Method and device for evaluating regional ecological environment quality, equipment and storage medium

Abstract

Embodiments of the present application relate to the field of ecological environment assessment, and particularly relate to a regional ecological environment quality assessment method and device, equipment and a storage medium. The method determines a plurality of element images according to remote sensing data of a target region. A convolution scale corresponding to each pixel position in each element image is determined according to a category image of the target region, to obtain a convolution scale map of each element image. The category image is used to represent the topographic category of each pixel position in the target region. Each element image is subjected to convolution processing according to the corresponding convolution scale map, to obtain an element convolution result. An assessment report of the target region is generated according to the element convolution results. Embodiments of the present application can improve the fidelity of spatial geometric shapes and the expression stability of ecological elements according to different topographic categories in the target region, and can match different convolution kernels for different pixel positions for different element images according to the topographic categories, to improve the explainability, reliability and cross-region applicability of the assessment results as a whole.

Description

Technical Field

[0001] This application relates to the field of ecological environment assessment, and includes, but is not limited to, a method, apparatus, equipment, and storage medium for assessing regional ecological environment quality. Background Technology

[0002] Remote sensing ecological indices (RSEIs) are a commonly used method for comprehensively evaluating regional ecological environment quality using remote sensing data. A typical RSEI consists of four sub-indices: Greenness (commonly NDVI), Wetness (commonly Tasseled Cap Wetness / TCW or other humidity indices), Dryness (commonly NDBSI, a combination of bare soil index and built-up area index), and Heat (commonly LST). These four factors are typically normalized and their positive / negative values ​​are unified. Principal component analysis (PCA) is then used to extract the first principal component as the comprehensive ecological index. Due to its simple index construction and wide applicability, RSEIs are widely used in urban ecological quality assessment, land use / cover change impact assessment, regional ecological red line monitoring, and temporal ecological evolution analysis. When using convolutional methods to extract features from the four types of factors mentioned above, related technologies suffer from several drawbacks. These include scale mismatch and boundary distortion caused by a uniform convolutional kernel and the lack of spatial processing based on land type conditions; ineffective handling of the fusion mismatch between multi-source resolution and heterogeneous data; and unstable fusion of thermal and spectral factors. Consequently, the RSEI evaluation results are inaccurate and lack interpretability. Summary of the Invention

[0003] In view of this, the regional ecological environment quality assessment method, apparatus, equipment, and storage medium provided in the embodiments of this application improve the interpretability, reliability, and cross-regional applicability of the ecological assessment results.

[0004] The regional ecological environment quality assessment method, apparatus, equipment, and storage medium provided in this application embodiment are implemented as follows:

[0005] One aspect of this application provides a method for assessing regional ecological environment quality, the method comprising:

[0006] Multiple feature images are determined based on remote sensing data of the target area;

[0007] Based on the category image of the target region, the convolution scale corresponding to each pixel position in each feature image is determined, and the convolution scale map of each feature image is obtained. The category image is used to characterize the landform category of each pixel position in the target region.

[0008] The convolution process is performed on each feature image according to the corresponding convolution scale map to obtain the feature convolution result;

[0009] An evaluation report for the target region is generated based on the convolution results of each element.

[0010] In one possible implementation, multiple feature images are determined based on remote sensing data of the target area, including:

[0011] Image preprocessing is performed on the remote sensing data of the target area;

[0012] Based on the preprocessed remote sensing data, determine the element images corresponding to multiple preset element dimensions.

[0013] In one possible implementation, the convolution scale corresponding to each pixel position in each feature image is determined based on the category image of the target region, resulting in a convolution scale map of each feature image, including:

[0014] Determine the set of convolution scales corresponding to the landform categories included in the categorized images;

[0015] Determine the terrain category of each pixel location in each feature image based on the category image;

[0016] For each feature image, the convolution scale corresponding to the pixel location is determined in the convolution scale set according to the corresponding scale selection rules and the terrain category of each pixel location, thus obtaining the convolution scale map of the feature image.

[0017] In one possible implementation, for each feature image, the convolution scale corresponding to the pixel location is determined from the convolution scale set according to the corresponding scale selection rule and the terrain category of each pixel location, resulting in a convolution scale map of the feature image, including:

[0018] For each feature image, candidate scales are selected from the set of convolution scales corresponding to each pixel location;

[0019] Based on the scale selection rules corresponding to the feature image, the convolution scale is determined from the convolution scale set according to the candidate scale corresponding to each pixel position, thus obtaining the convolution scale map of the feature image.

[0020] In one possible implementation, the feature images include greenness images, dryness images, humidity images, and heat images. For each feature image, candidate scales for each pixel location are selected from the set of convolution scales corresponding to that pixel location, including:

[0021] For greenness images, dryness images, and humidity images, a preset baseline scale is selected as the candidate scale for each pixel location from the set of convolution scales.

