In spatial analysis and modeling of urban crime, the spatial autocorrelation of model residuals poses an significant obstacle to model parameter estimation and produces deviations in analysis of the determinants of urban crime. The presence of significant spatial autocorrelation of model residuals and overdispersion of the model could lead to biased estimates and misleading inferences, even resulting in wrong conclusions. This study employed a new spatial regression method, namely Poisson regression with Eigenvector Spatial Filtering, to solve the problem of model residual spatial autocorrelation and model overdispersion to avoid subsequent biased estimation in model results. To explain the spatial variation of urban crime, we used two theories in spatial crime analysis: crime pattern theory and social disorganization theory. The case study focused on the main urban area of the Haining city in Zhejiang province, China, and the crime data that we used were larceny-theft over a four-year period, from January 2018 to September 2021. Other datasets that we employed for generating covariates included POI data of various facilities in Haining, the Luojia 1-01 nighttime light data, and the WorldPop global population data. We established a Poisson regression model with eigenvector spatial filtering and further identified several important determinants of larceny-theft with unbiased model parameters. The major findings are as follows: (1) The Poisson regression with eigenvector spatial filtering identified the spatial autocorrelation of model residuals, ensuring no significant spatial autocorrelation issue in model residuals. This can improve the model's goodness of fit, correct model parameter estimation, alleviate the impact of overdispersion, and retrieve omitted variables. More importantly, the eigenvector spatial filtering method could be applied to other generalized linear models such as Poisson regression; (2) The results of Emerging Hot Spot Analysis showed that the absolute number of larceny-theft decreased during the period of COVID-19 pandemic, and crime hot spots occurred in the central places of the main urban area of Haining while the cold spots exhibited a trend of multipoint distribution; (3) The level of relative deprivation measured by per capita nighttime light had a significant impact on larceny-theft in the unbiased model with eigenvector spatial filtering; (4) The crime generator, attractor and enabler in various built environment of interest had a significant impact on larceny-theft. The inconsistencies with the conclusions of previous studies were also discussed.
The harms of excessive drinking are important issues faced by China and other countries around the world. Previous studies have identified certain relationships between drinking behaviors and environmental factors, with limited consideration of the dynamic geographic context and spatiotemporal heterogeneity of these relationships. This study constructs a theoretical framework for the relationships between excessive drinking and geographical environment factors (i.e., alcohol availability, spatial disorder, and social disorder). Using multi-source spatiotemporal data, including the alcoholism cases received by Guangzhou Emergency Medical Command Center and mobile signaling data, the study explores the spatiotemporal patterns and the influencing factors of excessive drinking in a dynamic geographic context. Results show that: (1) Excessive drinking events mostly occur on weekends, at night, and in urban commercial centers. Its spatial distribution is balanced in the daytime but congregated in the nighttime; (2) Excessive drinking is influenced by alcohol availability (positively correlated with the density of KTVs and traffic stations). The positive impact of KTVs is prominent in the old urban area in the daytime and is prominent in more areas in the nighttime. The positive impact of traffic stations is mainly observed at the edge of central urban area and in suburban areas, and is relatively stronger in the daytime; (3) Excessive drinking is also affected by certain spatial disorder factors (positively correlated with the land development level), but has no significant relationship with building age and streetscape messiness. The positive impact of land development level is prominent in the new urban scenic core area and the urban villages at the urban-rural fringe, with a stronger impact in the daytime; (4) The relationship between excessive drinking and social disorder shows greater spatiotemporal differentiation and complexity. Overall, excessive drinking is negatively correlated with the proportion of nonlocal-born population, and positively correlated with the proportion of low-educated population and residential rent; (5) By analyzing different leading factors influencing excessive drinking in the daytime or nighttime, four types of spatial units can be identified, allowing for respective interventions in these zones to reduce the risk of excessive drinking. This study reveals the relationships between excessive drinking and geographical environment in a dynamic geographic context and at a smaller spatial scale, which fills the gap of lacking smaller-scale analysis, spatiotemporal heterogeneity perspective, and consideration of dynamic geographic context in previous studies. Our results provide insights for alcohol policy making, regional security and medical resource allocation, spatial planning, and environmental intervention related to health issues.
