Road network ecological bearing capacity estimation method and system based on multi-source remote sensing data

Abstract

The invention provides a road network ecological bearing capacity estimation method and system based on multi-source remote sensing data, and relates to the technical field of remote sensing technology and ecological evaluation. The method comprises the following steps: acquiring a multispectral remote sensing image and ecological system basic data for preprocessing; supervising and classifying the remote sensing image to extract linear ground feature pattern spots and construct a topological relation to obtain road network topological structure data; associating the historical traffic flow statistical data with the road network space and performing interpolation to generate day and night noise spatial and temporal distribution data; extracting habitat distribution data based on the habitat type thematic map and the historical field survey point of the species, and obtaining acoustic sensitive threshold data of the species in combination with behavior experiment records; according to the method, noise spatial-temporal distribution data and habitat distribution data are superposed, noise data in a habitat range are screened out, then the noise data are compared with an acoustic sensitive threshold pixel by pixel, the total area of pixels exceeding the threshold is counted to serve as a road network ecological interference index, and accurate estimation of road network ecological interference is achieved.

Description

Technical Field

[0001] This invention relates to the fields of remote sensing technology and ecological assessment technology, and in particular to a method and system for estimating the ecological carrying capacity of road networks based on multi-source remote sensing data. Background Technology

[0002] Against the backdrop of coordinated development between transportation infrastructure planning and ecological protection, the pressure exerted by road network construction on the regional ecological environment is receiving increasing attention. Ecological carrying capacity, as an indicator measuring the ability of a regional ecosystem to withstand human disturbance, is crucial for accurate estimation in the rational layout of road networks and the formulation of ecological mitigation measures. However, existing ecological carrying capacity assessment methods largely rely on statistical data and field surveys, making it difficult to finely characterize the dynamic disturbances of road network operation to the ecosystem on a large spatial scale. In particular, noise pollution generated by road networks is influenced by multiple factors such as traffic flow and topography, and traditional methods lack efficient and accurate estimation tools. Furthermore, different species exhibit varying sensitivities to noise, and existing assessment systems often ignore this biological characteristic, using a uniform noise threshold for evaluation, leading to discrepancies between estimated results and actual ecological impacts.

[0003] Therefore, how to comprehensively utilize multi-source data to achieve high-precision and targeted estimation of the ecological carrying capacity of the road network is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0004] To address the aforementioned problems in existing technologies, the first aspect of this invention proposes a method for estimating the ecological carrying capacity of road networks based on multi-source remote sensing data, comprising: S1: Acquire multispectral remote sensing images of the monitoring area and basic ecosystem data of the monitoring area; perform geometric correction and radiometric calibration on the multispectral remote sensing images to obtain preprocessed remote sensing images; and format the basic ecosystem data to obtain a standardized basic dataset. S2: Supervised classification of road network elements is performed on the preprocessed remote sensing image, linear land cover patches are extracted from the image, and topological relationships are constructed on the linear land cover patches to obtain road network topology data of the monitoring area. S3: Obtain historical traffic flow statistics for each road segment in the road network topology data, spatially correlate the historical traffic flow statistics with each road segment, and interpolate the correlated data based on the preset time slices to generate noise spatiotemporal distribution data of the road network in the day and night time dimension. S4: Obtain thematic maps of habitat types in the monitoring area and historical field survey data of target species in the monitoring area. Perform spatial overlay analysis on the historical field survey data and habitat type thematic maps to extract the spatial range of the target species' habitat and obtain the habitat distribution data of the target species. At the same time, obtain behavioral experimental records of the target species under different frequency noise stimuli and extract the critical noise intensity value when the target species exhibits avoidance behavior from the behavioral experimental records to obtain the acoustic sensitivity threshold data of the target species. S5: Spatially overlay the spatiotemporal distribution data of noise with the habitat distribution data of the target species, filter out the noise intensity data within the habitat distribution data range, and compare the filtered noise intensity data with the acoustic sensitivity threshold data of the target species pixel by pixel. Calculate the total area of ​​pixels within the habitat distribution data range where the noise intensity data exceeds the acoustic sensitivity threshold data, and obtain the road network ecological disturbance index used to characterize the degree of road network interference with the activities of the target species.

[0005] Secondly, this invention proposes a road network ecological carrying capacity estimation system based on multi-source remote sensing data. The system adopts a road network ecological carrying capacity estimation method based on multi-source remote sensing data proposed in any of the above embodiments, and the system includes: The data acquisition and preprocessing module is used to perform step S1: acquire multispectral remote sensing images of the monitoring area and basic ecosystem data of the monitoring area, perform geometric correction and radiometric calibration on the multispectral remote sensing images to obtain preprocessed remote sensing images, and perform formatting processing on the basic ecosystem data to obtain a standardized basic dataset. The road network topology extraction module is used to perform step S2: supervised classification of road network elements in the preprocessed remote sensing image, extract linear ground feature patches in the image, and construct the topological relationship of the linear ground feature patches to obtain the road network topology data of the monitoring area. The noise spatiotemporal distribution simulation module is used to perform step S3: obtain historical traffic flow statistics data of each road segment in the road network topology data, correlate the historical traffic flow statistics data with each road segment in terms of spatial location, and perform interpolation calculation on the correlated data according to the preset time slice to generate noise spatiotemporal distribution data of the road network in the day and night time dimension. The species sensitivity information acquisition module is used to perform step S4: acquire the thematic map of habitat types in the monitoring area and historical field survey data of the target species in the monitoring area, perform spatial overlay analysis of the historical field survey data and the thematic map of habitat types, extract the spatial range of the target species' habitat, obtain the habitat distribution data of the target species, and acquire behavioral experimental records of the target species under different frequency noise stimuli, extract the critical noise intensity value when the target species exhibits avoidance behavior from the behavioral experimental records, and obtain the acoustic sensitivity threshold data of the target species. The road network ecological disturbance index calculation module is used to perform step S5: spatially overlaying the spatiotemporal distribution data of noise with the habitat distribution data of the target species, filtering out the noise intensity data within the range of the habitat distribution data, comparing the filtered noise intensity data with the acoustic sensitivity threshold data of the target species pixel by pixel, and calculating the total area of ​​pixels within the habitat distribution data where the noise intensity data exceeds the acoustic sensitivity threshold data, thereby obtaining the road network ecological disturbance index used to characterize the degree of disturbance of the road network to the activities of the target species.

[0006] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention constructs a complete solution from data preprocessing to ecological disturbance index generation by organically combining multi-source remote sensing data and ecological data, which significantly improves the scientificity and accuracy of road network ecological carrying capacity estimation.

[0007] Specifically, step S1 eliminates geometric distortion and radiometric distortion generated during image acquisition by performing geometric correction and radiometric calibration on multispectral remote sensing images, laying a data foundation for the accurate extraction of subsequent road network information. Simultaneously, formatting of the ecosystem data ensures the compatibility and consistency of multi-source data in subsequent analysis. Step S2, based on the preprocessed remote sensing images, transforms visual linear land cover patches into spatially correlated road network topology data through supervised classification and topology construction, achieving a digital representation of the road network's spatial morphology and providing an accurate spatial carrier for noise simulation. Step S3 innovatively correlates historical traffic flow statistics with the spatial location of the road network and combines this with interpolation calculations in the time dimension to generate noise spatiotemporal distribution data in the day-night time dimension. This process not only considers the spatiotemporal heterogeneity of traffic flow but also quantifies noise, a dynamic interference factor, in both time and space, overcoming the limitations of traditional static assessments.

[0008] Building upon this foundation, step S4 incorporates the biological characteristics of the target species. Through spatial overlay analysis, the actual habitat distribution range of the species is extracted, and behavioral experimental records are combined to obtain the critical noise intensity value at which the species exhibits avoidance behavior—the acoustic sensitivity threshold data. This step refines the assessment object from the macro-ecosystem to specific sensitive species, fully considering the differentiated responses of different species to noise disturbance, making the assessment results more ecologically significant. Finally, step S5 spatially overlays the simulated spatiotemporal noise distribution data with the species habitat distribution data, filters out noise information within the habitat range, and accurately calculates the interference area exceeding the species' acoustic sensitivity threshold through pixel-by-pixel comparison, thereby obtaining the road network ecological disturbance index.

