A method, system, medium, and device for wild-field-farmland interface zoning

Abstract

The present application relates to the technical field of remote sensing image processing, and provides a wild field-farmland interface area division method, system, medium and equipment, which comprises the following steps: converting farmland grids into vector polygons to obtain a plurality of farmland patches, and filtering farmland patches with an area less than a first threshold value; constructing an outer expansion annular buffer zone for each retained farmland patch; marking the outer expansion annular buffer zone with a natural vegetation coverage greater than a second threshold value as a potential wild field-farmland interface area; dividing the potential wild field-farmland interface area into four quadrants with the geometric center point of the farmland patch as the origin; calculating the natural vegetation coverage of each quadrant sub-region; and identifying mixed wild field-farmland interface areas and interface-type wild field-farmland interface areas according to the quadrant-level natural vegetation coverage threshold value and the quadrant compliance number constraint. The present application effectively solves the quantification and identification problem of the forest-farmland transition area, can effectively identify the wild field-farmland interface area, and provides a basis for the wild fire risk assessment of the agriculture-dominant area.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of remote sensing image processing, and particularly relates to a wildland-farmland interface zoning method, system, medium and equipment. BACKGROUND

[0002] The statements in this section merely provide background information related to the present application and do not necessarily constitute prior art.

[0003] Under the dual background of global climate change intensification and human activity range expansion, the frequency and destructiveness of wild fires are increasingly prominent. In order to cope with the high risk of wild fires in urban fringe areas, the concept of wildland-urban interface (WUI) is proposed, which is defined as the transition zone where natural vegetation ecosystems and highly urbanized areas intersect or mix. The core risk of WUI comes from the close contact or high spatial mosaic of natural flammable vegetation and human settlements, buildings and facilities, greatly increasing the probability of fire caused by natural or human factors. Once a fire breaks out, it is extremely easy to spread to the building group, causing serious casualties and property losses. The expansion of this area is mainly driven by the uncontrolled spread of urban space and the increase of vegetation coverage in the surrounding area, resulting in high building density, large flammable vegetation coverage, and fragmented distribution of natural vegetation patches in the region, with obvious regional aggregation and complex dynamic changes in fire behavior.

[0004] However, the unique fire risk pattern of the agriculture-dominated area mainly shows that in terms of urban structure, most of the residential areas at the urban-rural junction present high density and are concentrated in a single piece, and their periphery is isolated from the natural environment through planning or buffer zone, and there are relatively few cases of high mixing and mosaic of dense building groups and natural vegetation inside. In contrast, a prominent feature of the landscape pattern in this area is the widespread existence of large areas of contiguous farmland, which are directly adjacent to forest, grassland, shrub and other natural vegetation areas, forming a broad contact boundary.

[0005] More importantly, there has been a long-term tradition of burning straw and other traditional agricultural fire habits in agricultural production. The agriculture-dominated area is generally faced with the risk pattern of wild fires in which large areas of farmland are directly adjacent to natural wildlands. However, the concept of WUI and its related technical system are based on the low-density single-family residential-dominated urban and rural structure, and when it is directly applied to the agriculture-dominated area with significantly different spatial patterns and farming traditions, there are fundamental limitations, making it difficult to effectively cover and reflect the main source of wild fire risk, accurately identify and quantify the wildland-farmland interlaced structure with significant heterogeneity and its inherent risk formation mechanism, resulting in insufficient pertinence of risk assessment and prevention measures. SUMMARY

[0006] To solve the technical problems in the background art, the present application provides a wild field-farmland interface area division method, system, medium and equipment, creates a multi-level progressive spatial verification architecture, effectively solves the quantitative identification problem of the forest-agriculture interlaced area, can effectively identify the wild field-farmland interface area suitable for the characteristics of the agricultural area, and provides a basis for the wild fire risk assessment of the agricultural dominant area.

[0007] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0008] The first aspect of the present application provides a wild field-farmland interface area division method, which comprises:

[0009] Obtaining a remote sensing image of a target area, obtaining a wild field area and a farmland area through image segmentation, and respectively encoding the wild field area and the farmland area into a grid;

[0010] Converting the farmland grid into a vector polygon to obtain a plurality of farmland patches, filtering the farmland patches with an area less than a first threshold value, constructing an outer expansion annular buffer zone for each retained farmland patch, calculating the natural vegetation coverage of the outer expansion annular buffer zone of each farmland patch, and marking the outer expansion annular buffer zone with a natural vegetation coverage greater than a second threshold value as a potential wild field-farmland interface area;

[0011] Dividing the potential wild field-farmland interface area into four quadrants with the geometric center point of the farmland patch as the origin, calculating the natural vegetation coverage of each quadrant sub-area, and identifying a mixed wild field-farmland interface area and an interface wild field-farmland interface area in the potential wild field-farmland interface area according to the quadrant-level natural vegetation coverage threshold value and the quadrant compliance number constraint;

[0012] Wherein, the natural vegetation coverage is the ratio of the number of wild field grids in the area to the total number of grids in the area.