Low-altitude space is an important component of urban space. Requisite measures for precise and meticulous management of urban low-altitude are indispensable. The urban low-altitude space should be characterized by spatial coordinates and possess significant geographic attributes. With the increasing adoption of low-altitude applications in urban areas, the intricate utilization of space, represented by low-altitude traffic, has transcended conventional airspace boundaries and encroached upon near-ground urban space, exerting an impact on urban architecture and human settlement environment. Emphasizing the 3D land space's utilization and service, the issue of the utilization and management of 3D land space has become increasingly conspicuous. The land parcel serves as the fundamental unit for urban land management, recording information of ownership relationships, spatial rights, and interests information. The establishment of easements between land parcels ensures the lawful utilization of relevant spaces. Given the inclusion of low-altitude flight activities within the purview of urban land management, there is an urgent imperative to elucidate the spatial utilization and impact of low-altitude passage processes on land space while establishing the service rights of urban low-altitude passage to safeguard parcel interests. However, the concept of easement is limited by the cognitive constraints imposed by a 2D land plane, neglecting the modeling and representation of 3D space. Precisely articulating the easement relationship formed by low-altitude passage activities in urban low-altitude space poses a significant challenge. Utilizing GIS technology for modeling urban spaces and facilitating the characterization and mapping from the physical to digital realms has consistently served as a crucial information tool in urban space management. Drawing upon the core principles of GIS modeling methodology and conceptual modeling, this paper presents a conceptual approach for describing low-altitude passage easement in urban areas. By analyzing the movement conditions of aircraft in urban low-altitude traffic, considering the impact on human settlement rights caused by overflight, and examining the path utilization of aircraft in three-dimensional space, we develop a constrained spatial modeling for low-altitude passage easement as a geometric description of the conceptual model. By integrating the spatial characteristics of 3D land parcels, we integrate the supply and service conditions of these parcels in terms of their space utilization, aircraft takeoff and landing modes, and path utilization. As a result, we propose a comprehensive supply and service model to address the demand for traffic within 3D parcels. The semantic relationship among low-altitude planning, space value, and space easement is established by extracting the concept of low-altitude access from existing research and regulations, thereby forming a conceptual model. Finally, we conduct an experimental demonstration using a logistics transportation case study in Shenzhen to instantiate the model and achieve 3D visualization. The findings demonstrate that the integrated approach of "space utilization conditions-data modeling-visual expression", implemented around the conceptual model, effectively describes urban low-altitude passage easement and conveys the equitable relationship of urban land space in low-altitude application scenarios, thereby providing valuable support for urban low-altitude management.
Spatial-temporal data missingness and sparsity are prevalent phenomena, for which spatial-temporal interpolation serves as a critical methodology to address these issues. Spatial-temporal interpolation constitutes a significant research domain within the field of Geographical Information Science. This technique enables the capture of dependencies in spatial-temporal data and the estimation of the geometric and attribute variations of geographical phenomena over time. With the advancement of geospatial technologies, particularly Geographic Information Systems, contemporary spatial-temporal interpolation methods predominantly rely on statistical, machine learning, and deep learning approaches that account for both temporal and spatial dimensions. These methods aim to reveal the evolutionary processes and spatial-temporal distribution patterns inherent in the data. However, a majority of such techniques often overlook long-term dependencies and contextual spatial information when interpolating. This study proposes an innovative model that intertwines Long Short-Term Memory (LSTM) networks with spatial attributes to address these limitations effectively. The proposed model operates through several key stages: (1) It employs a dedicated spatial layer to systematically eliminate weakly correlated information, focusing on extracting and feeding more significantly correlated spatial data into the LSTM network. (2) Given that conventional Artificial Neural Network (ANN) models are unable to consider the impact of the temporal dimension on interpolation, and unidirectional LSTM models can only factor in past moments' influence without utilizing future moment information, this research adopts a Bidirectional LSTM (BiLSTM) architecture. The BiLSTM inherently captures both spatial and temporal dependencies, thereby overcoming previous limitations. (3) To further enhance its performance by efficiently extracting comprehensive global spatial features while maintaining the advantages of bidirectional modeling offered by BiLSTM, we integrate a self-attention mechanism into the BiLSTM framework. This results in a novel, fused Bidirectional LSTM Interpolation Model with Spatial Layer-Self Attention (SL-BiLSTM-SA). In the experimental phase, the SL-BiLSTM-SA model is rigorously applied to a PM2.5 concentration dataset from Shandong Province to conduct a meticulous investigation into its interpolation capabilities. Upon comparative analysis against other models, it is evident that the SL-BiLSTM-SA model outperforms with notably lower error metrics, demonstrating substantial improvements in accuracy—by 39.83% and 36.63% when compared to Spatio-Temporal Ordinary Kriging (STOK) and Genetic Algorithm-optimized Spatio-Temporal Kriging (GA-STK) methods, respectively. Moreover, our model exhibits commendable precision in forecasting high and low concentration levels. By seamlessly integrating spatial information and coupling the strengths of BiLSTM with self-attention mechanisms, this research not only extends the suite of interpolation methods for spatiotemporal data analysis but also furnishes robust theoretical underpinnings and methodological support to facilitate sophisticated spatiotemporal data analyses.