[0022] For thermal images, candidate scales are selected from the set of convolution scales corresponding to each pixel location based on preset spatial support conditions.

[0023] In one possible implementation, for each feature image, the convolution scale corresponding to the pixel location is determined from the convolution scale set according to the corresponding scale selection rule and the terrain category of each pixel location, resulting in a convolution scale map of the feature image, further comprising:

[0024] Consistency correction is performed on the convolution scale map corresponding to each feature image.

[0025] In one possible implementation, an evaluation report of the target region is generated based on the convolution results of each feature, including:

[0026] Principal component analysis was performed based on the convolution results of each element to obtain the comprehensive ecological quality score;

[0027] An assessment report for the target area is generated based on the images of each element, the convolution results of each element, and the comprehensive ecological quality score of each element.

[0028] Another aspect of this application embodiment provides a regional ecological environment quality assessment device, the device comprising:

[0029] The image acquisition module is used to determine multiple feature images based on remote sensing data of the target area;

[0030] The scale determination module is used to determine the convolution scale corresponding to each pixel position in each feature image based on the category image of the target region, and obtain the convolution scale map of each feature image. The category image is used to characterize the landform category of each pixel position in the target region.

[0031] The convolution processing module is used to perform convolution processing on each feature image according to the corresponding convolution scale map to obtain the feature convolution result;

[0032] The quality assessment module is used to generate an assessment report for the target area based on the convolution results of each element.

[0033] In one possible implementation, the image acquisition module is further used for:

[0034] Image preprocessing is performed on the remote sensing data of the target area;

[0035] Based on the preprocessed remote sensing data, determine the element images corresponding to multiple preset element dimensions.

[0036] In one possible implementation, the scale determination module is further used for:

[0037] Determine the set of convolution scales corresponding to the landform categories included in the categorized images;

[0038] Determine the terrain category of each pixel location in each feature image based on the category image;

[0039] For each feature image, the convolution scale corresponding to the pixel location is determined in the convolution scale set according to the corresponding scale selection rules and the terrain category of each pixel location, thus obtaining the convolution scale map of the feature image.

[0040] In one possible implementation, the scale determination module is further used for:

[0041] For each feature image, candidate scales are selected from the set of convolution scales corresponding to each pixel location;

[0042] Based on the scale selection rules corresponding to the feature image, the convolution scale is determined from the convolution scale set according to the candidate scale corresponding to each pixel position, thus obtaining the convolution scale map of the feature image.

[0043] In one possible implementation, the feature images include greenness images, dryness images, humidity images, and heat images; the scale determination module is further used for:

[0044] For greenness images, dryness images, and humidity images, a preset baseline scale is selected as the candidate scale for each pixel location from the set of convolution scales.

[0045] For thermal images, candidate scales are selected from the set of convolution scales corresponding to each pixel location based on preset spatial support conditions.

[0046] In one possible implementation, the scale determination module is further used for:

[0047] Consistency correction is performed on the convolution scale map corresponding to each feature image.

[0048] In one possible implementation, the quality assessment module is further used for:

[0049] Principal component analysis was performed based on the convolution results of each element to obtain the comprehensive ecological quality score;

[0050] An assessment report for the target area is generated based on the images of each element, the convolution results of each element, and the comprehensive ecological quality score of each element.

[0051] The electronic device provided in this application includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the program, it implements the method described in this application.

[0052] The computer-readable storage medium provided in this application embodiment stores a computer program thereon, which, when executed by a processor, implements the method provided in this application embodiment.

[0053] In this embodiment, the method determines multiple element images based on remote sensing data of the target area. The convolution scale corresponding to each pixel location in each element image is determined based on the category image of the target area, resulting in a convolution scale map for each element image. The category image is used to characterize the landform category at each pixel location in the target area. Convolution processing is performed on each element image according to the corresponding convolution scale map to obtain the element convolution result. An evaluation report for the target area is generated based on the element convolution results. This embodiment can improve the fidelity of spatial geometry and the stability of ecological element representation based on different landform categories in the target area. Furthermore, it can selectively match convolution kernels for different pixel locations based on landform categories, thereby improving the interpretability, reliability, and cross-regional applicability of the evaluation results. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 A flowchart illustrating a regional ecological environment quality assessment method according to an embodiment of this application is shown.

[0056] Figure 2 A schematic diagram illustrating a regional ecological environment quality assessment process according to an embodiment of this application is shown;

[0057] Figure 3 A schematic diagram of a regional ecological environment quality assessment device according to an embodiment of this application is shown;

[0058] Figure 4 A schematic diagram of an electronic device according to an embodiment of this application is shown. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the specific technical solutions of this application will be further described in detail below with reference to the accompanying drawings of the embodiments of this application. The following embodiments are used to illustrate this application, but are not intended to limit the scope of this application.