[0013] Further, the step of identifying the mixed wild field-farmland interface area and the interface wild field-farmland interface area in the potential wild field-farmland interface area comprises: for a certain potential wild field-farmland interface area, if the number of quadrant sub-areas with a natural vegetation coverage greater than a third threshold value meets the standard, the potential wild field-farmland interface area is determined as a mixed wild field-farmland interface area; otherwise, converting the wild field grid into a vector polygon to obtain a plurality of wild field patches, filtering the wild field patches with an area less than a fourth threshold value, constructing a buffer zone for each retained wild field patch, detecting the spatial intersection relationship between the buffer zones of the farmland patches and the wild field patches corresponding to the potential wild field-farmland interface area, retaining the intersected farmland patches, and for the outer expansion annular buffer zone of a certain intersected farmland patch, if the number of quadrant sub-areas with a natural vegetation coverage greater than a fifth threshold value meets the standard, the potential wild field-farmland interface area is determined as an interface wild field-farmland interface area.

[0014] Further, the image segmentation step comprises:

[0015] For the target area remote sensing image, feature extraction is performed by standard convolution to obtain high-resolution features;

[0016] For the target area remote sensing image, feature extraction is performed by dilated convolution after down-sampling to obtain down-sampled features;

[0017] After up-sampling the down-sampled features, the normalized vegetation index is calculated to obtain enhanced features, and the gray level co-occurrence matrix is applied to the enhanced features to calculate the texture roughness and near-infrared reflectivity to obtain texture features and reflectivity features;

[0018] After feature fusion of the high-resolution features, the down-sampled features, the enhanced features, the texture features and the reflectivity features, the classifier is used to obtain shrubs, mosses, mangroves, grasslands, herbaceous wetlands and forest regions.

[0019] Further, the image segmentation step further comprises:

[0020] For the target area remote sensing image, the normalized water index is calculated, and the area with the normalized water index exceeding the set value is regarded as the herbaceous wetland, and the intersection with the herbaceous wetland obtained by the classifier is obtained to obtain the final herbaceous wetland;

[0021] The shrubs, mosses, mangroves, grasslands, herbaceous wetlands and forest regions are merged into wild regions;

[0022] For the target area remote sensing image, the area other than the wild region is regarded as the farmland pre-classification region.

[0023] Further, the image segmentation step further comprises: based on the histogram of oriented gradients, the consistency of the field ridge direction of the farmland pre-classification region is quantified to screen out the farmland region.

[0024] The second aspect of the present application provides a wild-farmland interface area division system, which comprises:

[0025] An image segmentation module configured to: acquire a target area remote sensing image, obtain a wild region and a farmland region by image segmentation, and encode the wild region and the farmland region into a grid respectively;

[0026] A preliminary division module configured to: convert the farmland grid into a vector polygon to obtain a plurality of farmland patches, filter the farmland patches with an area less than a first threshold value, construct an outer expansion annular buffer zone for each retained farmland patch, calculate the natural vegetation coverage of the outer expansion annular buffer zone of each farmland patch, and mark the outer expansion annular buffer zone with a natural vegetation coverage greater than a second threshold value as a potential wild-farmland interface area.

[0027] a final division module configured to: divide the potential wild field-cropland interface with the cropland patch geometric center point as the origin into four quadrants, calculate the natural vegetation coverage of each quadrant sub-region, and identify the mixed wild field-cropland interface and the interface wild field-cropland interface in the potential wild field-cropland interface according to the quadrant-level natural vegetation coverage threshold and the number-of-quadrant constraint;

[0028] wherein the natural vegetation coverage is the ratio of the number of wild field grids in the region to the total number of grids in the region.

[0029] Further, the step of identifying the mixed wild field-cropland interface and the interface wild field-cropland interface in the potential wild field-cropland interface comprises: for a certain potential wild field-cropland interface, if the number of quadrant sub-regions with natural vegetation coverage greater than a third threshold meets the standard, the potential wild field-cropland interface is determined as a mixed wild field-cropland interface; otherwise, the wild field grids are converted into vector polygons to obtain a plurality of wild field patches, and wild field patches with an area less than a fourth threshold are filtered, a buffer zone is constructed for each retained wild field patch, the spatial intersection relationship between the buffer zones of the wild field patch and the cropland patch corresponding to the potential wild field-cropland interface is detected, and the intersected cropland patches are retained, and for the outer expansion annular buffer zone of a certain intersected cropland patch, if the number of quadrant sub-regions with natural vegetation coverage greater than a fifth threshold meets the standard, the potential wild field-cropland interface is determined as an interface wild field-cropland interface.