Geographical flows are formed by the movement and exchange of substances, information, energy, and other elements between different locations. The Geographical Multiple Flow (GMF) is the combination of various geographical flows. Human activities such as international trade, population migration, and air transportation have formed multiple global scale GMF, which will provide new information for the study of global issues like international relation. GMF is composed of single geographic flows, containing various types of information. It can effectively compensate for the shortcomings of single geographic flows and reveal the patterns that cannot be reflected by them. Thus, how to extract information from it is an urgent problem. The Global Database of Events, Language, and Tone (GDELT) is a real-time updated global news database. It records all events reported by global news media since January 1st,1979, extracting key information including participants, locations, event types, and emotional tendencies through text analysis. The paper takes the international relation research using GDELT data as the application background, exploring the measurement and community detection methods of GMF. We use networked methods to analyze GMF. Multi-layer network methods are applied to measure the structure of GMF, and the feasibility is verified on the GDELT dataset. Firstly, we select data from GDELT Event Database to construct single geographic flow networks including media networks, legislature networks, and insurgent networks, and then uses them to construct GMF networks based on multiplex network. Secondly, we use multi-layer network methods for measurement and community. For calculating GMF networks, the corresponding single-layer network methods are also used to extract features of each layer. Finally, the temporal variation characteristics of the GMF networks are analyzed. We take one year as a cycle to analyze the similarities and differences of network features at different times and discover the evolution of things reflected in it. Results show that: ① GMF network contains more information than single geographic flow network, which can reflect the characteristics of the system from a more comprehensive perspective; ② The characteristics of GMF can reflect the combined effects of multiple elements; ③ The multi-layer network methods can directly characterize the overall features of geographic multiple flows, avoiding errors caused by comprehensive analysis; ④ Combining GMF networks and geographic flow networks can yield more comprehensive results from different scales. This research provides reference for the development of GMF spatiotemporal analysis methods and the exploration of spatiotemporal big data analysis methods for international relation.
In China, the majority of cities are currently transitioning from incremental planning to inventory planning. The implementation of inventory planning still relies on incremental planning as its foundation, with the two approaches complementing each other and developing in coordination. During this transition, the new national spatial planning system emphasizes the hierarchical transmission of planning, guiding and constraining urban development progressively from macro to micro levels. This study is based on a multi-level Vector Cellular Automata (VCA) model. We use land use data from Jiangyin City in 2012 as the foundation to simulate and analyze land use changes in 2017, and then verify the accuracy of the VCA model. Subsequently, adhering to the spatiotemporal patterns of stock renewal and incremental development, parameters are adjusted and area allocations are made separately to integrate these two indicators. They are then incorporated into the multi-level VCA model. This approach allows for a multi-level simulation of urban spatial growth, considering both stock renewal and incremental development. Then, this approach is applied to predict and analyze multiple scenarios of land use growth in Jiangyin City in 2027 within the context of stock planning. The conclusions of the study are as follows: ① The multi-level VCA decomposes differential driving forces from top to bottom and transmits them collaboratively from bottom to top, achieving both the implementation of upper-level control indicators and fine-grained control of land use simulation. ② The multi-level VCA assigns differentiated speed parameters to different regions, fully considering the spatial heterogeneity of driving factors and land distribution. The overall Figure of Merit (FoM) reaches 24.6%, which is 2.5% higher than that of the single-level VCA. At the local detail level, the multi-level VCA exhibits fewer misclassifications of different land use types and less encroachment on prohibited areas such as water bodies compared to single-level simulations, with better simulation results for linear land parcels. ③ Under the constraints of "baseline control-layered control-indicator control," with the increase in stock renewal speed, the expansion of newly predicted construction land in different scenarios remains within the range of planned control. The scale of expansion for newly predicted construction land is reduced, and it is concentrated in the West, South, and Southeast of Jiangyin city. This transition aligns with the growth boundaries and land layout structure designated in the master planning outline. In different scenarios, stock land renewal is dominated by industrial land, followed by rural construction land. ④ In scenarios (d) and (e), industrial renewal tends toward saturation, achieving overall balanced growth, serving as a key reference for future land use pattern development in Jiangyin City.