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0061] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0062] It should be noted that the terms "first, second, third" used in the embodiments of this application are used to distinguish similar or different objects and do not represent a specific order of objects. It can be understood that "first, second, third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0063] The regional ecological environment quality assessment method of this application embodiment can be executed by any electronic device, including but not limited to mobile phones, wearable devices (such as smartwatches, smart bracelets, smart glasses, etc.), tablet computers, laptops, vehicle terminals, PCs (Personal Computers), etc. The functions implemented by this method can be achieved by the processor in the electronic device calling program code. Of course, the program code can be stored in a computer storage medium. It can be seen that the electronic device includes at least a processor and a storage medium.

[0064] The regional ecological environment quality assessment method of this application can be used in any application scenario that requires an assessment of the ecological environment quality of a region. For example, this application embodiment can be applied to urban ecological quality assessment and monitoring in the application scenario of urban development and planning management. Alternatively, it can also be applied to dynamic monitoring of nature reserves / national parks in the application scenario of ecosystem protection and restoration.

[0065] Current regional ecological environment quality assessments are typically conducted by calculating the Regional Remote Sensing Ecological Index (RSEI), which uses remote sensing data to comprehensively evaluate the regional ecological environment quality. A typical RSEI consists of four sub-indices: Greenness (commonly NDVI), Wetness (commonly Tasseled Cap Wetness / TCW or other humidity indices), Dryness (commonly NDBSI, a combination of bare soil index and built-up area index), and Heat (commonly LST). These four factors are usually normalized and their positive / negative values ​​are unified. Principal component analysis (PCA) is then used to extract the first principal component as the comprehensive ecological index. Due to its simple index construction and wide applicability, the RSEI is widely used in urban ecological quality assessment, land use / cover change impact assessment, regional ecological red line monitoring, and temporal ecological evolution analysis.

[0066] The data sources for remote sensing ecological indices typically include multispectral (such as Landsat and Sentinel-2) and thermal infrared (for LST inversion), and require preprocessing such as atmospheric correction, cloud shadow masking, resampling and registration, and temporal synthesis. To suppress stripes, speckles, and random noise, traditional workflows often employ spatial smoothing techniques (such as mean filtering, Gaussian filtering, and fixed-size convolution / window operations), and these often use globally uniform window or convolution kernel sizes.

[0067] In related technologies, numerous studies have attempted to replace or extend the four factors (such as introducing Albedo, soil moisture index, vegetation cover, etc.), adjust weights (PCA variants, entropy weights, random forest weights), or perform RSEI calculations by region / land type (statistical analysis and fusion separately according to functional zone or land cover type). However, in spatial processing, the vast majority of works still use globally uniform smoothing or fixed convolution kernel sizes. In the fields of remote sensing semantic segmentation and object detection, methods such as multi-scale convolution, dynamic convolution, deformable convolution, and selective convolution kernels have been used to improve the adaptability of feature extraction. However, these methods are mostly used for feature learning in recognition and segmentation tasks, and are rarely directly embedded into the sub-index spatial processing and fusion stages of RSEI, and a standard procedure of "adaptively selecting convolution kernel size according to land cover type" has not yet been formed.

[0068] Meanwhile, LST often has resolution differences with multispectral data. Traditional approaches often use uniform resampling and uniform window smoothing, which lacks targeted processing for the scale differences and noise characteristics of different land cover types.

[0069] Therefore, it can be seen that the relevant technologies have the following technical problems in assessing the quality of the ecological environment by calculating the remote sensing ecological index of the region:

[0070] The use of uniform convolution kernels and spatial processing lacking land use classification leads to scale mismatch and boundary distortion. Existing methods often smooth all land use types with a single, globally uniform convolution kernel size, ignoring the significant differences in spatial texture, structural scale, and noise spectrum among forests, grasslands, farmland, water bodies, and built-up areas. The result is that in high-texture and hard-boundary areas such as urban edges, street networks, shorelines, and waterfront transitions, uniform kernels become overly smoothed, smoothing out details, widening boundaries, and even causing "overflowing colors." Conversely, in high-frequency speckled areas such as deserts / bare land, uniform kernels tend to be undersmoothed, leaving behind salt-pepper noise and random undulations, disrupting spatial continuity and the interpretability of ecological semantics. Furthermore, spatial convolution does not adapt to land use classification, easily producing a "mixed pixel" effect and blurred boundaries at category boundaries and transition zones, causing RSEI to exhibit evaluation distortion, hotspot misjudgment, and local instability in highly heterogeneous areas.