[0030] Further, the step of image segmentation comprises:

[0031] For the target area remote sensing image, feature extraction is performed through standard convolution to obtain high-resolution features;

[0032] After down-sampling of the target area remote sensing image, feature extraction is performed through hole convolution to obtain down-sampled features;

[0033] After up-sampling of the down-sampled features, the normalized vegetation index is calculated to obtain enhanced features, and the gray level co-occurrence matrix is applied to the enhanced features to calculate the texture roughness and near-infrared reflection intensity to obtain texture features and reflection features;

[0034] After feature fusion of the high-resolution features, the down-sampled features, the enhanced features, the texture features and the reflection features, the classifier is used to obtain shrub, moss, mangrove, grassland, herbaceous wetland and forest regions.

[0035] The third aspect of the application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the wild field-cropland interface zoning method described above.

[0036] The fourth aspect of the present application provides a computer device comprising a computer readable storage medium, a processor and a computer program stored on the computer readable storage medium and executable on the processor, wherein the processor implements the steps of a wildland-farmland interface area division method as described above when executing the program.

[0037] Compared with the prior art, the present application has the following beneficial effects:

[0038] The present application creates a multi-level progressive spatial verification architecture, effectively solving the quantitative identification problem of the forest-agricultural interlaced area: firstly, the macro risk area is preliminarily screened by using the global buffer zone analysis, secondly, the four-quadrant rigid geometric segmentation method along the north-south / east-west axis is innovatively introduced, combined with the quadrant-level vegetation coverage threshold and the quadrant compliance number constraint, the mixed WCI risk characteristics of the deep penetration inside the farmland are accurately analyzed, and finally the interface conduction type WCI area is locked through the continuous wildland patch boundary identification and single-quadrant coverage verification; the present application overcomes the defect of the traditional radial buffer zone in the insufficient perception of heterogeneous space, and for the first time realizes the physical separation of the two types of disaster mechanisms of "spatial penetration" and "boundary contact" in the landscape fragmentation area.

[0039] The present application constructs a grid-vector dual-engine collaborative mechanism at the spatial calculation level, uses grid pixel statistics to ensure the micro coverage calculation accuracy, and realizes the dynamic analysis of the geometric characteristics of farmland through vector spatial operation; in the coverage calculation, the geometric constraint mechanism is introduced, the rigid north-south / east-west axis four-quadrant segmentation is used instead of the traditional free direction division, the spatial structure distortion error is eliminated, and a transferable wild-farmland interaction quantification tool chain is formed.

[0040] The present application constructs a wildland-farmland interface theory framework suitable for the characteristics of agricultural areas, innovates the traditional wild fire risk recognition system centered on towns, and provides a new theoretical basis for wild fire risk assessment in agricultural dominant areas. BRIEF DESCRIPTION OF DRAWINGS

[0041] The drawings accompanying the specification of the present application serve to provide a further understanding of the present application, and the illustrative embodiments of the present application and their descriptions serve to explain the present application, and do not constitute an improper limitation on the present application.

[0042] Figure 1 is a flowchart of a wildland-farmland interface area division method of the present application embodiment one;

[0043] Figure 2 is a multi-element semantic segmentation identification model structure diagram based on the HRNetV2 network of the present application embodiment one;

[0044] Figure 3 is a wildland-farmland interface (WCI) determination flowchart of the present application embodiment one;

[0045] Figure 4 Figure 1 is a structural schematic diagram of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0046] In order to make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0047] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0048] Embodiment One

[0049] The embodiment provides a wildland-cropland interface area division method.

[0050] The wildland-cropland interface area division method provided by the embodiment proposes a wildland-cropland interface (WCI) which is a wild fire high-risk area specific to an agriculture-dominated area. In the WCI area, a fire source in a natural vegetation area or a straw burning activity in a cropland is extremely easy to break the boundary and ignite the adjacent combustible material. In particular, it is worth noting that the vast cropland itself can become a channel for long-distance and rapid spread of wild fire, threatening vegetation areas or communities farther away, which not only poses a serious threat to important ecological resources such as forests and grasslands in the region, but also brings great risks to valuable cropland resources, agricultural production facilities and food security.

[0051] The wildland-cropland interface area division method provided by the embodiment innovatively proposes a multi-level progressive spatial verification architecture and a mixed / interface WCI grading system, aiming to accurately depict different modes of spatial interaction between cropland and wildland, so as to realize physical separation of risk mechanisms and provide direct and effective technical support for scientific identification and demarcation of WCI, fire risk assessment and regional planning management.