Large-scale sudden environmental incidents often disrupt network infrastructures such as fiber cables and base stations. Establishing a special self-organizing network for emergency monitoring data transmission is essential for effective disaster response. Utilizing spatial analysis techniques, this paper guides the emergency communication vehicle to a strategically advantageous position. Building on this, considering factors such as the distribution of emergency personnel, communication performance, and climate conditions, this research intelligently deploys communication drones and establishes communication links to form an emergency terrestrial-aerial communication network using a hybrid hierarchical genetic algorithm. This includes designing a spatio-temporal hierarchical chromosome matrix, a comprehensive fitness evaluation model, and a co-evolutionary mechanism to achieve adaptive transmission of monitoring data. In the event of unexpected changes such as the addition or failure of communication resources or adjustments to communication requirements, the algorithm adaptively updates the deployment positions of emergency communication resources and communication links. This method was validated and analyzed using a tailings pond in Xinjiang as an experimental area. The deployment solutions were visualized on a 3D Earth platform, and the fitness convergence ability, dynamic adjustability, and implementation effectiveness were evaluated. The algorithm demonstrated an approximate 30% improvement in fitness after 260 iterations, showing a higher convergence speed and improvement amplitude compared to other five algorithms such as random search. The communication nodes achieved comprehensive coverage and balanced distribution of monitoring points under different scenarios within the endurance period, meeting the requirements for on-site emergency monitoring response, thereby proving the algorithm's reliability and feasibility.
Soil salinization is a significant issue that not only leads to ecological problems like land desertification but also poses a threat to food security by reducing the quantity and quality of arable land. Therefore, it is crucial to rapidly and accurately obtain information about soil salinization for timely and effective soil management. In recent years, the development of microwave remote sensing has provided new methods for large-scale and rapid monitoring of soil salinization, with Synthetic Aperture Radar (SAR) data-based soil salinization monitoring becoming a hotspot in remote sensing research. Recent advancements in SAR remote sensing for soil salinization monitoring can be summarized in four main aspects: (1) Microwave scattering mechanism of saline soil: Research has clarified the correlation between soil salinity and radar backscattering coefficients, providing a basis for estimating soil salinity using SAR data. Understanding the microwave scattering mechanism of saline soil is essential for accurately interpreting SAR data and extracting meaningful information about soil salinization. (2) Construction and extraction of characteristic parameters of saline soil: The construction and extraction of characteristic parameters of saline soil have shown a trend towards diversification and integration. Various parameters, such as backscattering coefficients, polarization ratios, and texture features, are used to characterize the properties of saline soil. By utilizing a combination of these parameters, researchers can obtain a more comprehensive understanding of soil salinization. (3) Classification of saline soil: The classification methods for saline soil have shifted from traditional machine learning to deep learning methods. Deep learning algorithms, such as convolutional neural networks, have shown promising results in accurately classifying saline soil areas. These advanced techniques enable the identification and mapping of different levels of soil salinity, aiding in effective soil management strategies. (4) Inversion of soil salinity: The inversion of soil salinity has transitioned from regression analysis to inversion methods that combine scattering mechanisms. By considering the microwave scattering mechanisms and using multiple data sources, more accurate estimations of soil salinity can be obtained. This approach allows for a better understanding of the spatial distribution and variability of soil salinity, facilitating targeted interventions and management practices. Despite these advancements, there are still challenges and issues in the current research on soil salinization monitoring using SAR remote sensing. Some of these challenges include the influence of multiple factors on the relationship between soil salinity and backscattering coefficients, the need for further analysis of mechanisms, the construction of characteristic parameters, and the fusion of multi-source data for achieving high-precision soil salinization monitoring using SAR remote sensing.