[0071] The mismatch between multi-source resolution and heterogeneous data fusion was not effectively addressed, leading to unstable fusion of thermal and spectral factors. RSEI typically fuses LST (thermal infrared) and multispectral sub-indices (NDVI / humidity / dryness), which exhibit systematic differences in spatial resolution, noise spectrum, and sampling support (e.g., LST has coarser resolution and stronger thermal noise and field-of-view effects). Uniform spatial convolution cannot select the optimal smoothing scale for thermal and spectral factors according to land type, nor can it align the "effective spatial support" of the two types of data. This introduces artifacts and misaligned hotspots during the fusion stage, disrupting the stability of the covariance structure, causing PCA load direction fluctuations, principal component instability, and reduced comparability of time series and cross-regional comparisons, ultimately affecting the robustness and application reliability of RSEI.

[0072] Therefore, the technical problem solved by the embodiments of this application is how to improve the interpretability, reliability and cross-regional applicability of ecological assessment results.

[0073] The regional ecological environment quality assessment scheme of this application embodiment will be described in detail below with reference to the accompanying drawings.

[0074] Figure 1 A flowchart illustrating a regional ecological environment quality assessment method according to an embodiment of this application is shown. Figure 1 As shown, the regional ecological environment quality assessment method of this application embodiment may include the following steps S10-S40.

[0075] For ease of description, the regional ecological environment quality assessment method of this application embodiment is described using an electronic device as the execution subject. It should be understood that the execution subject of this application embodiment can also be a processor or chip in an electronic device, and this application embodiment does not impose any limitations.

[0076] Step S10: Determine multiple feature images based on remote sensing data of the target area.

[0077] In one possible implementation, electronic devices can assess the ecological environment quality of a target area by determining a remote sensing ecological index through acquiring remote sensing data of the target area. The remote sensing data can include multiple images of the target area acquired simultaneously at different times using different bands, used to determine corresponding element images across multiple preset element dimensions. Each element image corresponds to a specific element dimension, representing the distribution of features of the target area within that element dimension.

[0078] For example, the remote sensing data in this application embodiment may include images such as visible light images, near-infrared light images, and thermal infrared light images of the target area. These can be acquired, for instance, using Landsat 8 / 9 OLI / TIRS, Sentinel-2 with its accompanying LST product, or acquired using a land surface temperature inversion method paired with a TIRS (thermal infrared sensor). These multiple feature images may include multiple images used to calculate remote sensing ecological indices. That is, the feature dimensions in this application embodiment may include greenness, dryness, humidity, and heat, with corresponding greenness images, dryness images, humidity images, and heat images determined respectively.

[0079] In some embodiments, remote sensing data may be subject to interference noise during the acquisition process due to issues such as cloud cover. Therefore, before determining multiple feature images based on remote sensing data, electronic devices can first perform image preprocessing on the remote sensing data of the target area, and then determine the feature images corresponding to multiple preset feature dimensions based on the image preprocessed remote sensing data.

[0080] Optionally, this preprocessing process may include radiometric calibration, topographic correction, and atmospheric correction of the remote sensing data, while identifying cloud-obscured areas as invalid regions using cloud masking. For example, radiometric calibration can be performed using LEDAPS or LaSRC, topographic correction can be performed using DEM, atmospheric correction can be performed using Sen2Cor (a dedicated atmospheric correction processor for Sentinel-2), and invalid regions can be identified using Fmask. Simultaneously, the preprocessing process can also perform geometrical fine correction and resampling to a uniform resolution for each image data. Furthermore, invalid regions identified after cloud masking can be directly marked as invalid or interpolated for completion.

[0081] In some embodiments, after preprocessing remote sensing data, the electronic device can directly calculate multiple element images based on the element images corresponding to each element dimension, using the rule for determining these images. For example, the rule for determining the greenness image (G) can be: NDVI = (NIR - Red) / (NIR + Red), where NIR is the near-infrared band and Red is the red band. The rule for determining the humidity image (W) can be: using the Tasseled Cap humidity component (Landsat coefficient) or employing NDWI / modified humidity index; prioritizing the Tasseled Cap W component to balance physical meaning and stability. The rule for determining the dryness image (D) can be: NDBSI (combining the bare soil index SI and the built-up area index IBI, or directly using a weighted average of NDBI and SI); ensuring a negative correlation with ecological quality. The rule for determining the thermal image (L) can be: surface temperature LST (estimated by combining surface emissivity with the TIRS single-window or split-window algorithm).