[0009] In summary, this method employs a series of interconnected steps. It interprets the spatial structure of the road network using remote sensing technology, simulates the dynamic distribution of noise using traffic flow data, and then combines this with species-specific sensitivity thresholds for spatial analysis. Ultimately, it quantifies the abstract ecological impact into a specific disturbance area index. The entire process fully leverages the macroscopic nature of multi-source remote sensing data, the dynamic nature of traffic statistics, and the precision of ecological experimental data. This enables accurate estimation of road network ecological disturbances from a broad to a specific level, and from general to targeted approaches, providing strong data support for road network planning and ecological protection decision-making. Attached Figure Description

[0010] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0011] Figure 1 The diagram shown is a flowchart illustrating a method for estimating the ecological carrying capacity of a road network based on multi-source remote sensing data, according to an embodiment of the present invention. Figure 2 The diagram shown is a schematic representation of a road network ecological carrying capacity estimation system based on multi-source remote sensing data, according to an embodiment of the present invention. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0013] The specific embodiments of the present invention will be described below.

[0014] Example 1 like Figure 1 As shown, the first aspect of this invention proposes a method for estimating the ecological carrying capacity of road networks based on multi-source remote sensing data, comprising: S1: Acquire multispectral remote sensing images of the monitoring area and basic ecosystem data of the monitoring area; perform geometric correction and radiometric calibration on the multispectral remote sensing images to obtain preprocessed remote sensing images; and format the basic ecosystem data to obtain a standardized basic dataset. S2: Supervised classification of road network elements is performed on the preprocessed remote sensing image, linear land cover patches are extracted from the image, and topological relationships are constructed on the linear land cover patches to obtain road network topology data of the monitoring area. S3: Obtain historical traffic flow statistics for each road segment in the road network topology data, spatially correlate the historical traffic flow statistics with each road segment, and interpolate the correlated data based on the preset time slices to generate noise spatiotemporal distribution data of the road network in the day and night time dimension. S4: Obtain thematic maps of habitat types in the monitoring area and historical field survey data of target species in the monitoring area. Perform spatial overlay analysis on the historical field survey data and habitat type thematic maps to extract the spatial range of the target species' habitat and obtain the habitat distribution data of the target species. At the same time, obtain behavioral experimental records of the target species under different frequency noise stimuli and extract the critical noise intensity value when the target species exhibits avoidance behavior from the behavioral experimental records to obtain the acoustic sensitivity threshold data of the target species. S5: Spatially overlay the spatiotemporal distribution data of noise with the habitat distribution data of the target species, filter out the noise intensity data within the habitat distribution data range, and compare the filtered noise intensity data with the acoustic sensitivity threshold data of the target species pixel by pixel. Calculate the total area of ​​pixels within the habitat distribution data range where the noise intensity data exceeds the acoustic sensitivity threshold data, and obtain the road network ecological disturbance index used to characterize the degree of road network interference with the activities of the target species.

[0015] This embodiment provides a method for estimating the ecological carrying capacity of road networks based on multi-source remote sensing data. The implementation process begins with data acquisition and preprocessing. In step S1, multispectral remote sensing images covering the monitoring area need to be acquired. Multispectral remote sensing images refer to remote sensing images containing multiple spectral bands (e.g., red, green, blue, near-infrared, etc.). Different land features have unique reflectance characteristics in different bands, making it possible to identify and distinguish road network elements using spectral information. Simultaneously, basic ecosystem data for the monitoring area also needs to be collected. Basic ecosystem data refers to various types of data describing the ecological environment of the area, such as land use status data, vegetation type distribution data, water system network data, and soil type data. This data can be obtained from existing national geographic census results, forestry survey data, or records from environmental monitoring stations. For the acquired multispectral remote sensing images, geometric correction is required. Geometric correction aims to eliminate geometric deformations caused by sensor attitude changes, terrain undulations, and Earth curvature during image acquisition, ensuring that each pixel in the image accurately corresponds to a specific location in the actual geographic coordinate system. The correction process typically utilizes ground control points and digital elevation models. Next, radiometric calibration is performed. Radiometric calibration is the process of converting the raw digital values ​​recorded by the sensor into physically meaningful radiance values ​​or surface reflectance. This eliminates the influence of atmospheric conditions, solar altitude angle, and differences in sensor performance on image brightness, making images acquired at different times and by different sensors comparable. This results in preprocessed remote sensing images that can be used for quantitative analysis. For ecosystem baseline data, since its sources may include various types such as vector-format thematic maps and tabular statistical data, formatting is required. Formatting involves converting these diverse data sources into a unified data format, coordinate system, and attribute structure. For example, all vector data can be standardized into a common shapefile format and a unified projected coordinate system, thus obtaining a standardized baseline dataset. This preprocessing provides an accurate and consistent data foundation for all subsequent spatial analyses, avoiding cumulative errors caused by raw data quality issues and ensuring the reliability of road network extraction and ecological analysis results.

[0016] After data preprocessing, step S2 is executed to extract road network topology data of the monitored area. This step is achieved through supervised classification of the preprocessed remote sensing images. Supervised classification is a sample-based classification method that requires the operator to select representative road network sample areas on the image as training samples, such as clearly visible local areas of highways, national roads, or urban roads. Classification algorithms (such as maximum likelihood, support vector machines, or random forests) will establish discrimination rules based on the spectral characteristics of these samples, and then classify each pixel in the image into a road or non-road category, thereby segmenting linear feature patches composed of pixels from the image. These patches initially reflect the spatial distribution of roads. However, the patches obtained from supervised classification are often discrete areal regions. These discrete patches alone are insufficient to describe the connectivity of the road network as a system. Therefore, it is necessary to construct topological relationships for these linear feature patches. Topological relationship construction refers to establishing the adjacency and connection relationships between patches. For example, identifying which patches belong to the continuous part of the same road, determining the connection mode of road intersections, and the location of road endpoints. Finally, a road network topology data in vector format is generated, consisting of nodes (representing road intersections or road endpoints) and arcs (representing road segments). This structured data not only contains the precise geographical location of the roads but also fully expresses the spatial association between roads through the node-arc topological relationship. This provides an indispensable spatial framework for accurately linking traffic flow data to specific road segments and simulating noise propagation along the road network. By combining supervised classification with topology construction, high-precision road network information can be efficiently obtained from remote sensing images, avoiding the heavy workload of manual field measurement or screen digitization, and ensuring the timeliness of road network data, so that the evaluation results can reflect the latest road network construction status.

[0017] Next, step S3 is executed to generate noise spatiotemporal distribution data of the road network in the day-night time dimension. After obtaining the road network topology data, it is necessary to obtain historical traffic flow statistics for each road segment. Historical traffic flow statistics refer to data that reflects the number of vehicles traveling on the road and the time-varying patterns, recorded over a long period by traffic monitoring departments through equipment such as underground induction coils, radars, or cameras installed on the roadside. This data is usually stored in time series form, for example, recording two-way traffic flow every hour or every five minutes. These historical traffic flow statistics are then spatially correlated with the road network topology data. That is, based on the geographical identifiers or location information of the road segments, the records in the flow data table are matched to the corresponding arc segments in the road network topology, so that each road segment arc segment is accompanied by a set of data reflecting its traffic load changes over time. Given the significant temporal fluctuations in traffic flow, such as the different flow patterns on weekdays and weekends, and the substantial differences between daytime and nighttime flow, it is necessary to interpolate the correlated data based on preset time slices. These time slices can be set according to the needs of the ecological assessment; for example, a day can be divided into daytime and nighttime slices. For any missing periods in the monitoring data, linear interpolation or trend interpolation based on historical data from the same period can be used to fill in the gaps, thereby generating representative flow values ​​for each road segment in each time slice. Then, this flow data is combined with a noise propagation model to simulate the spatial distribution of noise intensity generated by the road network at different times of day and night. The key to this step is the fusion of static road network spatial data with dynamic traffic flow temporal data. This allows the final generated spatiotemporal noise distribution data to accurately reflect the temporal and spatial variations in noise interference caused by road network operations to the surrounding environment. For example, the noise impact range expands during busy daytime traffic and significantly shrinks at night. This dynamic characteristic cannot be captured by traditional static noise assessments and provides high-resolution spatiotemporal data support for subsequent assessments of the potential impact of noise on the diurnal activity rhythms of species.