ICESat-2, a new-generation satellite for spaceborne lidar altimetry, adopts a multi-beam single-photon counting regime. The presence of a large number of noisy photons in its detection data is well known, and the large amount of photonic data poses a challenge to data transmission and processing due to the limited processor performance and satellite storage resources on board the satellite. Therefore, in order to efficiently denoise the raw detection data in orbit and reduce the data volume for transmission from space to ground, this study proposes a histogram denoising algorithm based on the Poisson distribution characteristics of the raw detection photons. The method includes creating vertical and skewed histograms of the photon point cloud. Firstly, the point cloud data segment is divided into a two-dimensional grid to form a vertical histogram based on the photon transmission distance. The thresholds for distinguishing signal and noise photons are calculated from the mean and standard deviation of the number of photons in the histogram box, and signal photons are assigned with low, medium, and high confidence labels to characterize signal reliability. This is the process of vertical histogramming. Secondly, the slope information is obtained by linearly fitting the medium- and high-confidence photons from the first step. The photon transmission distance is projected to the vertical direction along the slope to form a tilted histogram, and the signal is recognized twice, and the confidence labels are merged. Finally, the noise is eliminated to achieve the purpose of compressing the original data for downlinking. This study also compares the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm and Ordering Points to Identify the Clustering Structure (OPTICS) algorithm to evaluate the performance of the histogram denoising algorithm. Experiments were conducted using ATL02 data for eight surface types. The results show that DBSCAN and OPTICS are not suitable for data denoising in urban areas, with F-values of 0.766 and 0.765, respectively. In contrast, the histogram algorithm is robust and adaptable to different landform types, with F-values of more than 0.90 in all the experimental areas. The DBSCAN algorithm results in the loss of signal photons, while the OPTICS algorithm produces spurious signal clusters. In contrast, these problems are effectively avoided due to the inclusion of signal rate constraints and horizontal to overlapping division histogram processing techniques in the histogram algorithm. Among the histogram algorithms, the vertical histogram algorithm achieves a signal photon recall (R) of 1, and the average precision (P) and F-value are more than 0.90. This improves the operation efficiency by 12 times, 3473 times, and 1528 times compared with the vertical/skewed histogram, DBSCAN, and OPTICS algorithms, respectively. In addition, the average operation time of the vertical histogram algorithm is only 0.048 seconds, realizing efficient data denoising and compression. The denoising result of the vertical histogram algorithm initially meets the requirement of on-orbit processing efficiency (<0.25 seconds). This study can provide a technical reference for future on-orbit denoising of spaceborne photon-counting LiDAR data.
With the rapid changes in remote sensing platforms, there is a noticeable exponential increase in the quantity of remote sensing images. Choosing the appropriate remote sensing images from extensive remote sensing big data is now a fundamental challenge in remote sensing applications. Currently, utilizing deep Convolutional Neural Networks (CNNs) for extracting deep features from images has become the main approach for remote sensing image retrieval due to its effectiveness. However, the high feature dimensions pose challenges for similarity measurement in the image retrieval, resulting in decreased processing speed and retrieval accuracy. The hash method maps images into compact binary codes from a high-dimensional space, which can be used in remote sensing image retrieval to efficiently reduce feature dimensions. Therefore, this paper proposes a ResNet-based adaptive dilated and structural embedding asymmetric hashing algorithm for the remote sensing image retrieval. Firstly, an adaptive dilated convolution module is designed to adaptively capture multi-scale features of remote sensing images without introducing additional model parameters. Secondly, to address the issue of insufficient extraction of structural information in remote sensing imagery, the current structural embedding module has been optimized and improved to effectively extract geometric structure features from remote sensing images. Lastly, to tackle the problem of low retrieval efficiency caused by intra-class differences and inter-class similarities, pairwise similarity-based constraints are introduced to preserve the similarity of remote sensing images in both the original feature space and the hash space. Experimental comparisons with four datasets (i.e. UCM, NWPU, AID, and PatternNet) were conducted to demonstrate the effectiveness of the proposed method. The mean average precision rates for 64-bit hash codes were 98.07%, 93.65%, 97.92%, and 97.53% with these four datasets, respectively, proving the superiority of our proposed approach over other existing deep hashing image retrieval methods. In addition, four ablation experiments were carried out to verify each module of the proposed method. The ablation experimental results showed that the mean average precision rate was 68.9% by only using the ResNet18 backbone network. The rate will rise to 81.71% after introducing the structural self-similarity coding module, indicating an improvement of 12.81%. Meanwhile, introducing the adaptive dilated convolution module increased the average precision rate by 10.53%. The additional implementation of the pairwise similarity constraints module further increased the average precision rate to 98.07%, indicating a rise of 5.83%. In summary, the experimental results confirm the efficiency of the proposed network framework, which can improve the retrieval accuracy of remote sensing images while maintaining the advantages of deep hashing features.
