper, GS-DeepLabV3+ network is proposed, which is an improvement upon DeepLabV3+ and aims to effectively enhance disease identification and segmentation rates. Initially, a gated pyramid feature fusion structure is proposed, utilizing a special module to merge features of different dimensions to enhance differentiation of similar features. Secondly, an improved grouped attention fusion mechanism is introduced, which allocates channel and spatial weights to different feature groups, enhancing the extraction of detailed features. Lastly, we substituted the backbone network of DeepLabV3+ with the more lightweight MobileNetV2, which significantly reduces the number of parameters while more stably extracting multidimensional features. In comparison with the existing segmentation networks on a self-created oil tea leaf disease data set, GS-DeepLabV3+ achieved a mIoU of 87.77% and an average pixel accuracy of 94.55%. Compared to the base DeepLabV3+ network, there was a 4.14% improvement in mIoU, a 3.58% increase in mPA, and a 1.4% rise in precision. The improved attention mechanism contributed to a 1.98% increase in mIoU, and the gated multidimensional fusion mechanism resulted in a 1.03% improvement. These enhancements can provide technical references for crop protection and pest disease segmentation.

The Impact of Green Finance and Financial Technology on Regional Green Energy Technological Innovation Based on the Dual Machine Learning and Spatial Econometric Models

Regional green energy technological innovation is an important means to alleviate economic–environmental contradictions. The purpose of this study was to explore the mechanisms of green finance, financial technology, and regional green energy technological innovation. In this study, we constructed dual machine learning models, spatial econometric models, and panel threshold effect models to investigate the effects of green finance and financial technology on regional green energy technological innovation, using panel data from 266 cities nationwide from 2009 to 2021. The research findings are as follows: (1) Both green finance and financial technology significantly promote regional green energy technological innovation. (2) Based on a spatial weight matrix embedded in economic geography, both green finance and financial technology generate positive spatial spillover effects on regional green energy technological innovation. (3) The interaction between green finance and financial technology significantly contributes to regional green energy technological innovation. Financial technology can strengthen the positive local and neighboring effects of green finance on regional green energy technological innovation. (4) Based on the threshold effect of financial technology, green finance cannot significantly promote regional green energy technological innovation when financial technology is in an underdeveloped stage. With the advancement of financial technology, green finance continues to have a positive impact on regional green energy technological innovation. Based on this analysis and our conclusions, we propose practical policy recommendations that can provide a more sustainable approach to green energy technology innovation.

Residual spatial fusion network for RGB-thermal semantic segmentation

Semantic segmentation plays an important role in widespread applications such as autonomous driving and robotic sensing. This work focuses on RGB-Thermal (RGB-T) semantic segmentation, since RGB images are easily affected by lighting conditions, e.g., darkness, and thermal images are robust to the night scenario as a compensating modality. However, existing works either simply fuse RGB-T images or adopt the encoder with the same backbone for the two modalities, neglecting the semantic difference under varying lighting conditions. Therefore, we present a Residual Spatial Fusion Network (RSFNet) for RGB-T semantic segmentation. Specifically, RSFNet employs an asymmetric encoder to learn the compensating features of RGB and thermal images. To fuse the dual-modality features, it generates pseudo-labels by adopting the saliency detection skill to supervise feature learning, and develops the Residual Spatial Fusion (RSF) module with structural re-parameterization to learn more promising features by spatially fusing the cross-modality features. RSF adopts a hierarchical feature fusion to aggregate multi-level features, and applies the spatial weights with the residual connection to adaptively control the multi-spectral feature fusion by the confidence gate. Extensive experiments were carried out on two benchmarks, i.e., MFNet database and PST900 database. The results have shown the state-of-the-art segmentation performance of our method, which strikes a good balance between accuracy and speed.

HETEROSKEDASTICITY ROBUST SPECIFICATION TESTING IN SPATIAL AUTOREGRESSION

Spatial autoregressive (SAR) and related models offer flexible yet parsimonious ways to model spatial and network interactions. SAR specifications typically rely on a particular parametric functional form and an exogenous choice of the so-called spatial weight matrix with only limited guidance from theory in making these specifications. Also, the choice of a SAR model over other alternatives, such as spatial Durbin (SD) or spatial lagged X (SLX) models, is often arbitrary, raising issues of potential specification error. To address such issues, this paper develops a new specification test within the SAR framework that can detect general forms of misspecification including that of the spatial weight matrix, the functional form and the model itself. The test is robust to the presence of heteroskedasticity of unknown form in the disturbances and the approach relates to the conditional moment test framework of Bierens ([1982, Journal of Econometrics 20, 105–134], [1990, Econometrica 58, 1443–1458]). The Bierens test is shown to be inconsistent in general against spatial alternatives and the new test introduces modifications to achieve test consistency in the spatial setting. A central element is the infinite-dimensional endogeneity induced by spatial linkages. This complexity is addressed by introducing a new component to the omnibus test that captures the effects of potential spatial matrix misspecification. With this modification, the approach leads to a simple pivotal test procedure with standard critical values that is the first test in the literature to have power against misspecifications in the spatial linkages. We derive the asymptotic distribution of the test under the null hypothesis of correct SAR specification and prove consistency. A Monte Carlo study is conducted to study its finite sample performance. An empirical illustration on the performance of the test in modeling tax competition in Finland is included.

Unveiling land use-carbon Nexus: Spatial matrix-enhanced neural network for predicting commercial and residential carbon emissions

Carbon emissions play a pivotal role in driving global warming, with commercial and residential sectors ranking as the third-largest source after industrial and transportation sectors. Consequently, establishing a framework for spatial characterizing and predicting carbon emissions from these sectors is vital for effective carbon reduction planning. This study proposes an optimized Backpropagation Neural Network, incorporating a Spatial Weight Matrix, to model the relationship between land use and carbon emissions, taking spatial effects into account. Additionally, an optimized Random Forest model with variable search step lengths is developed for large-scale and multi-class land use prediction. Beijing, China serves as the study site, yielding the following findings: (1) The optimized Backpropagation Neural Network outperforms traditional models, achieving goodness of fit of 0.999 and 0.997 in training and testing datasets, respectively. (2) The optimized Random Forest model, while reducing prediction precision for ecological areas (Root Mean Square Error increases by 4.12 %), enhances overall model performance by 73.95 % in a single iteration. (3) By 2025, commercial and residential carbon emissions in Beijing are projected to reach 8.91 million tons. (4) The study characterizes the spatial patterns of commercial and residential carbon emissions at a 1 km resolution for the next 15 years, revealing statistical features. (5) Expansion of mixed commercial and residential land relies on service-oriented development, positively impacting regional carbon emission intensity control.