[0018] Next, step S4 is executed to obtain habitat distribution data and acoustic sensitivity threshold data for the target species. First, a habitat type thematic map of the monitoring area needs to be obtained. This thematic map, based on land cover or vegetation type, reflects the suitability of potential habitats for different species. It can be obtained by using object-oriented image classification methods on high-resolution satellite remote sensing images. Classification categories typically include forests, shrublands, grasslands, wetlands, water bodies, and farmland, each corresponding to different ecological functions. Simultaneously, historical field survey data for the target species within the monitoring area also needs to be obtained. The target species refers to wildlife species that require close monitoring due to road network influences, such as ungulates, carnivores, or rare and endangered species. Historical field survey data can be derived from years of tracking and monitoring records from research institutions, such as GPS collars used to locate animals, or animal traces, droppings, or direct observation points recorded through transect surveys. These historical field survey data points are spatially overlaid with habitat type thematic maps in the form of vector point layers. This involves utilizing the spatial connectivity function of a Geographic Information System (GIS) to extract the habitat patches located at each survey point. Then, the habitat patches covered by all survey points are merged to delineate the actual habitat space used by the target species, obtaining habitat distribution data. This process defines habitats based on actual species occurrence records, which is more consistent with the species' ecological learning and actual spatial utilization patterns compared to relying solely on theoretical inferences from habitat type maps. On the other hand, it is also necessary to obtain behavioral experimental records of the target species under different frequency noise stimuli. These records typically come from bioacoustic laboratories or controlled field experiments. The experimental design involves playing pre-recorded or synthesized traffic noise of different frequencies and intensities in a quiet environment, while simultaneously using video equipment or behavioral observation devices to record the behavioral responses of individual target species, such as whether they exhibit avoidance behaviors such as raising their heads in alarm, stopping foraging, or fleeing the sound source area. From these behavioral experimental records, statistical analysis is used to determine the lowest noise intensity value that causes 50% or more individuals to exhibit avoidance behavior; this is the acoustic sensitivity threshold data for the species. This threshold data serves as a biological benchmark reflecting a species' tolerance to noise interference, providing an objective quantitative standard for subsequent judgments on whether noise constitutes substantial interference.

[0019] Finally, step S5 is executed to calculate the road network ecological disturbance index. The spatial overlay analysis is performed on the noise spatiotemporal distribution data generated in step S3 and the habitat distribution data obtained in step S4. Specifically, in the geographic information system software, the noise raster layer and the habitat vector layer are overlaid. Using the mask extraction tool, with the habitat vector layer as the mask area, all pixels exactly inside the habitat boundary are extracted from the noise raster, obtaining the noise intensity raster data within the habitat. Then, this noise intensity data within the habitat is compared pixel-by-pixel with the acoustic sensitivity threshold data of the target species. That is, for each raster pixel within the habitat, a conditional statement is executed: if the noise intensity value of the pixel is greater than the critical noise intensity value of the species, the pixel is marked as a disturbing pixel and assigned a value of 1; otherwise, it is marked as a non-disturbing pixel and assigned a value of 0. The total number of disturbing pixels assigned a value of 1 is counted, and this total number of pixels is multiplied by the actual geographical area represented by a single pixel to calculate the noise disturbance area within the habitat distribution data range. This area is the road network ecological disturbance index, which quantitatively characterizes the spatial extent of actual disturbance caused by road network noise within the target species' habitat using a straightforward area value. It directly reflects the degree of impact of road network operation on the species' normal activities. Through this complete process from multi-source data acquisition and processing to professional model analysis and final index generation, this invention achieves refined, species-specific, and spatially explicit estimation of road network ecological carrying capacity, providing scientific data support for route avoidance, ecological compensation measure design, and environmental impact assessment in transportation planning. This method organically combines remote sensing technology, geographic information system spatial analysis, traffic engineering, and animal behavior, with each step interconnected and logically progressive. The data processing results of the previous stage are direct inputs for the analysis of the next stage, ensuring the accuracy and ecological significance of the final estimation results.

[0020] In some implementations, S2 includes: S21: Perform multi-scale segmentation on the preprocessed remote sensing image to generate multiple image objects composed of homogeneous pixels, and obtain a candidate road network patch object set; S22: Perform geometric feature calculations on each object in the candidate road network patch object set to obtain the aspect ratio, shape index, and area parameters of each object, and obtain the geometric morphological feature set of the candidate patches; S23: Compare the aspect ratio of the geometric feature set with the preset aspect ratio threshold of linear features, compare the shape index with the preset road smoothness threshold, and filter out objects that simultaneously meet the aspect ratio threshold of linear features and the road smoothness threshold to obtain the target road network patch object set. S24: Extract centerlines from objects in the target road network patch object set, and capture endpoints and construct connection relationships for the extracted centerlines to generate road network topology data composed of nodes and arc segments.

[0021] This embodiment further defines the specific method for extracting road network topology data in step S2. During implementation, when executing step S21, the preprocessed remote sensing image is first segmented at multiple scales. Multi-scale segmentation is an object-oriented image analysis technique that, based on the spectral and shape heterogeneity of image pixels and according to operator-preset scale parameters, shape factors, and compactness factors, divides the image into multiple homogeneous regions, which are called image objects. The segmentation scale determines the size of the generated objects; the larger the scale parameter, the larger the area of ​​the generated objects and the more pixels they contain. For road network extraction, it is necessary to select appropriate scale parameters so that each road segment can be completely segmented into one or more objects, while avoiding merging roads with other adjacent land features (such as buildings or bare land) into the same object. The resulting candidate road network patch object set contains a large number of image objects that may be road networks, but also includes other linear land features with similar shapes, such as rivers, ditches, field ridges, or ridgelines, as well as some non-linear land features, such as residential areas and ponds. Therefore, in step S22, it is necessary to calculate the geometric features of each object in the candidate road network patch object set to quantify its morphological characteristics. Specifically, the aspect ratio, shape index, and area parameter of each object are calculated. The aspect ratio is the ratio of the longer side to the shorter side of the object's smallest bounding rectangle, reflecting the object's elongation. Roads, as typical linear features, usually have a much larger aspect ratio than areal features. The shape index is usually defined as the ratio of the square of the object's perimeter to its area. It is a dimensionless index that reflects the complexity or smoothness of the object's boundary. Regular and smooth objects (such as artificial roads) have a smaller shape index, while natural linear features (such as streams) often have more tortuous boundaries and a relatively larger shape index. The area parameter helps to exclude excessively small noise patches (which may be misclassified) or excessively large non-road features (such as plazas or large building roofs). By calculating these parameters, the geometric morphological feature set of each object can be obtained, providing data support for subsequent screening.

[0022] Next, in step S23, the parameters in the geometric feature set are compared and filtered using preset thresholds. Specifically, the aspect ratio of each object is compared with a preset linear feature aspect ratio threshold. This threshold is an empirical value or a value obtained through statistical training samples, such as 3 or 5. Only objects with an aspect ratio greater than this threshold are considered potential linear features. Simultaneously, the shape index of each object is compared with a preset road smoothness threshold. This threshold is also an empirical threshold, such as a range of values. Only objects with a shape index less than this threshold are considered to have smooth boundaries and conform to road characteristics. Through this dual filtering using these two thresholds, non-road features that are long and narrow but have extremely irregular boundaries (such as jagged dried-up riverbeds) and other features that are smooth but have insufficient aspect ratios (such as circular farmland or square buildings) can be effectively removed from the candidate object set. Finally, objects that simultaneously meet both the linear feature aspect ratio threshold and the road smoothness threshold are selected, resulting in the target road network patch object set. This screening process significantly improves the purity and accuracy of road network extraction, reduces the possibility of misidentifying non-road features as roads, and ensures that subsequent analysis is based on accurate road network information. After obtaining accurate road network patches, step S24 is performed to extract centerlines from these patches. Centerline extraction can employ a morphological skeletonization algorithm, which repeatedly peels away pixels from the boundaries of planar patches until only a single-pixel-width skeleton line remains, or a vectorized centerline extraction tool can be used to convert planar polygons into linear features. Then, endpoint capture and connection relationship construction are performed on the extracted centerlines. Endpoint capture refers to automatically moving and connecting the endpoints of centerlines from different directions that should be connected but have slight gaps due to digitization errors at road intersections, ensuring uninterrupted connectivity of the road network. Ultimately, a road network topology data is generated, consisting of nodes (representing road intersections or the start and end points of roads) and arcs (representing continuous road segments between nodes). This data format not only stores the spatial coordinates of roads but also records, in the form of a topology table, the arcs connected to each node, the start and end nodes of each arc, and the adjacency relationships between arcs. This provides a crucial spatial data foundation for accurately attaching traffic flow data to each road segment and simulating noise propagation along the road network. Through the progressive refinement of this embodiment, the road network extraction process becomes more rigorous and accurate, significantly improving the efficiency and quality of the conversion from raw remote sensing images to structured road network data. The resulting topology data serves as the core support for all subsequent road network-related analyses.