Accurate and effective extraction of water body information is crucial for water resource management, monitoring, and application. The diversity in the shape, size, and distribution of water bodies, coupled with the complexity of scenes, poses challenges in efficiently and accurately extracting water bodies from remote sensing images. Existing active contour model algorithms for water extraction are primarily tailored for specific data types or water body types and are significantly affected by noise, often resulting in unclear segmentation boundaries and low accuracy in water extraction. In response to these issues, this paper proposes a rapid segmentation method using the Chan-Vese (CV) model that integrates both local and global features of the target. The energy functional of this improved method comprises global, local, and regularization terms. By incorporating local image information into the CV model's energy functional and introducing convolution operators in the local term to compute the mean grayscale difference between the interior and exterior of the evolution curve, using difference images instead of the original images effectively limits erroneous movements during the processing of uneven grayscale images. Additionally, the regularization term consists of a length constraint and a new penalty energy. The length constraint effectively limits the evolution curve's length, preventing excessive boundary gradients and resulting in smoother and more precise target boundaries. The penalty energy avoids the re-initialization steps common in traditional level set methods, enhancing efficiency. This paper utilizes complex land background images from Sentinel-1 and Sentinel-2 remote sensing satellites to validate the practicality of the proposed algorithm. Experiments on the segmentation of lakes, rivers, and small water bodies in remote sensing images show that for SAR (Synthetic Aperture Radar) images, the improved CV model achieves segmentation accuracies of 96.15%, 95.19%, and 83.64% with F1 scores of 95.77%, 91.06%, and 75.78%, respectively. For optical images, the accuracies are 97.71%, 95.12%, and 93.97%, with F1 scores of 97.15%, 93.67%, and 86.78%, respectively. In urban central areas, the SAR data segmentation accuracy and F1 score are 97.2% and 89.2%, respectively; the optical data accuracy and F1 scores are 92.12% and 89.37%. The improved algorithm demonstrates high segmentation accuracy for complex, multi-type water bodies and urban water bodies, achieving high-precision water body extraction in remote sensing images, thus proving highly practical.
Wetland plant species diversity, as a quantifiable indicator reflecting the level of organization in an ecosystem’s community, can reveal the community organization and stability of wetland ecosystems. Accurate assessments of wetland health, degradation, and restoration status are crucial for effective wetland management and protection. Therefore, timely understanding of the current status of wetland plant community species diversity is of great importance. However, traditional field survey methods are time-consuming and labor-intensive, limited by temporal costs, and cannot achieve large-scale synchronous observation. Meanwhile, hyperspectral technology, with its high resolution, can capture more abundant spectral information, providing an opportunity for the realization of this goal. To investigate how to accurately invert wetland plant species diversity using hyperspectral technology, we investigated the wetland plants in Hanzhong Crested Ibis National Nature Reserve in Shaanxi Province and simultaneously acquired hyperspectral images of the plant canopy. Species diversity was characterized by four indicators: Simpson (DS), Margalef (DM), Shannon-Weiner (H'), and Pielou (J). The inverse model was established using three methods: Random Forest (RF), Back Propagation Neural Network (BPNN), and Partial Least Squares (PLS). Finally, the inverse projection of regional species diversity was realized. The outcomes indicate that spectral differentiation complicates the association between spectra and species diversity indices, producing a range of sensitive bands. Notably, the first-order differential transform is superior in extracting sensitive bands compared to the second-order differential transform. Furthermore, correlating species diversity indices can be enhanced through the integration of vegetation indices from various bands. When applying the RF model to analyze differential spectra and vegetation indices, it was found that both using original features and combinations of features, the model's inversion results demonstrated similar and high accuracy (R2 > 0.40). Particularly, in predicting H' and J, the model exhibited strong precision (R2 > 0.6), and in terms of DS, R2 also exceeded 0.5, indicating potential predictive capabilities. However, in reverting another measurement of DM, the model showed lower accuracy (R2 < 0.5), suggesting challenges in improving the model's predictive power. This study demonstrates the effectiveness of UAV hyperspectral technology in the accurate inversion of wetland plant species diversity and confirms the reliability of the method for species diversity inversion at the Unmanned Aerial Vehicle (UAV) scale, achieved through spectral differential transformation and feature variable extraction combined with the random forest model. This technique can provide technical support for the large-scale detection of wetland biodiversity and offer references for decision-making by relevant management departments.