[0082] In other embodiments, the electronic device in this application can first determine candidate element images corresponding to multiple element dimensions using the above method, and then perform pre-standardization and polarity checks on each candidate element image to obtain element images that pass rapid quality inspection. The pre-standardization process can involve normalizing each candidate element image to ensure that the pixel values ​​at each pixel position are between [0,1]. The polarity check process also includes polarity correction, used to check that greenness and humidity images are positively correlated with ecological quality, while dryness and heat images are negatively correlated with ecological quality. For element images that fail the polarity check, polarity correction can be performed by inverting the polarity.

[0083] Step S20: Determine the convolution scale corresponding to each pixel position in each element image based on the category image of the target region, and obtain the convolution scale map of each element image.

[0084] In one possible implementation, the electronic device acquires a category image of the target region, where each pixel in the category image indicates the terrain category of the corresponding location within the target region. The resolution of the category image can be the same as that of each feature image, used to indicate the terrain category corresponding to each pixel location in each feature image. Based on the corresponding terrain category, the convolution scale corresponding to each pixel location in each feature image is determined, resulting in a convolution scale map of the feature images.

[0085] In some embodiments, the electronic device can generate category images of the target area using land cover classification networks (such as DeepLabv3+, U-Net, SegFormer) or traditional classification methods (random forest, SVM). These land cover categories can include vegetation, water bodies, built-up land, bare land / desert, wetlands, farmland, woodland, roads, etc. Optionally, the electronic device can also calculate a boundary distance map based on the category images and set a boundary bandwidth b (2–3 pixels recommended); generate a mask for each category for intra-category k-smoothing and water / land separation processing. This category division process ensures an overall accuracy of ≥85% for the category images and good boundary quality. Morphological refinement is provided for slender features (roads, canals).

[0086] Optionally, in this embodiment, each landform category can be pre-defined with a corresponding set of convolutional scales, which includes multiple candidate convolutional scales. When determining the convolutional scale map corresponding to the feature image, the electronic device can match the corresponding set of convolutional scales based on the landform category at each pixel location, and select a convolutional scale from the corresponding set of convolutional scales. That is, the electronic device can first determine the set of convolutional scales corresponding to the landform categories included in the category image, and then determine the landform category at each pixel location in each feature image based on the category image. Then, for each feature image, the convolutional scale corresponding to the pixel location is determined from the set of convolutional scales according to the corresponding scale selection rules and the landform category at each pixel location, thus obtaining the convolutional scale map of the feature image.

[0087] For example, the landform categories in the embodiments of this application may include built-up land / roads, water bodies, vegetation (forests / grasslands / farmland), bare land / desert, bare rock / rocky mountains, sand dunes / aeolian sand, snow / glaciers, paddy fields, intertidal zones / mudflats, and urban green spaces / parks. Different landform categories can be pre-defined with corresponding convolution scale sets: Built-up land / roads: Kc={3,5,7}, Water bodies: Kc={5,7,9}, Vegetation (forest / grassland / farmland): Kc={7,9,11}, Bare land / desert: Kc={5,7,9}, Bare rock / rocky mountains: Kc={3,5,7}, Sand dunes / aeolian sand: Kc={7,9,11}, Snow / glaciers: Kc={7,9,11}, Paddy fields: Kc={5,7,9}, Intertidal zone / mudflats: Kc={3,5,7}, Urban green space / parks: Kc={3,5,7}.

[0088] Optionally, the set of convolutional scales for each landform category in this embodiment may further include at least one baseline scale, which is one of multiple convolutional scales and is used as the default convolutional scale. For example, the baseline scale for construction land / roads may be determined as k_c=3 or 5 (in case of preserving detail and preventing color overflow at the pixel location). The baseline scale for water bodies may be determined as k_c=5 or 7 (in case of uniformity within the pixel location and clear outline). The baseline scale for vegetation (forest / grass / farmland) may be determined as k_c=7 or 9 (in case of large patches at the pixel location and enhanced coherence). The baseline scale for bare land / desert may be determined as k_c=5 or 7. The baseline scale for bare rock / rocky mountains may be determined as k_c=5 to ensure the coexistence of hard boundaries and prevent widening. The baseline scale for sand dunes / aeolian sand may be determined as k_c=9. The baseline scale for snow / glaciers may be determined as k_c=9. The baseline scale for paddy fields may be determined as k_c=7. The baseline scale for intertidal zones / mudflats is determined to be k_c=5. The baseline scale for urban green spaces / parks is determined to be k_c=5.

[0089] In some embodiments, when determining the convolution scale map of a feature image, the electronic device can first obtain a set of convolution scales corresponding to each pixel location, select one convolution scale from it as a candidate scale, and then refine it based on the candidate scale according to a preset scale selection rule to accurately determine the corresponding convolution scale. That is, for each feature image, the electronic device selects a candidate scale for each pixel location from the set of convolution scales corresponding to each pixel location. Then, according to the scale selection rule corresponding to the feature image, it determines the convolution scale from the set of convolution scales based on the candidate scale for each pixel location, thus obtaining the convolution scale map of the feature image.