[0023] In some implementations, S22 includes: S221: For each image object in the candidate road network patch object set, the edge detection operator is used to calculate the gradient, identify the edge points where the pixel values ​​inside the object change abruptly, and connect the edge points into a closed boundary line according to the eight-neighbor tracking algorithm to obtain the contour line data of each image object. S222: Based on the contour data of each image object, count the total number of pixels contained in the area enclosed by the contour lines, and multiply the total number of pixels by the actual ground area represented by a single pixel to obtain the area parameter of the image object. S223: Based on the contour data and area parameters of each image object, measure the total length of the contour as the perimeter parameter, and use the ratio of the square of the perimeter parameter to the area parameter as the shape index of the image object; S224: Based on the contour data of each image object, the minimum bounding rectangle algorithm is used to fit the region enclosed by the contour data, calculate the length of the long side and the length of the short side of the minimum rectangle that completely contains the contour, and use the ratio of the length of the long side to the length of the short side as the aspect ratio of the image object. S225: Associate the aspect ratio, shape index, and area parameter as attribute fields to generate a feature record for each image object containing the aspect ratio, shape index, and area parameter. After summarizing the feature records of all image objects, a geometric shape feature set is obtained.

[0024] This embodiment further specifies the geometric feature calculation step S22 in detail, clarifying how to extract specific geometric parameters from image objects. When implementing step S221, for each image object in the candidate road network patch object set, an edge detection operator is first used to calculate the gradient. Edge detection operators (such as the Canny operator, Sobel operator, or Laplacian operator) can identify edge points where pixel values ​​inside the object undergo a step change by performing first or second derivative operations on the brightness value of the image object. These edge points typically correspond to the boundary between the feature outline and the background. After obtaining the edge points, these discrete edge points need to be connected into a closed boundary line according to the eight-neighborhood tracking algorithm. The principle of the eight-neighborhood tracking algorithm is to start from any edge point and search for other edge points in its eight neighboring directions (up, down, left, right, and four diagonal directions). If they exist, they are included in the contour line, and the search continues until returning to the starting point or no new neighboring points are found, thus forming a continuous and closed contour line. This contour line is the contour line data for each image object. This process transforms abstract image objects, composed of sets of pixels, into geometric figures with well-defined boundaries that can be mathematically described. After obtaining the contour data, step S222 is executed to calculate the area parameter. Based on the contour data of each image object, the total number of pixels contained within the area enclosed by the contour is counted, i.e., all pixels within the closed area are counted. This can be achieved through the area statistics function of the geographic information system software or the raster calculator. Then, the total number of pixels is multiplied by the actual ground area represented by a single pixel. This actual ground area depends on the spatial resolution of the remote sensing image. For example, for an image with a spatial resolution of meters, a single pixel represents one square meter on the ground; if the resolution is 0.5 meters, then a single pixel represents 0.25 square meters. Through this multiplication operation, the actual area parameter of the image object is obtained, in square meters or hectares. This area is the basis for subsequent calculations of the shape index and for determining the scale of road network patches and whether they are components of roads.

[0025] Given the area parameter, step S223 calculates the shape index. First, the total length of the contour line is measured as the perimeter parameter, which can be obtained by calculating the sum of the Euclidean distances between all adjacent pixels on the contour line. For vectorized contour lines, their length attribute can be directly read. Then, the ratio of the square of the perimeter parameter to the area parameter is taken as the shape index of the image object. The shape index is a dimensionless value that describes the complexity of the object's boundary: for a shape of the same area, the longer the perimeter, the larger the shape index, meaning the more irregular and fragmented the boundary; while regular shapes (such as circles, squares, or smooth rectangles) have smaller shape indices. For roads, due to artificial construction, their boundaries are usually relatively straight and smooth, so the shape index is small. The boundaries of natural features are often tortuous due to factors such as terrain and vegetation, resulting in a larger shape index. Therefore, the shape index is an important feature for distinguishing between artificial roads and natural linear features. On the other hand, step S224 calculates the aspect ratio. Based on the contour line data of each image object, the minimum bounding rectangle algorithm is used for fitting. The minimum bounding rectangle algorithm rotates a rectangle until it finds a rectangle that completely contains the outline and has the smallest area. This rectangle's direction is usually aligned with the object's principal axis (i.e., the direction of longest extension). The lengths of the longer and shorter sides of this minimum bounding rectangle are calculated, and the ratio of the longer to shorter side lengths is used as the aspect ratio of the image object. The aspect ratio effectively reflects the elongation or narrowness of an object. Roads, as typical linear features, typically have a much larger aspect ratio than other areal features; for example, the aspect ratio of buildings is often close to 1. After calculating the aspect ratio, shape index, and area parameter, step S225 associates these parameters with attribute fields. Each image object corresponds to a feature record, which creates an aspect ratio field, a shape index field, and an area parameter field, and fills in the calculated values ​​accordingly. Summarizing the feature records of all image objects constitutes a geometric feature set containing quantitative indicators of each object's geometric shape.

[0026] Through the refined calculations in this embodiment, the geometric features of each candidate road network patch are precisely quantified. These features provide objective and comparable quantitative indicators for subsequent threshold screening, ensuring the scientific nature and repeatability of the screening process. This avoids the subjectivity and uncertainty of relying solely on visual judgment, thereby further improving the accuracy and automation level of road network extraction.

[0027] In some implementations, S3 includes: S31: Obtain historical traffic flow time series data for each road segment released by the traffic monitoring department, classify and statistically analyze the historical traffic flow time series data according to weekdays and rest days, and obtain the basic traffic flow data of the classified road segments. S32: The traffic flow data of the classified road segments are summed separately according to the daytime and nighttime periods to obtain the first average flow value of each road segment during the daytime period and the second average flow value during the nighttime period. S33: Obtain digital elevation model data of the monitoring area, extract surface roughness raster from the digital elevation model data, use the surface roughness raster as the surface attenuation factor for sound wave propagation, and use the first average flow rate and the second average flow rate as the intensity parameters of the line sound source, respectively, input into the sound wave propagation geometric attenuation model for calculation, and generate daytime noise intensity raster and nighttime noise intensity raster. S34: Combine the daytime noise intensity raster map with the nighttime noise intensity raster map by band to generate noise spatiotemporal distribution data containing both day and night time dimensions.

[0028] This embodiment provides a detailed description of the specific implementation of step S3, namely, generating the spatiotemporal distribution data of noise. When implementing step S31, it is necessary to obtain historical traffic flow time-series data for each road segment from traffic monitoring departments. This data typically records the number of standard vehicles passing through the road at fixed time intervals (e.g., every hour or every 15 minutes). The data source can be records from highway toll systems, traffic monitoring station data from urban intelligent transportation systems, or periodic manual traffic surveys. Considering the significant differences in people's travel patterns between weekdays and weekends—for example, weekdays usually have two commuting peaks in the morning and evening with sparse traffic at night, while weekends may have a midday travel peak with a relatively flat traffic distribution throughout the day—it is necessary to classify and statistically analyze this time-series data according to weekdays and weekends. The original data is divided into two datasets based on calendar dates: weekday traffic flow datasets and weekend traffic flow datasets, to obtain a more accurate classification of the road segment traffic flow data reflecting real traffic patterns. After classification, step S32 is executed to divide the traffic flow of each road segment into day and night time periods. The categorized traffic flow data for each road segment is summed separately for daytime and nighttime periods. The specific division of daytime and nighttime periods can be set based on the activity rhythm of the target species or assessment needs; for example, daytime can be defined as 6:00 AM to 6:00 PM, and the remaining time as nighttime. By summing the data and dividing by the number of hours within each period, the average hourly flow rate for each road segment during the daytime can be calculated as the first average flow rate value, and the average hourly flow rate during the nighttime can be calculated as the second average flow rate value. These two flow rate values ​​represent the average noise source intensity of the road segment during the two typical daytime and nighttime periods, respectively. They are important input parameters for subsequent noise simulation, reflecting the differences in traffic load across different time periods and providing a basis for generating a time-dimensional noise distribution.