The accurate extraction of street trees is of great significance for the construction of ecological garden cities and the promotion of urban intelligence. However, in vehicle LiDAR point cloud data, there are many situations where the street tree and the adjacent ground objects, such as buildings or other vegetation, are occluded or connected to each other. These situations make it difficult to accurately extract street trees. To solve the above problems, a hierarchical dynamic region-growing method for street tree extraction was proposed in this paper. Firstly, the ground points were filtered out using the point cloud rasterization method, and street trees were preliminarily extracted according to the projection features of ground objects. Secondly, the point cloud data was stratified with equal height intervals based on the distribution characteristics of ground objects. A hierarchical point cloud space was constructed to further obtain information about street trees and interfering objects. Then, the dynamic region-growing operation was carried out in the hierarchical point cloud space, and the attribute information of the point cloud between the same layer and the adjacent layer was obtained. Based on the obtained attribute information, point cloud clusters were generated to distinguish street trees from interfering objects. Finally, the interfering objects were filtered out to achieve accurate street tree extraction using the geometric features of the interfering objects and the rod-like features of the street trees. In this paper, competition data provided by LiDAR conference and point cloud data provided by Open DataLab website were selected for our experiment. The street data of Lille and Paris from the Open DataLab website were mainly used. The experimental results demonstrate that the accuracy and completeness of the proposed method for extracting street trees are above 98.69% and 97.73%, respectively. The proposed method can achieve accurate and complete street tree extraction even when street trees and adjacent ground objects are occluded or connected with each other. Overall, the experimental results strongly support the effectiveness and reliability of the proposed method for street tree extraction. Furthermore, the hierarchical dynamic region-growing method proposed in this paper for street tree extraction demonstrates remarkable data applicability. This method exhibits a robust capacity to adapt to different datasets, making it suitable for a wide range of scenarios. Notably, even when the independence of street trees is not immediately evident, the proposed approach can still effectively extract them.
The use of Unmanned Aerial Vehicles (UAVs) for road image collection is advantageous owing to their large scope and cost-effectiveness. However, the size and shape of road damages vary significantly, making them challenging to predict. Furthermore, due to the limitations of computational resources, generalized target detection algorithms are only applicable to small-size images (512 pixels× 512 pixels or 640 pixels× 640 pixels). This makes them unsuitable for direct application to large-size UAV images (5 472 pixels× 3 648 pixels or 7 952 pixels × 5 304 pixels). The utilization of traditional methods for the detection of multi-scale targets in large-size images is associated with a number of issues, including the slicing of large-size targets and the failure to detect small-size targets. To address these challenges, this paper presents an innovative solution that combines the global-local multiscale fusion strategy with YOLOv5-RDD. First, a YOLOv5-RDD model is constructed, and based on the existing YOLOv5 model, a multiscale C3 (MSC3) module and a Contextual Feature Pyramid Network (CFPN) are designed to improve the detection capability of multiscale targets. Additionally, we introduce an extra detection head for larger-size targets. Then, a global-local multiscale fusion strategy is proposed, which uses resizing and slicing means to obtain global and local information of large UAV images, and then superimposes the global and local multiscale information to obtain the multi-scale information of the whole large image. The detection results are optimized using the center non-maximum value suppression algorithm. Specifically, the global-local multiscale fusion strategy first trains the YOLOv5-RDD using multiscale training strategy to learn complete multiscale features. Then, YOLOv5-RDD predicts multiscale road damages in large-size images using a multiscale prediction strategy to avoid directly applying it to these images. Finally, we use center non-maximum suppression to eliminate redundant object detection boxes. To verify the effectiveness of the proposed method and meet real-world requirements, a UAV-RDD dataset specialized for UAV road disease detection is created. The experimental results show that compared with the original YOLOv5 model, the new model YOLOv5-RDD improves the mAP by 5.8%, while the global-local multiscale fusion strategy improves the mAP by 9.73% compared with the traditional method. The MSC3 achieves the maximum enhancement of mAP@0.5, with an improvement of 2.6%, contributing only 0.8 M parameters. The CFPN yields an improvement of 0.2% in mAP@0.5 while reducing the number of parameters by 8 M. These results fully prove the effectiveness and superiority of the method in this paper.