[0090] Optionally, for different types of feature images, the electronic device can directly select a baseline scale as a candidate scale from the set of convolutional scales corresponding to each pixel location, or select a candidate scale according to other rules. For example, in the case where the feature images in this embodiment include greenness images, dryness images, humidity images, and thermal images, for greenness images, dryness images, and humidity images, a preset baseline scale is selected as a candidate scale for each pixel location from the set of convolutional scales corresponding to each pixel location. For thermal images, a candidate scale for each pixel location is selected from the set of convolutional scales corresponding to each pixel location according to preset spatial support conditions.

[0091] Further, the spatial support conditions in this embodiment may include first calculating q = r_LST / r_MS using the target raster resolution r_MS and the native resolution r_LST of the thermal image, and then determining the candidate scale coverage ≈ q × q neighborhood (example: r_LST = 100 m, r_MS = 30 m, q ≈ 3.3 → k is 9 or 11). Then, a choice is made between the two scales using simplified adaptive and boundary protection methods. The simplified adaptive process may involve calculating the local variance or entropy (a 5×5 window is recommended, with thresholds set according to intra-class statistical quantiles), taking the intra-class 25th quantile q25 and 75th quantile q75. If the local variance ≤ q25 → use a larger kernel; ≥ q75 → use a smaller kernel; if it falls between these values, use the baseline scale. The boundary protection process may include using the class boundary distance d and the boundary bandwidth b (2–3 pixels), d ≤ b → force the use of the smallest kernel k_min in Kc; d > b → select k_c according to the above rules or fine-tune it. Further, through pixel-level... Perform 3×3 median smoothing to avoid abrupt changes and ensure that the scales of the four-factor convolution kernels of the same pixel do not differ by more than one level in the corresponding convolution scale set; at the same time, force k_LST to be no less than other factors.

[0092] In other embodiments, after determining the candidate scales for each pixel position in the feature image, the present application embodiments can correct the candidate scales using the scale selection rules corresponding to each feature image to obtain the corresponding convolution scale.

[0093] For example, when the element image in the embodiment of this application includes a greenness image, aridity image, humidity image and thermal image, the scale selection rule for the greenness image may include: when the pixel location belongs to the vegetation inner area with high NDVI and low texture, selecting a convolution scale one level higher based on the candidate scale in the convolution scale set; when the pixel location belongs to urban green space, broken vegetation or seasonal leaf fall, selecting a convolution scale one level lower based on the candidate scale in the convolution scale set; when the pixel location belongs to the waterfront or the boundary of built-up area, selecting a convolution scale one level lower based on the candidate scale in the convolution scale set.

[0094] The scale selection rules for humidity images may include: when the pixel location is within a wetland / paddy field area, selecting a convolution scale one level higher from the convolution scale set based on the candidate scale; when the pixel location is near a water body boundary (d≤b) or a mixed pixel, selecting a convolution scale one level lower from the convolution scale set based on the candidate scale; when the pixel location is in a dry area with high noise (wind and sand, thin clouds), using the candidate scale as the corresponding convolution scale.

[0095] The scale selection rules for dryness images can include: when the pixel location belongs to a strip of bare farmland / cultivated texture area, select a convolution scale that is one level higher from the convolution scale set based on the candidate scale; when the pixel location belongs to a high-D and high-gradient area in a built-up area, select a convolution scale that is one level lower from the convolution scale set based on the candidate scale; when the pixel location belongs to a micro-textured area in the hinterland of a desert, select a convolution scale that is one level higher from the convolution scale set based on the candidate scale; when the pixel location belongs to a high-contrast fissure in a rocky mountain, select a convolution scale that is one level lower from the convolution scale set based on the candidate scale.

[0096] The scale selection rules for thermal images may include: when the pixel location is in a uniform open area (lake center, desert hinterland), select a convolution scale that is one level higher from the convolution scale set based on the candidate scale; when the pixel location is in an industrial heat island / point heat source or shoreline thermal gradient area, select a convolution scale that is one level lower from the convolution scale set based on the candidate scale.

[0097] In another embodiment, after determining the convolution scale map corresponding to each feature image, the electronic device can further improve the accuracy of the final convolution result by performing consistency correction based on the convolution scale maps of different feature images. This consistency correction process may include performing a consistency check on each convolution scale map to ensure that the convolution scale corresponding to the heat image at the same pixel location is the largest convolution scale among multiple feature images, and that the difference in convolution scales corresponding to different feature images at the same pixel location does not exceed one level within the convolution scale set. Furthermore, the electronic device can also perform 3×3 median smoothing on each convolution scale map to prevent abrupt changes in convolution scale while avoiding scale jumps and cross-class convergence by performing the smoothing only within the same type of mask (not across classes).