[0029] Next, step S33 is executed to simulate the spatial propagation of noise. First, digital elevation model (DEM) data of the monitoring area is acquired. DEM data is ground elevation information stored in raster form, with each pixel recording the altitude of that location. This data can be obtained from the Space Shuttle Radar Terrain Mission or stereo imagery. A surface roughness raster map is extracted from the DEM data. Surface roughness can be obtained by calculating the elevation difference or slope variability between each pixel in the DEM and its neighboring pixels. It quantifies the degree of surface undulation and fragmentation. Rough surfaces (such as mountainous areas) scatter, absorb, and block sound waves, thus accelerating noise attenuation. Therefore, the surface roughness raster map is used as the surface attenuation factor for sound wave propagation input into the model. Simultaneously, the previously calculated first and second average flow rates are used as intensity parameters of the line sound source and input into the sound wave propagation geometric attenuation model for calculation. The geometric attenuation model for sound wave propagation is based on the acoustic theory of point or line sound source propagation. For a typical line sound source like a road, the model typically divides the road into several small segments, each considered a point source. Factors such as distance attenuation (sound pressure level decreases logarithmically with increasing distance), air absorption attenuation, ground absorption attenuation, and terrain shielding effects are then considered, and the total sound pressure level at the receiving point is calculated by superposition. Substituting daytime and nighttime flow values ​​into the model, noise intensity raster maps for daytime and nighttime scenarios can be calculated separately, i.e., daytime noise intensity raster map and nighttime noise intensity raster map. These two raster maps reflect the equivalent continuous A-weighted sound level distribution at various locations within the monitoring area during the two typical daytime and nighttime time periods. Finally, step S34 combines the daytime and nighttime noise intensity raster maps by band. This involves using a raster data processing tool to merge the two single-band raster files into a single raster file containing two bands: the first band stores the daytime noise intensity data, and the second band stores the nighttime noise intensity data. In this way, noise spatiotemporal distribution data containing both day and night time dimensions are generated. This multi-band raster data not only depicts the spatial diffusion pattern of noise, but also reveals its dynamic changes on the day and night time scale through different bands. It provides rich temporal dimension information for subsequent assessment of the potential impact of road network noise on species activities at different time periods (such as daytime foraging and nighttime resting), enabling ecological disturbance assessment to more precisely match the behavioral rhythms of species.

[0030] In some implementations, S4 includes: S41: Acquire satellite remote sensing images covering the monitoring area, use an object-based image classification method to classify land cover types in the satellite remote sensing images, and generate habitat type thematic maps containing information on forest, grassland, water, and farmland categories; S42: Obtain the set of spatial locations recorded by GPS collar tracking of the target species in the wildlife monitoring database, clean the set of spatial locations to remove outliers, and obtain historical field survey location data of the target species. S43: The cleaned historical field survey data is overlaid onto the habitat type thematic map in the form of vector point layers. All habitat type patches covered by the historical field survey data are extracted and merged to obtain the habitat distribution data of the target species. S44: Obtain hearing test data recorded in the playback experiment of the target species in the bioacoustics laboratory, analyze the hearing threshold of the target species under different frequency pure tones from the hearing test data, and determine the minimum value of the hearing threshold as the critical noise intensity value for the target species to produce avoidance behavior, thereby obtaining the acoustic sensitivity threshold data of the target species.

[0031] This embodiment further specifies the method for obtaining habitat distribution data and acoustic sensitivity threshold data of the target species in step S4. During implementation, when performing step S41, it is necessary to acquire satellite remote sensing images covering the monitoring area. Satellite remote sensing images refer to images acquired by sensors carried by artificial satellites, such as multispectral data from the Landsat series, Sentinel series, or my country's Gaofen series satellites. These images typically contain multiple bands and can reflect the spectral characteristics of different land features. An object-based image classification method is used to classify land cover types from the satellite remote sensing images. Object-based image classification is a high-resolution remote sensing image analysis technique. It first segments the image into several pixel clusters with similar spectral and texture features; these clusters are called image objects. Then, based on the object's spectral mean, standard deviation, shape index, texture features, and spatial relationships between objects, each object is assigned to a preset category using supervised classification or fuzzy classification rules. This method can generate habitat type thematic maps containing information on categories such as forests, grasslands, water bodies, and farmland. Forests typically refer to areas with tree cover exceeding a certain threshold, characterized by high canopy closure; grasslands are dominated by herbaceous plants and may include steppes and meadows; water bodies include rivers, lakes, reservoirs, and other areas with perennial or seasonal water accumulation; and farmland refers to arable land used for crop cultivation, including paddy fields and dry land. These habitat types are fundamental to the survival of wild animals, and different species exhibit specific selectivity in habitat types. Therefore, habitat type thematic maps provide a spatial framework for subsequent delineation of species habitats.

[0032] After obtaining the habitat type thematic map, step S42 is executed to acquire historical field survey location data for the target species. This historical field survey location data comes from a wildlife monitoring database, which records a set of spatial locations of individual target species obtained by researchers using GPS collars. A GPS collar is a device worn around an animal's neck, containing a built-in GPS receiver. It automatically records the latitude and longitude coordinates of the animal's location at preset time intervals (e.g., several times per hour or per day) and transmits this data back to ground stations via satellite or radio. Over time, this data accumulates to form a set of locations reflecting the animal's activity trajectory and home range. After acquiring this raw data, the spatial location set needs to be cleaned to remove outliers. Outliers may include outliers that significantly deviate from the actual activity range due to device signal drift, positioning errors, or data transmission errors, such as isolated points located in the ocean or kilometers away from other locations. Cleaning can be achieved by setting reasonable spatial range thresholds (e.g., based on the species' known maximum daily activity distance) or by using velocity-based filtering algorithms (e.g., removing displacements between adjacent points that exceed the species' maximum movement speed). This process removes anomalous points that clearly do not conform to the biological activity patterns, thereby obtaining high-quality historical field survey data for the target species. These locations accurately reflect where the species has appeared or been active, and are the most direct evidence for determining its actual habitat range.

[0033] Next, step S43 is performed to extract habitat distribution data for the target species. The cleaned historical field survey data is overlaid onto the habitat type thematic map as a vector point layer. Using the spatial connectivity or spatial query function of the Geographic Information System (GIS), all habitat type patches covered by the historical field survey data are extracted. "Covered" means that the survey point falls within a habitat patch or is within a certain tolerance range (e.g., considering positioning errors). After extracting these patches, adjacent or nearby patches belonging to the same habitat type are merged to eliminate small gaps between patches, thus obtaining one or more continuously distributed areas. These areas together constitute the habitat distribution data for the target species. The rationale for this method is that the actual location of the species is direct evidence of its habitat selection. Habitat ranges delineated based on these locations are closer to reality than ranges delineated solely based on habitat type suitability assessments. This avoids mistakenly including suitable habitats that the species has never used in core habitats, allowing subsequent disturbance assessments to focus more on the areas the species truly depends on, thus improving the relevance and accuracy of the assessment.

[0034] On the other hand, step S44 is performed to obtain acoustic sensitivity threshold data for the target species. This step requires acquiring hearing test data recorded from playback experiments conducted on the target species in a bioacoustic laboratory. Hearing tests are typically conducted in a controlled acoustic environment (such as an anechoic chamber or soundproof room), playing pure tone signals of different frequencies (such as low, mid, and high frequencies) to individuals of the target species. Simultaneously, behavioral observations (such as head turning and ear movements) or physiological indicators (such as auditory brainstem response) are used to monitor and record the individual's response to sound, determining the minimum sound intensity they can hear at different frequencies, i.e., the hearing threshold. From the hearing test data, the hearing thresholds of the target species at different frequency pure tones can be analyzed; these thresholds constitute the species' hearing curve. Among these hearing thresholds, the minimum value is selected as the threshold corresponding to the most sensitive frequency point for the target species. More importantly, to assess noise interference, it is necessary to determine the critical noise intensity value at which the species exhibits avoidance behavior. This is typically achieved through behavioral experiments, which involve playing traffic noise of varying intensities or specific frequencies and observing whether the species exhibits avoidance behaviors such as fleeing, becoming wary, or ceasing normal activities. The minimum noise intensity that elicits avoidance behavior in 50% or more individuals is defined as the critical noise intensity value. This critical value is the acoustic sensitivity threshold data for the target species, reflecting the species' behavioral response threshold to noise disturbance and serving as a biological benchmark for determining whether noise causes substantial disturbance to the species.

[0035] By combining the habitat distribution of a species with its acoustic sensitivity threshold, this embodiment provides a species-specific key input for the subsequent calculation of the road network ecological disturbance index. This ensures that the assessment results truly reflect the impact of road network noise on specific protected objects, avoiding the blind use of a uniform noise threshold and significantly improving the relevance and accuracy of ecological carrying capacity estimation. Simultaneously, the use of historical field survey sites to delineate habitats ensures the realism of habitat boundaries; and the acquisition of acoustic sensitivity thresholds through behavioral experiments ensures the biological validity of the thresholds. The entire process is interconnected, laying a scientific foundation for accurate assessment.