[0052] The wildland-cropland interface area division method provided by the embodiment, as shown in Figure 1 includes the following steps:

[0053] Step 1: as shown in Figure 2As shown, the high-resolution 2 remote sensing image of the target area is acquired, image preprocessing is performed, and the trained high-resolution network (High-Resolution Representations for Labeling Pixels and Regions, HRNetV2) model is used for semantic segmentation of the target area remote sensing image to realize land class classification and merging, and the wild and farmland segmentation results are obtained.

[0054] Step 101, retrieve the high-resolution 2 multispectral image data covering the target area from the high-resolution satellite data center, and the cloud coverage of the target area remote sensing image needs to be less than 10%.

[0055] Step 102, use the satellite's supporting radiation calibration coefficient file to convert the original pixel brightness (DN) value to apparent radiance.

[0056] Step 103, for the target area remote sensing image, use atmospheric correction (FLAASH), input the imaging time, geographic location, aerosol model, combine with the radiation transfer model, eliminate the atmospheric scattering effect, and output the ground reflectance data.

[0057] Step 104, perform blue / green / red / near-infrared four-band standardization on the preprocessed high-resolution 2 remote sensing image data, use the NDWI index (Normalized Difference Water Index, Normalized Difference Water Index) (NDWI=(Green-NIR) / (Green+NIR)) to preliminarily screen the region with NDWI>0.3 as the herbaceous wetland pixel, and enhance the near-infrared reflectivity of the herbaceous wetland pixel to 1.5 times the original value. Among them, Green is the green band, and NIR is the near-infrared band.

[0058] Step 105, initialize the double-flow HRNetV2 model, and process the target area remote sensing image in double-flow, the main branch processes 512×512 original resolution images to capture details, and the secondary branch processes 128×128 down-sampled images to identify large-scale features.

[0059] Step 106, the high-resolution branch (i.e., the main branch) extracts shrub and moss features through a 3×3 standard convolution kernel, and the down-sampling branch (i.e., the secondary branch) uses a dilated convolution with an expansion rate of 3 to capture wetland, mangrove, and grassland features.

[0060] The target area remote sensing image has four bands of blue / green / red / near-infrared, and the weight of the first four bands before step 105 is 1:1:1:1. In the main and secondary branch convolution layers, the weight of the near-infrared band (NIR) is expanded by 1.25 times, becoming 1:1:1:1.25, and the near-infrared channel weight is increased by 25% to strengthen the vegetation response.

[0061] Step 107, after the sub-branch extracted features are up-sampled to 512x512, an NDVI (Normalized Difference Vegetation Index) feature layer is added, NDVI enhanced features are calculated, NDVI=(NIR-R) / (NIR+R), to highlight the herbaceous wetland, wherein NIR is the near-infrared band, and R is the red band; the result (NDVI enhanced features) obtained after adding the NDVI feature layer is applied to a gray level co-occurrence matrix, the texture roughness and the near-infrared reflectivity are calculated, the texture features and the reflection features are obtained, and the calculation formula is as follows:

[0062] ;

[0063] wherein C is the texture roughness, P(i,j) is the probability of the gray level i and j co-occurrence in the gray level co-occurrence matrix, and N is the number of quantized gray levels;

[0064] ;

[0065] wherein NIR i is the near-infrared reflectivity, NIR p is the near-infrared reflectivity value at the pixel P, and k is the total number of pixels in the calculation window.

[0066] Step 108, the high-resolution features extracted by the main branch, the down-sampled features extracted by the sub-branch, the NDVI enhanced features, the texture features and the reflection features are fused by a connection operation, and for the fused features, a classifier is used to obtain shrub, moss, mangrove, grassland, herbaceous wetland and forest area.

[0067] Step 109, the herbaceous wetland pixels with NDWI>0.3 in step 104 are intersected with the herbaceous wetland pixels obtained in step 108, and are marked as herbaceous wetland.

[0068] Step 110, after the original image is subtracted from the shrub, moss, mangrove, grassland area, forest area obtained in step 108, and the herbaceous wetland area obtained in step 109, the remaining area is regarded as a farmland pre-classification area, the direction gradient histogram (HOG) is used to quantify the ridge direction consistency in the farmland pre-classification area, and the rule strip with more than 75% uniform direction is marked as farmland category, and the specific calculation formula is as follows:

[0069] ;

[0070] wherein R is the main direction concentration ratio, i is the direction number (the present application takes 10° interval as one direction, and there are 36 directions), H i is the gradient amplitude sum of the i-th direction interval, and max(H i) is the amplitude of the main direction interval, and when R≥0.75, it indicates that the area has a dominant direction, so it is determined as a farmland category.