[0098] Step S30: Perform convolution processing on each of the feature images according to the corresponding convolution scale map to obtain the feature convolution result.

[0099] In one possible implementation, after determining the convolution scale map corresponding to each element image, the electronic device performs Gaussian convolution on each element image in parallel based on the convolution scale map of each element image to obtain the element convolution result corresponding to each element image.

[0100] Step S40: Generate an evaluation report for the target region based on the convolution results of each of the aforementioned elements.

[0101] In one possible implementation, after completing the convolution of each element image to obtain multiple convolution scale maps, the electronic device can generate an assessment report of the target region based on the convolution results of each element. The assessment report also includes a comprehensive ecological quality score calculated using remote sensing ecological indices based on the aforementioned element convolution results. That is, the electronic device can perform principal component analysis based on the element convolution results to obtain a comprehensive ecological quality score. Then, it generates an assessment report of the target region based on the element images, the element convolution results, and the comprehensive ecological quality scores.

[0102] In some embodiments, the principal component analysis (PCA) process may include uniformly performing Z-score or robust Z-score (MAD) standardization on the convolution results of each element and aligning their polarities. The processed convolution results are then subjected to PCA, and the first principal component, PC1, is taken as the comprehensive ecological quality score. The electronic device can also check the sign based on the component loading: if PC1 is opposite to the ecological quality direction, it is negative, and then linearly normalized to [0, 1] to obtain the remote sensing ecological index. Furthermore, if PCA is sensitive to anomalies during PCA, robust PCA or factor analysis can be used instead; alternatively, correlation constraints can be applied to PCA pairs, with the convolution results of elements corresponding to greenness and humidity images contributing positively, and the convolution results of elements corresponding to dryness and heat images contributing negatively.

[0103] Optionally, after the above calculations are completed, morphological closing operations can be performed on the small holes at each pixel location marked as invalid, and invalid values ​​can be filled in through neighborhood interpolation. Simultaneously, the electronic device can also perform scale consistency checks (multi-time comparison) and calculate the correlation with external ecological indicators (such as land cover rate and surface moisture sampling points). After completing the above processing, an assessment report for the target area is generated based on the images of each element, the convolution results of each element, and the comprehensive ecological quality score. The report outputs the images of each element, the convolution results of each element, the comprehensive ecological quality score, and the assessment report for the target area. It can also output the sensor information and parameters used to acquire the remote sensing data, as well as the date the assessment report was generated.

[0104] Figure 2 This diagram illustrates a regional ecological environment quality assessment process according to an embodiment of this application. Figure 2 As shown, in this embodiment of the application, when assessing the ecological environment quality of a target area, remote sensing data of the target area can be acquired first. Based on the remote sensing data, greenness images, dryness images, humidity images, and thermal images are determined, along with category images (land cover category maps) representing the landform categories at various locations within the target area. Further, the correspondence between each landform type in the category images and the convolutional scale set is determined. Simultaneously, after normalizing the greenness, dryness, humidity, and thermal images, the electronic device determines the convolutional scale map of each element image based on the correspondence between each landform type in the category images and the convolutional scale set. Then, based on the convolutional scale maps of each element image, PCA principal component analysis, post-processing, and quality control are performed to obtain the Remote Sensing Ecological Index (RSEI), thereby generating an ecological environment quality assessment report for the target area.

[0105] Based on the aforementioned technical features, the embodiments of this application can adaptively adjust the spatial scale according to the category, balancing detail preservation and boundary realism. Specifically, by actively selecting or weighting different convolutional kernel sizes based on the land cover category and its typical texture scale, each sub-index is smoothed and enhanced at the "most suitable" spatial scale. Small kernels are preferentially used for areas with high texture and complex boundaries, such as built-up areas, shorelines, and roads. This suppresses local random fluctuations while maximizing the preservation of detailed structures and clear boundaries, significantly reducing edge smoothing and "overflowing" effects caused by cross-class mixing. Medium to large kernels are used for blocky, structurally continuous areas such as forests and cultivated land to improve patch consistency and regional coherence, avoiding mottled and fragmented appearances caused by excessively small windows. Overall, this category-matching scale strategy simultaneously improves the fidelity of spatial geometry and the stability of ecological element representation, providing clearer and more reliable input features for subsequent RSEI fusion. Furthermore, the robust denoising and multi-source resolution collaboration in this application for complex scenes enhance fusion stability and spatiotemporal consistency. Specifically, for areas with high-frequency spots and significant random undulations, such as bare land / deserts, a larger kernel or multi-scale weighting strategy is adopted to effectively suppress salt-pepper noise, reduce isolated small spots, and improve the spatial continuity and statistical robustness of sub-indices. Simultaneously, an "effective spatial support" alignment strategy and a category-adaptive window are introduced to mitigate resolution differences and sampling mismatches between land surface temperature (LST) and multispectral indicators, reducing artifacts and misaligned hotspots caused by the mixing of coarse and fine resolutions. After this processing, the covariance structure of the thermal factor, greenness, humidity, and aridity in PCA is more stable, the principal component direction is clearer, the fusion results are less sensitive to outliers and noise, and the time-series comparisons are more consistent, thus improving the overall interpretability, reliability, and cross-regional applicability of the assessment results.