[0036] In some implementations, S5 includes: S51: In the geographic information system software, the raster layer containing the spatiotemporal distribution data of noise is spatially overlaid with the vector layer containing the habitat distribution data of the target species. The mask extraction tool is then used to extract all noise pixels located within the boundary of the habitat distribution data to obtain the raster data of noise intensity within the habitat. S52: Perform logical operations on the noise intensity data of each pixel in the noise intensity raster data of the habitat and the acoustic sensitivity threshold data of the target species. Assign a value of 1 to pixels whose noise intensity data is greater than the acoustic sensitivity threshold data, and assign a value of 0 to the rest of the pixels to generate a noise interference Boolean raster map. S53: Calculate the total number of pixels with a value of 1 in the Boolean raster map of noise interference, multiply the total number of pixels by the actual geographical area represented by a single pixel, and use the noise interference area as the road network ecological interference index.

[0037] This embodiment further specifies the method for calculating the road network ecological disturbance index in step S5. During implementation, when executing step S51, it is necessary to spatially overlay the raster layer containing the noise spatiotemporal distribution data with the vector layer containing the target species' habitat distribution data in the geographic information system software. The noise spatiotemporal distribution data is a multi-band raster containing noise intensity values ​​for both daytime and nighttime time dimensions, with each pixel representing the equivalent continuous A-weighted sound level at that location. The habitat distribution data is a vector layer stored in polygonal form, precisely outlining the core area utilized by the target species. By running the mask extraction tool, using the polygon of the habitat distribution data as the mask range, all noise pixels exactly located within the boundary of the polygon are extracted from the noise raster layer, resulting in a new raster dataset, namely, the habitat noise intensity raster data. This step achieves precise spatial cropping, ensuring that subsequent analysis focuses only on the area where the species is actually active, excluding irrelevant areas outside the habitat, making the assessment results directly related to species conservation, and avoiding the miscalculation of noise impacts on non-habitat areas into the disturbance index.

[0038] After obtaining the noise intensity raster data within the habitat, step S52 is executed to perform pixel-by-pixel logical operations. The noise intensity data of each pixel in the habitat noise intensity raster data is compared with the acoustic sensitivity threshold data of the target species. A conditional statement is used: if the noise intensity value of the pixel is greater than the acoustic sensitivity threshold data of the target species, the ground location represented by that pixel is considered to be subject to noise interference exceeding the species' tolerance, and the pixel is assigned a value of 1; if the noise intensity value is less than or equal to the sensitivity threshold, the location is considered to be in a noise environment acceptable to the species, and the value is assigned a value of 0. After this logical operation, the raster that originally stored continuous noise intensity values ​​is converted into a raster with only two values, 0 and 1, called a noise interference Boolean raster map. This Boolean map clearly identifies which areas within the habitat are affected by excessive noise and which areas remain relatively quiet, providing direct input for subsequent area statistics.

[0039] Next, step S53 is executed to count the total number of pixels with a value of 1 in the noise interference Boolean raster map, which represents the total number of all affected pixels. This total number of pixels is then multiplied by the actual geographic area represented by each pixel. For example, if the raster resolution is in the meter range, each pixel corresponds to one square meter on the ground; the product is the noise interference area in square meters. This area value is the road network ecological disturbance index. This index quantitatively characterizes the actual encroachment or impact range of road network noise on the target species' habitat using a direct physical area. The larger the value, the more severe the interference of road network operations on the species' living environment, and the greater the pressure on the ecological carrying capacity.

[0040] Through the refinement of this embodiment, the calculation process of the road network ecological disturbance index becomes specific and transparent. Each step is supported by clear spatial analysis operations, ensuring the computability and repeatability of the index and providing a unified metric for comparative assessments between different regions and species. More importantly, this index is directly related to the species' sensitivity threshold and actual habitat, giving the assessment results a clear ecological interpretation: the disturbance area is precisely the portion of the habitat that the species cannot normally utilize due to noise, thus intuitively reflecting the degree to which the road network compresses the species' living space.

[0041] In some implementations, S5 also includes: S54: Acquire land use cover raster data of the monitoring area, and assign resistance values ​​to the degree of obstruction to the migration of target species according to different land use types, and construct species migration resistance surface raster data. S55: Obtain the two largest patches in the habitat distribution data and use them as the source and destination points of migration, respectively. Input the source, destination, and species migration resistance surface raster data into the cost connectivity tool, run the shortest path algorithm, and simulate and generate theoretical potential ecological corridor vector line data. S56: Convert the noise interference Boolean raster map into a vector polygon to obtain the noise interference area vector data. Then, spatially overlay the potential ecological corridor vector line data with the noise interference area vector data, cut out the corridor segments traversed by the noise interference area vector data, measure the total length of the traversed corridor segments, obtain the corridor connectivity loss data, and use the corridor connectivity loss data as a component of the road network ecological interference index.

[0042] This embodiment further introduces the dimension of ecological corridor connectivity loss into the calculation of the road network ecological disturbance index, enriching the connotation of the evaluation index. In the implementation process, step S54 first requires acquiring land use cover raster data of the monitoring area. This data can be a land use map obtained based on remote sensing image classification, containing various land cover types such as cultivated land, construction land, forest land, and water bodies. Resistance values ​​are assigned to the degree of obstruction to the migration of target species based on these land use types, constructing species migration resistance surface raster data. The resistance values ​​are set based on the ecological learning characteristics of the species. For example, for forest species, forest land is a suitable low-resistance area for traversal, while open farmland or densely populated construction land is a high-resistance area that is difficult to traverse. Water bodies may have different resistance values ​​depending on the swimming ability of the species. By assigning corresponding resistance values ​​to each land use type, the entire study area becomes a resistance surface where each pixel has a traversal cost, providing a basis for subsequent simulation of the ease or difficulty of species movement in the landscape. The rationality of the resistance surface construction directly affects the realism of subsequent corridor simulations; therefore, it is necessary to refer to biological literature or expert knowledge of the species when assigning values.

[0043] In step S55, the two largest patches are identified from the habitat distribution data, serving as the source and destination points of migration, respectively. These two patches are typically the species' primary concentrated distribution areas or core habitats, and their connectivity is crucial for gene flow and population dispersal. These two patches, along with the previously constructed species migration resistance surface raster data, are input into the cost connectivity tool of the Geographic Information System (GIS) software. This tool runs a shortest path algorithm, which seeks the path with the lowest cumulative cost from the source to the destination on the resistance surface—the migration route most likely chosen by the species. The algorithm considers bypassing obstacles and selecting areas with lower resistance, ultimately generating one or more theoretically potential ecological corridor vector lines. These corridors represent key channels for species migration and dispersal in the landscape, and their connectivity directly impacts the long-term survival of populations, making them a key spatial element of concern in ecological conservation.

[0044] Finally, step S56 converts the noise interference Boolean raster map into vector polygons to obtain noise interference zone vector data. Areas with a value of 1 in the noise interference Boolean raster map represent areas subjected to excessive noise interference. Using a raster-to-vector tool, these continuously distributed interference pixels can be merged into polygons to form noise interference zones. The potential ecological corridor vector line data is spatially overlaid with the noise interference zone vector data. Spatial clipping or intersection analysis is used to extract corridor segments traversed by the noise interference zone vector data. The total length of these traversed corridor segments is measured to obtain corridor connectivity loss data. This data reflects the extent to which road network noise obstructs the connectivity of species migration corridors. The longer the barrier formed by the noise interference zone on the corridor, the greater the difficulty for species to cross, and the more severe the damage to the corridor's ecological function. Including corridor connectivity loss data as a component of the road network ecological disturbance index means that the final index includes not only the noise-affected area within habitats but also connectivity loss along key ecological corridors. This comprehensively characterizes the negative impact of the road network on species' ecological space from both point and line dimensions, making the assessment more comprehensive and multi-dimensional. It provides a more targeted quantitative basis for ecological compensation and corridor restoration. For example, if a corridor is completely cut off by a noise-affected area, special measures may be needed in the construction of noise barriers or the planning of alternative corridors.

[0045] In some implementations, S54 includes: S541: Acquire global land cover data products for the monitored area and extract land cover type patches for cultivated land, construction land, forest land, and water areas from them; S542: Based on the target species’ crossing preferences for different land features recorded in biological literature, assign a first resistance value to cultivated land features, a second resistance value to construction land features, a third resistance value to forest land features, and a fourth resistance value to water features. The second resistance value is greater than the first resistance value, the first resistance value is greater than the fourth resistance value, and the fourth resistance value is greater than the third resistance value. S543: Perform raster conversion and merging on the assigned land cover types to generate species migration resistance surface raster data covering the entire monitoring area, with each pixel carrying a resistance value.