[0071] Step 111, combine the six types of land objects, forest land, grassland, shrub, herbaceous wetland, mangrove and moss lichen, into a single "wild land" layer, and keep the original classification attribute field.

[0072] Step 112, for the combined wild land-farmland binary classification map, use a 5x5 window median filter to eliminate salt and pepper noise, and remove small patches below 100 square meters by patch area calculation.

[0073] Step 113, combine the wild land layer (containing the attribute fields of the six types of land objects, forest land, grassland, shrub, herbaceous wetland, mangrove and moss lichen) and the farmland layer, and encode to generate a dual-channel GeoTIFF (Geographic Tagged Image File Format) raster, with a spatial resolution consistent with the original resolution of the Gaofen 2 remote sensing image (channel 1 wild land classification / channel 2 farmland mask).

[0074] Step 2: Convert the farmland raster data obtained by image segmentation into vector polygons, create an outer expansion buffer for each farmland patch, calculate the vegetation coverage in the buffer, and mark it as a potential WCI area.

[0075] Step 201, convert the farmland raster to vector polygons to obtain several farmland patches, automatically filter farmland patches with an area <0.1 square kilometers, and generate a unique ID and spatial attribute table (including area, perimeter, etc.) for each farmland patch.

[0076] Step 202, build a 2.4 km outer ring buffer for each retained farmland patch.

[0077] Step 203, calculate the natural vegetation coverage for each farmland patch buffer, using the binary wild land raster as input, and the statistical formula is defined as:

[0078] (buffer wild land pixel count / total buffer pixel count) x 100%;

[0079] It should be noted that pixel and raster are the same concept.

[0080] In the statistical process, the original resolution of Gaofen 2 is counted, and the buffer with natural vegetation coverage ≥25% within the buffer boundary is selected as a potential WCI area, and the rest is considered as a non-interface domain and removed from the processing flow.

[0081] Step 3: As Figure 3The buffer of the potential WCI area is divided into four quadrants, and the vegetation coverage is calculated quadrant by quadrant to determine the mixed WCI area.

[0082] Step 301: The farmland buffer area marked as potential WCI is divided into four quadrants (B1, B2, B3, and B4) with the farmland patch geometric center point as the origin, and the four quadrants are divided into north (N, 0°), east (E, 90°), south (S, 180°), and west (W, 270°).

[0083] Step 302: The vegetation coverage is calculated independently for each quadrant sub-area, and the coverage calculation method is consistent with step 203. That is, the vegetation coverage within the four-quadrant polygon is calculated.

[0084] Step 303: When the coverage of three or more quadrants in the buffer area corresponding to a single farmland patch is greater than or equal to 50%, the farmland patch buffer area is determined to be mixed WCI, and the final output is a mixed WCI vector set with an attribute identification field; otherwise, it is determined to be a potential non-mixed WCI farmland patch.

[0085] Step 4: The large wild patch is intersected with the potential non-mixed WCI farmland patch through a 2.4 km buffer, and the intersected patch is retained to perform four-quadrant coverage calculation to determine the interface type WCI area.

[0086] Step 401: The wild grid data after image segmentation and merging is converted into vector polygons to obtain a number of wild patches, the area of each wild patch is calculated, and large continuous wild patches with an area greater than or equal to 5 km 2 are retained.

[0087] Step 402: A 2.4 km buffer is constructed for the retained large wild patch, and the spatial intersection relationship between the wild patch and the potential WCI area farmland patch (i.e., potential non-mixed WCI farmland patch) that is not marked as mixed WCI is detected, and the intersected farmland patch is retained. That is, whether the polygon is located near a dense vegetation area with an area greater than or equal to 5 km 2 is verified.

[0088] Step 403: The retained intersected farmland patch is subjected to four-quadrant coverage calculation, and the calculation method is consistent with step 203.

[0089] Step 404: When the coverage of one or more quadrants in the buffer area corresponding to a single farmland patch is greater than or equal to 50%, the farmland patch buffer area is determined to be interface type WCI, and the final output is an interface type WCI vector set with an attribute identification field.

[0090] Step 5: The mixed WCI and interface type WCI spatial distribution data are integrated to generate a wild-farmland interface area zoning map.

[0091] The embodiment provides a wild field-farmland interface area division method, constructs a wild field-farmland interface (WCI) theoretical framework suitable for the characteristics of agricultural areas, and innovates the traditional wild fire risk cognition system centered on towns. The framework first systematically explains the dual disaster attribute of the farmland ecosystem: it is not only a medium for the outward spread of wild fires through straw burning and other agricultural activities, but also a core disaster-bearing body for the reverse invasion of wild fires from natural areas. Compared with traditional models, the theory focuses on the fire dynamics mechanism dominated by spatial connectivity, reveals the key role of farmland as a physical channel in cross-border fire spread, and provides a new theoretical basis for wild fire risk assessment in agricultural-dominated areas.

[0092] The embodiment provides a wild field-farmland interface area division method, creates a multi-level progressive spatial verification architecture, and effectively solves the quantitative identification problem of the forest-farmland interlaced area. First, a 2.4 km global buffer analysis is used to realize macro risk area preliminary screening (vegetation coverage ≥ 25%); second, a rigid geometric partition method of four quadrants in the north-south / east-west direction is innovatively introduced, combined with a quadrant-level vegetation coverage threshold (≥ 50%) and a quadrant compliance number constraint (≥ 3), to accurately analyze the mixed WCI risk characteristics of the deep penetration of farmland; finally, through a ≥ 5 km 2 continuous wild field patch boundary identification and single-quadrant coverage verification (boundary contact area coverage ≥ threshold), the interface conduction type WCI area is locked. The method overcomes the defect of insufficient heterogeneity space perception of traditional radial buffer, and for the first time realizes the physical separation of the two disaster mechanisms of "space penetration" and "boundary contact" in the landscape fragmentation area.

[0093] The embodiment provides a wild field-farmland interface area division method, which deeply integrates deep learning and spatial topology calculation technology, and constructs a multi-scale collaborative analysis engine. The HRNetV2 deep learning model is used to realize high-precision pixel-level synchronous segmentation of six types of natural features such as forest land and grassland, and generate a semantic structured binary base. A grid-vector dual-engine collaborative mechanism is constructed at the spatial calculation level, the grid pixel statistics are used to guarantee the micro coverage calculation accuracy, and the vector space operation is used to realize the dynamic analysis of the geometric characteristics of farmland. In the coverage calculation, a geometric constraint mechanism is introduced, a rigid north-south / east-west four-quadrant partition is used instead of the traditional free direction division, the spatial structure distortion error is eliminated, and a transferable farmland-wildland interaction quantitative tool chain is formed.

[0094] The embodiment provides a wild field-farmland interface area division method, establishes a mixed / interface WCI grading system, and provides a precise space guide for differentiated prevention and control strategies. In view of the deep penetration characteristics of vegetation in the mixed area, it is suggested to implement a grid fire prevention isolation belt construction project and all-weather agricultural fire control; in view of the boundary contact characteristics of large-scale combustibles in the interface area, a boundary buffer zone fire retardant project and a fire source activity time scheduling mechanism are designed. The technical framework realizes the transformation from homogenization prevention and control to "risk mechanism identification-prevention and control strategy adaptation".

[0095] Embodiment two

[0096] The embodiment provides a wild field-farmland interface area division system, which specifically comprises:

[0097] An image segmentation module configured to: acquire a remote sensing image of a target area, obtain a wild field area and a farmland area through image segmentation, and encode the wild field area and the farmland area into grids respectively;

[0098] A preliminary division module configured to: convert the farmland grid into a vector polygon to obtain a plurality of farmland patches, filter a farmland patch with an area less than a first threshold value, construct an outer expansion annular buffer zone for each reserved farmland patch, calculate the natural vegetation coverage of the outer expansion annular buffer zone of each farmland patch, and mark an outer expansion annular buffer zone with a natural vegetation coverage greater than a second threshold value as a potential wild field-farmland interface area.

[0099] A final division module configured to: divide the potential wild field-farmland interface area into four quadrants with the geometric center point of the farmland patch as the origin, calculate the natural vegetation coverage of each quadrant sub-area, and identify a mixed wild field-farmland interface area and an interface wild field-farmland interface area in the potential wild field-farmland interface area according to a quadrant-level natural vegetation coverage threshold and a quadrant compliance number constraint.

[0100] A spatial data integration module configured to: spatial topological fusion of the mixed WCI and the interface WCI, generate a standardized division map and a statistical report.

[0101] Wherein, the natural vegetation coverage is a ratio of the number of wild field grids in the area to the total number of grids in the area.

[0102] The step of identifying the mixed wild field-crop field junction and the interface wild field-crop field junction in the potential wild field-crop field junction region comprises: for a potential wild field-crop field junction, if the number of the quadrant sub-regions with the natural vegetation coverage greater than the third threshold value meets a criterion, the potential wild field-crop field junction is determined as the mixed wild field-crop field junction; otherwise, the wild field grid is converted into a vector polygon to obtain a plurality of wild field patches, and a wild field patch with an area less than a fourth threshold value is filtered out, a buffer zone is constructed for each retained wild field patch, the spatial intersection relationship between the buffer zones of the wild field patch and a crop field patch corresponding to the potential wild field-crop field junction is detected, and an intersected crop field patch is retained, and for an outer expansion annular buffer zone of the intersected crop field patch, if the number of the quadrant sub-regions with the natural vegetation coverage greater than a fifth threshold value meets a criterion, the potential wild field-crop field junction is determined as the interface wild field-crop field junction.