[0046] This embodiment further specifies the method for constructing the raster data of species migration resistance surface in step S54. During implementation, when executing step S541, it is necessary to acquire global land cover data products for the monitoring area, such as data from the European Space Agency's Global Land Cover Data (ESA CCI Land Cover) or 30-meter global land cover data products developed in my country. These data products have undergone professional processing, providing a unified land cover classification system with a spatial resolution typically of 30 meters or higher, meeting the needs of regional-scale analysis. Specific land cover type patches, such as cultivated land, construction land, forest land, and water areas, can be extracted from these products. Cultivated land refers to land used for planting crops, including paddy fields, dry land, and orchards; construction land refers to artificial surfaces used for human habitation, industry, mining, and transportation, including urban and rural settlements, roads, and industrial and mining land; forest land refers to areas covered by trees, including closed forests, sparse forests, and shrublands; and water areas include rivers, lakes, reservoirs, ponds, and other areas with perennial or seasonal water accumulation. These land cover types are the main landscape elements that influence species movement, and they all present different levels of resistance to species crossing.

[0047] When performing step S542, resistance values ​​need to be assigned based on the target species' crossing preferences for different land cover types, as recorded in biological literature. Crossing preferences refer to the selective or avoidant behaviors exhibited by a species when facing different land cover types, usually obtained through field tracking, experimental observation, or literature reviews. For example, for a certain forest bird or mammal, literature may indicate that they tend to move under forest cover, avoid open farmland, have difficulty crossing densely populated human-built areas, and face obstacles in wide bodies of water. Based on this information, a first resistance value is assigned to farmland features, a second resistance value to built-up land features, a third resistance value to forest land features, and a fourth resistance value to water features. The relative magnitude of the resistance values ​​reflects the ease or difficulty of crossing. Typically, a second resistance value greater than the first resistance value indicates that built-up land is more difficult to cross than farmland; a first resistance value greater than the fourth resistance value indicates that farmland is more difficult to cross than water; and a fourth resistance value greater than the third resistance value indicates that water is more difficult to cross than forest. This ranking makes forest land the substrate with the least resistance, followed by farmland and water, with built-up land having the greatest resistance. The resistance value can be any positive number; the important thing is to maintain the correct relative relationship. For ease of calculation, the minimum resistance value is usually set to 1, with others increasing sequentially. These assignments should be based on sound ecological evidence, avoiding subjective arbitrariness.

[0048] Finally, step S543 performs raster conversion and merging of the assigned land cover types. Vector-formatted farmland, construction land, forest land, and water bodies are converted into rasters, with each cell receiving a corresponding resistance value. Then, using raster mosaicking or merging tools, the raster data of all patches are combined into a single raster data set covering the entire monitoring area, with each cell carrying a resistance value. This raster data reflects the cumulative cost of species movement throughout the landscape, providing a unified resistance field for subsequent ecological corridor simulations.

[0049] Through the refinement of this embodiment, the construction process of the resistance surface has a clear ecological basis. The resistance value is set based on the biological characteristics of the species, avoiding the drawbacks of arbitrary assignment. This makes the simulated ecological corridor more consistent with the actual movement patterns of species, thereby improving the reliability of corridor connectivity analysis. Integrating this resistance surface data into the calculation of the road network ecological disturbance index means that the disturbance index not only focuses on the direct interference of noise but also indirectly relates to changes in landscape connectivity, reflecting the comprehensive impact of the road network on the structure and function of the ecosystem. For example, when the noise disturbance area happens to be located in a critical location of the corridor, even if the disturbance area is small, it may lead to serious connectivity loss. This loss is quantified through corridor connectivity loss data and thus reflected in the comprehensive index.

[0050] Example 2 like Figure 2 As shown, in a second aspect, the present invention proposes a road network ecological carrying capacity estimation system based on multi-source remote sensing data. The system adopts a road network ecological carrying capacity estimation method based on multi-source remote sensing data proposed in any of the above embodiments, and the system includes: The data acquisition and preprocessing module is used to perform step S1: acquire multispectral remote sensing images of the monitoring area and basic ecosystem data of the monitoring area, perform geometric correction and radiometric calibration on the multispectral remote sensing images to obtain preprocessed remote sensing images, and perform formatting processing on the basic ecosystem data to obtain a standardized basic dataset. The road network topology extraction module is used to perform step S2: supervised classification of road network elements in the preprocessed remote sensing image, extract linear ground feature patches in the image, and construct the topological relationship of the linear ground feature patches to obtain the road network topology data of the monitoring area. The noise spatiotemporal distribution simulation module is used to perform step S3: obtain historical traffic flow statistics data of each road segment in the road network topology data, correlate the historical traffic flow statistics data with each road segment in terms of spatial location, and perform interpolation calculation on the correlated data according to the preset time slice to generate noise spatiotemporal distribution data of the road network in the day and night time dimension. The species sensitivity information acquisition module is used to perform step S4: acquire the thematic map of habitat types in the monitoring area and historical field survey data of the target species in the monitoring area, perform spatial overlay analysis of the historical field survey data and the thematic map of habitat types, extract the spatial range of the target species' habitat, obtain the habitat distribution data of the target species, and acquire behavioral experimental records of the target species under different frequency noise stimuli, extract the critical noise intensity value when the target species exhibits avoidance behavior from the behavioral experimental records, and obtain the acoustic sensitivity threshold data of the target species. The road network ecological disturbance index calculation module is used to perform step S5: spatially overlaying the spatiotemporal distribution data of noise with the habitat distribution data of the target species, filtering out the noise intensity data within the range of the habitat distribution data, comparing the filtered noise intensity data with the acoustic sensitivity threshold data of the target species pixel by pixel, and calculating the total area of ​​pixels within the habitat distribution data where the noise intensity data exceeds the acoustic sensitivity threshold data, thereby obtaining the road network ecological disturbance index used to characterize the degree of disturbance of the road network to the activities of the target species.

[0051] This system corresponds to the method proposed in Example 1, and will not be described in detail here.

[0052] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for estimating the ecological carrying capacity of road networks based on multi-source remote sensing data, characterized in that, include: S1: Acquire multispectral remote sensing images of the monitoring area and basic ecosystem data of the monitoring area; perform geometric correction and radiometric calibration on the multispectral remote sensing images to obtain preprocessed remote sensing images; and format the basic ecosystem data to obtain a standardized basic dataset. S2: Supervised classification of road network elements is performed on the preprocessed remote sensing image, linear land cover patches are extracted from the image, and topological relationships are constructed on the linear land cover patches to obtain road network topology data of the monitoring area. S3: Obtain historical traffic flow statistics for each road segment in the road network topology data, spatially correlate the historical traffic flow statistics with each road segment, and interpolate the correlated data based on the preset time slices to generate noise spatiotemporal distribution data of the road network in the day and night time dimension. S4: Obtain thematic maps of habitat types in the monitoring area and historical field survey data of target species in the monitoring area. Perform spatial overlay analysis on the historical field survey data and habitat type thematic maps to extract the spatial range of the target species' habitat and obtain the habitat distribution data of the target species. At the same time, obtain behavioral experimental records of the target species under different frequency noise stimuli and extract the critical noise intensity value when the target species exhibits avoidance behavior from the behavioral experimental records to obtain the acoustic sensitivity threshold data of the target species. S5: Spatially overlay the spatiotemporal distribution data of noise with the habitat distribution data of the target species, filter out the noise intensity data within the habitat distribution data range, and compare the filtered noise intensity data with the acoustic sensitivity threshold data of the target species pixel by pixel. Calculate the total area of ​​pixels within the habitat distribution data range where the noise intensity data exceeds the acoustic sensitivity threshold data, and obtain the road network ecological disturbance index used to characterize the degree of road network interference with the activities of the target species.

2. The method for estimating the ecological carrying capacity of a road network based on multi-source remote sensing data according to claim 1, characterized in that, S2 include: S21: Perform multi-scale segmentation on the preprocessed remote sensing image to generate multiple image objects composed of homogeneous pixels, and obtain a candidate road network patch object set; S22: Perform geometric feature calculations on each object in the candidate road network patch object set to obtain the aspect ratio, shape index, and area parameters of each object, and obtain the geometric morphological feature set of the candidate patches; S23: Compare the aspect ratio of the geometric feature set with the preset aspect ratio threshold of linear features, compare the shape index with the preset road smoothness threshold, and filter out objects that simultaneously meet the aspect ratio threshold of linear features and the road smoothness threshold to obtain the target road network patch object set. S24: Extract centerlines from objects in the target road network patch object set, and capture endpoints and construct connection relationships for the extracted centerlines to generate road network topology data composed of nodes and arc segments.