[0103] The step of image segmentation comprises:

[0104] For the target area remote sensing image, feature extraction is performed through standard convolution to obtain high-resolution features;

[0105] For the target area remote sensing image, feature extraction is performed through standard convolution to obtain high-resolution features;

[0106] After upsampling the down-sampling features, the normalized vegetation index is calculated to obtain enhanced features, and the gray level co-occurrence matrix is applied to the enhanced features to calculate the texture roughness and near-infrared reflection intensity to obtain texture features and reflection features;

[0107] After feature fusion of the high-resolution features, the down-sampling features, the enhanced features, the texture features and the reflection features, the shrubs, the mosses, the mangroves, the grasslands, the herbaceous wetlands and the forest regions are obtained through the classifier;

[0108] For the target area remote sensing image, the normalized water index is calculated, the region with the normalized water index exceeding a set value is regarded as the herbaceous wetland, and the intersection between the herbaceous wetland and the herbaceous wetland obtained by the classifier is obtained to obtain the final herbaceous wetland;

[0109] The shrubs, the mosses, the mangroves, the grasslands, the herbaceous wetlands and the forest regions are merged into the wild field region.

[0110] For the target area remote sensing image, the region other than the wild field region is regarded as the crop field pre-classification region.

[0111] For the crop field pre-classification region, the direction gradient histogram is used to quantify the consistency of the field ridge direction, and the crop field region is screened out.

[0112] It should be noted that each module in the embodiment corresponds to each step in Embodiment One, and the specific implementation process is the same, which will not be repeated here.

[0113] Example 3

[0114] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the field-farmland boundary zoning method described in Embodiment 1 above.

[0115] Example 4

[0116] This embodiment provides a computer device, such as... Figure 4 As shown, the system includes a computer-readable storage medium 1003, a processor 1001, a communication interface 1002, and a computer program stored on the computer-readable storage medium 1003 and executable on the processor 1001. The processor 1001, communication interface 1002, and computer-readable storage medium 1003 can be connected via a bus or other means. The communication interface 1002 is used to receive and send data. When the processor 1001 executes the program, it implements the steps in the field-farmland boundary zoning method described in Embodiment 1 above.

[0117] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method of dividing a wild-field interface area, characterized by, The method comprises the following steps: acquiring a remote sensing image of a target region, obtaining a wild region and a farmland region by image segmentation, and encoding the wild region and the farmland region into grids respectively; converting the farmland grid into a vector polygon to obtain a plurality of farmland patches, filtering out farmland patches with an area less than a first threshold value, constructing an outer expansion annular buffer zone for each retained farmland patch, calculating a natural vegetation coverage of the outer expansion annular buffer zone of each farmland patch, and marking an outer expansion annular buffer zone with a natural vegetation coverage greater than a second threshold value as a potential wild-farmland interface region; dividing the potential wild-farmland interface region into four quadrants with the geometric center of the farmland patch as the origin, calculating the natural vegetation coverage of each quadrant sub-region, and identifying a mixed wild-farmland interface region and an interface wild-farmland interface region in the potential wild-farmland interface region according to a quadrant-level natural vegetation coverage threshold value and a quadrant compliance number constraint; wherein the natural vegetation coverage is a ratio of the number of wild grids in a region to the total number of grids in the region.

2. A method of dividing a wild-field interface zone according to claim 1, characterized in that The step of identifying the mixed wild-farmland interface region and the interface wild-farmland interface region in the potential wild-farmland interface region comprises: for a certain potential wild-farmland interface region, if the number of quadrant sub-regions with a natural vegetation coverage greater than a third threshold value meets the standard, the potential wild-farmland interface region is determined as a mixed wild-farmland interface region; otherwise, the wild grid is converted into a vector polygon to obtain a plurality of wild patches, wild patches with an area less than a fourth threshold value are filtered out, a buffer zone is constructed for each retained wild patch, the spatial intersection relationship between the buffer zones of the wild patches and the farmland patches corresponding to the potential wild-farmland interface region is detected, and the intersecting farmland patches are retained; for an outer expansion annular buffer zone of a certain intersecting farmland patch, if the number of quadrant sub-regions with a natural vegetation coverage greater than a fifth threshold value meets the standard, the potential wild-farmland interface region is determined as an interface wild-farmland interface region.