3. The method for estimating the ecological carrying capacity of a road network based on multi-source remote sensing data according to claim 2, characterized in that, S22 includes: S221: For each image object in the candidate road network patch object set, the edge detection operator is used to calculate the gradient, identify the edge points where the pixel values ​​inside the object change abruptly, and connect the edge points into a closed boundary line according to the eight-neighbor tracking algorithm to obtain the contour line data of each image object. S222: Based on the contour data of each image object, count the total number of pixels contained in the area enclosed by the contour lines, and multiply the total number of pixels by the actual ground area represented by a single pixel to obtain the area parameter of the image object. S223: Based on the contour data and area parameters of each image object, measure the total length of the contour as the perimeter parameter, and use the ratio of the square of the perimeter parameter to the area parameter as the shape index of the image object; S224: Based on the contour data of each image object, the minimum bounding rectangle algorithm is used to fit the region enclosed by the contour data, calculate the length of the long side and the length of the short side of the minimum rectangle that completely contains the contour, and use the ratio of the length of the long side to the length of the short side as the aspect ratio of the image object. S225: Associate the aspect ratio, shape index, and area parameter as attribute fields to generate a feature record for each image object containing the aspect ratio, shape index, and area parameter. After summarizing the feature records of all image objects, a geometric shape feature set is obtained.

4. The method for estimating the ecological carrying capacity of a road network based on multi-source remote sensing data according to claim 1, characterized in that, S3 include: S31: Obtain historical traffic flow time series data for each road segment released by the traffic monitoring department, classify and statistically analyze the historical traffic flow time series data according to weekdays and rest days, and obtain the basic traffic flow data of the classified road segments. S32: The traffic flow data of the classified road segments are summed separately according to the daytime and nighttime periods to obtain the first average flow value of each road segment during the daytime period and the second average flow value during the nighttime period. S33: Obtain digital elevation model data of the monitoring area, extract surface roughness raster from the digital elevation model data, use the surface roughness raster as the surface attenuation factor for sound wave propagation, and use the first average flow rate and the second average flow rate as the intensity parameters of the line sound source, respectively, input into the sound wave propagation geometric attenuation model for calculation, and generate daytime noise intensity raster and nighttime noise intensity raster. S34: Combine the daytime noise intensity raster map with the nighttime noise intensity raster map by band to generate noise spatiotemporal distribution data containing both day and night time dimensions.

5. The method for estimating the ecological carrying capacity of a road network based on multi-source remote sensing data according to claim 1, characterized in that, S4 include: S41: Acquire satellite remote sensing images covering the monitoring area, use an object-based image classification method to classify land cover types in the satellite remote sensing images, and generate habitat type thematic maps containing information on forest, grassland, water, and farmland categories; S42: Obtain the set of spatial locations recorded by GPS collar tracking of the target species in the wildlife monitoring database, clean the set of spatial locations to remove outliers, and obtain historical field survey location data of the target species. S43: The cleaned historical field survey data is overlaid onto the habitat type thematic map in the form of vector point layers. All habitat type patches covered by the historical field survey data are extracted and merged to obtain the habitat distribution data of the target species. S44: Obtain hearing test data recorded in the playback experiment of the target species in the bioacoustics laboratory, analyze the hearing threshold of the target species under different frequency pure tones from the hearing test data, and determine the minimum value of the hearing threshold as the critical noise intensity value for the target species to produce avoidance behavior, thereby obtaining the acoustic sensitivity threshold data of the target species.

6. The method for estimating the ecological carrying capacity of a road network based on multi-source remote sensing data according to claim 1, characterized in that, S5 include: S51: In the geographic information system software, the raster layer containing the spatiotemporal distribution data of noise is spatially overlaid with the vector layer containing the habitat distribution data of the target species. The mask extraction tool is then used to extract all noise pixels located within the boundary of the habitat distribution data to obtain the raster data of noise intensity within the habitat. S52: Perform logical operations on the noise intensity data of each pixel in the noise intensity raster data of the habitat and the acoustic sensitivity threshold data of the target species. Assign a value of 1 to pixels whose noise intensity data is greater than the acoustic sensitivity threshold data, and assign a value of 0 to the rest of the pixels to generate a noise interference Boolean raster map. S53: Calculate the total number of pixels with a value of 1 in the Boolean raster map of noise interference, multiply the total number of pixels by the actual geographical area represented by a single pixel, and use the noise interference area as the road network ecological interference index.

7. The method for estimating the ecological carrying capacity of a road network based on multi-source remote sensing data according to claim 6, characterized in that, S5 also includes: S54: Acquire land use cover raster data of the monitoring area, and assign resistance values ​​to the degree of obstruction to the migration of target species according to different land use types, and construct species migration resistance surface raster data. S55: Obtain the two largest patches in the habitat distribution data and use them as the source and destination points of migration, respectively. Input the source, destination, and species migration resistance surface raster data into the cost connectivity tool, run the shortest path algorithm, and simulate and generate theoretical potential ecological corridor vector line data. S56: Convert the noise interference Boolean raster map into a vector polygon to obtain the noise interference area vector data. Then, spatially overlay the potential ecological corridor vector line data with the noise interference area vector data, cut out the corridor segments traversed by the noise interference area vector data, measure the total length of the traversed corridor segments, obtain the corridor connectivity loss data, and use the corridor connectivity loss data as a component of the road network ecological interference index.

8. The method for estimating the ecological carrying capacity of a road network based on multi-source remote sensing data according to claim 7, characterized in that, S54 includes: S541: Acquire global land cover data products for the monitored area and extract land cover type patches for cultivated land, construction land, forest land, and water areas from them; S542: Based on the target species’ crossing preferences for different land features recorded in biological literature, assign a first resistance value to cultivated land features, a second resistance value to construction land features, a third resistance value to forest land features, and a fourth resistance value to water features. The second resistance value is greater than the first resistance value, the first resistance value is greater than the fourth resistance value, and the fourth resistance value is greater than the third resistance value. S543: Perform raster conversion and merging on the assigned land cover types to generate species migration resistance surface raster data covering the entire monitoring area, with each pixel carrying a resistance value.

9. A road network ecological carrying capacity estimation system based on multi-source remote sensing data, characterized in that, The system employs a road network ecological carrying capacity estimation method based on multi-source remote sensing data as described in any one of claims 1 to 8, and the system comprises: The data acquisition and preprocessing module is used to perform step S1: acquire multispectral remote sensing images of the monitoring area and basic ecosystem data of the monitoring area, perform geometric correction and radiometric calibration on the multispectral remote sensing images to obtain preprocessed remote sensing images, and perform formatting processing on the basic ecosystem data to obtain a standardized basic dataset. The road network topology extraction module is used to perform step S2: supervised classification of road network elements in the preprocessed remote sensing image, extract linear ground feature patches in the image, and construct the topological relationship of the linear ground feature patches to obtain the road network topology data of the monitoring area. The noise spatiotemporal distribution simulation module is used to perform step S3: obtain historical traffic flow statistics data of each road segment in the road network topology data, correlate the historical traffic flow statistics data with each road segment in terms of spatial location, and perform interpolation calculation on the correlated data according to the preset time slice to generate noise spatiotemporal distribution data of the road network in the day and night time dimension. The species sensitivity information acquisition module is used to perform step S4: acquire the thematic map of habitat types in the monitoring area and historical field survey data of the target species in the monitoring area, perform spatial overlay analysis of the historical field survey data and the thematic map of habitat types, extract the spatial range of the target species' habitat, obtain the habitat distribution data of the target species, and acquire behavioral experimental records of the target species under different frequency noise stimuli, extract the critical noise intensity value when the target species exhibits avoidance behavior from the behavioral experimental records, and obtain the acoustic sensitivity threshold data of the target species. The road network ecological disturbance index calculation module is used to perform step S5: spatially overlaying the spatiotemporal distribution data of noise with the habitat distribution data of the target species, filtering out the noise intensity data within the range of the habitat distribution data, comparing the filtered noise intensity data with the acoustic sensitivity threshold data of the target species pixel by pixel, and calculating the total area of ​​pixels within the habitat distribution data where the noise intensity data exceeds the acoustic sensitivity threshold data, thereby obtaining the road network ecological disturbance index used to characterize the degree of disturbance of the road network to the activities of the target species.

Images