3. The method for delineating the boundary area between farmland and field as described in claim 1, characterized in that, The step of image segmentation comprises: performing feature extraction on the target region remote sensing image by standard convolution to obtain high-resolution features; performing feature extraction on the target region remote sensing image after down-sampling by hole convolution to obtain down-sampled features; after up-sampling the down-sampled features, calculating a normalized vegetation index to obtain enhanced features, and applying a gray level co-occurrence matrix to the enhanced features to calculate texture roughness and near-infrared reflectivity to obtain texture features and reflectivity features; after feature fusion of the high-resolution features, the down-sampled features, the enhanced features, the texture features, and the reflectivity features, the classifier obtains shrub, moss, mangrove, grassland, herbaceous wetland, and forest regions.

4. A method of dividing a wild-field interface zone according to claim 3, characterized in that The step of image segmentation further comprises: for the target region remote sensing image, calculating a normalized water index, regarding a region with a normalized water index exceeding a set value as herbaceous wetland, and performing an intersection operation between the herbaceous wetland obtained by the classifier and the herbaceous wetland to obtain the final herbaceous wetland; merging the shrub, moss, mangrove, grassland, herbaceous wetland, and forest regions into a wild region; for the target region remote sensing image, regarding a region other than the wild region as a farmland pre-classification region.

5. A method of dividing a wild-field interface zone according to claim 4, characterized in that The image segmentation step further comprises: pre-classifying the farmland, quantifying the field ridge direction consistency based on a histogram of oriented gradients, and screening out the farmland region.

6. A wild-field interface zone zoning system characterized by, The method comprises: an image segmentation module configured to: acquire a remote sensing image of a target region, obtain a wildland region and a farmland region through image segmentation, and encode the wildland region and the farmland region into grids respectively; a preliminary division module configured to: convert the farmland grid into a vector polygon to obtain a plurality of farmland patches, filter out farmland patches with an area less than a first threshold, construct an outer expansion annular buffer zone for each retained farmland patch, calculate the natural vegetation coverage of the outer expansion annular buffer zone of each farmland patch, and mark an outer expansion annular buffer zone with a natural vegetation coverage greater than a second threshold as a potential wildland-farmland interface region; a final division module configured to: divide the potential wildland-farmland interface region into four quadrants with the geometric center of the farmland patch as the origin, calculate the natural vegetation coverage of each quadrant sub-region, and identify a mixed wildland-farmland interface region and an interface wildland-farmland interface region in the potential wildland-farmland interface region according to a quadrant-level natural vegetation coverage threshold and a quadrant compliance number constraint; wherein the natural vegetation coverage is the ratio of the number of wildland grids in a region to the total number of grids in the region.

7. A wild-field interface zone system according to claim 6, c h a r a c t e r i z e d in that The step of identifying the mixed wildland-farmland interface region and the interface wildland-farmland interface region in the potential wildland-farmland interface region comprises: for a certain potential wildland-farmland interface region, if the number of quadrant sub-regions with a natural vegetation coverage greater than a third threshold meets the standard, the potential wildland-farmland interface region is determined as a mixed wildland-farmland interface region; otherwise, the wildland grid is converted into a vector polygon to obtain a plurality of wildland patches, wildland patches with an area less than a fourth threshold are filtered out, a buffer zone is constructed for each retained wildland patch, the spatial intersection relationship between the buffer zones of the wildland patches and the farmland patches corresponding to the potential wildland-farmland interface region is detected, and the intersecting farmland patches are retained; for the outer expansion annular buffer zone of a certain intersecting farmland patch, if the number of quadrant sub-regions with a natural vegetation coverage greater than a fifth threshold meets the standard, the potential wildland-farmland interface region is determined as an interface wildland-farmland interface region.

8. A wild-field interface zone system according to claim 6, characterized in that The image segmentation step comprises: performing feature extraction on the target region remote sensing image through standard convolution to obtain high-resolution features; performing feature extraction on the target region remote sensing image after down-sampling through a dilated convolution to obtain down-sampled features; performing up-sampling on the down-sampled features, calculating a normalized vegetation index to obtain enhanced features, and applying a gray level co-occurrence matrix to the enhanced features to calculate texture roughness and near-infrared reflectance intensity to obtain texture features and reflectance features; performing feature fusion on the high-resolution features, the down-sampled features, the enhanced features, the texture features, and the reflectance features, and obtaining shrub, moss, mangrove, grassland, herbaceous wetland, and forest regions through a classifier.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The program, when executed by a processor, implements the steps of a wildland-farmland interface region division method according to any one of claims 1-5.

10. A computer device, comprising a computer readable storage medium, a processor, and a computer program stored on the computer readable storage medium and executable on the processor, wherein, The processor implements the steps of a wild field-farmland interface area division method as claimed in any one of claims 1-5 when executing the program.

Images