[0106] It should be understood that although the steps in the above flowcharts are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the above flowcharts may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0107] Based on the foregoing embodiments, this application provides a regional ecological environment quality assessment device. The device includes various modules and units included in each module, which can be implemented by a processor; of course, it can also be implemented by specific logic circuits. In the implementation process, the processor can be a central processing unit (CPU), microprocessor (MPU), digital signal processor (DSP), or field programmable gate array (FPGA), etc.

[0108] Figure 3 A schematic diagram of a regional ecological environment quality assessment device according to an embodiment of this application is shown. Figure 3 As shown, the regional ecological environment quality assessment device in this application embodiment includes:

[0109] Image acquisition module 30 is used to determine multiple feature images based on remote sensing data of the target area;

[0110] The scale determination module 31 is used to determine the convolution scale corresponding to each pixel position in each feature image based on the category image of the target region, so as to obtain the convolution scale map of each feature image. The category image is used to characterize the landform category of each pixel position in the target region.

[0111] The convolution processing module 32 is used to perform convolution processing on each feature image according to the corresponding convolution scale map to obtain the feature convolution result;

[0112] The quality assessment module 33 is used to generate an assessment report for the target area based on the convolution results of each element.

[0113] In one possible implementation, the image acquisition module 30 is further used for:

[0114] Image preprocessing is performed on the remote sensing data of the target area;

[0115] Based on the preprocessed remote sensing data, determine the element images corresponding to multiple preset element dimensions.

[0116] In one possible implementation, the scale determination module 31 is further used for:

[0117] Determine the set of convolution scales corresponding to the landform categories included in the categorized images;

[0118] Determine the terrain category of each pixel location in each feature image based on the category image;

[0119] For each feature image, the convolution scale corresponding to the pixel location is determined in the convolution scale set according to the corresponding scale selection rules and the terrain category of each pixel location, thus obtaining the convolution scale map of the feature image.

[0120] In one possible implementation, the scale determination module 31 is further used for:

[0121] For each feature image, candidate scales are selected from the set of convolution scales corresponding to each pixel location;

[0122] Based on the scale selection rules corresponding to the feature image, the convolution scale is determined from the convolution scale set according to the candidate scale corresponding to each pixel position, thus obtaining the convolution scale map of the feature image.

[0123] In one possible implementation, the feature images include greenness images, dryness images, humidity images, and heat images, and the scale determination module 31 is further used for:

[0124] For greenness images, dryness images, and humidity images, a preset baseline scale is selected as the candidate scale for each pixel location from the set of convolution scales.

[0125] For thermal images, candidate scales are selected from the set of convolution scales corresponding to each pixel location based on preset spatial support conditions.

[0126] In one possible implementation, the scale determination module 31 is further used for:

[0127] Consistency correction is performed on the convolution scale map corresponding to each feature image.

[0128] In one possible implementation, the quality assessment module 33 is further used for:

[0129] Principal component analysis was performed based on the convolution results of each element to obtain the comprehensive ecological quality score;

[0130] An assessment report for the target area is generated based on the images of each element, the convolution results of each element, and the comprehensive ecological quality score of each element.

[0131] The descriptions of the above device embodiments are similar to those of the above method embodiments, and have similar beneficial effects. For technical details not disclosed in the device embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0132] It should be noted that, in the embodiments of this application... Figure 3The module division of the regional ecological environment quality assessment device shown is illustrative and represents only a logical functional division; in actual implementation, other division methods may be used. Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, exist as separate physical units, or be integrated into one unit with two or more units. The integrated units can be implemented in hardware, as software functional units, or a combination of both.

[0133] It should be noted that, in the embodiments of this application, if the above-described methods are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks. Thus, the embodiments of this application are not limited to any specific hardware and software combination.

[0134] Figure 4 A schematic diagram of an electronic device according to an embodiment of this application is shown. For example... Figure 4 As shown in the figure, this application provides an electronic device, which can be a server, and its internal structure diagram can be as follows. Figure 4 As shown, the electronic device includes a processor 420, a memory, and a transceiver 440 connected via a system bus 410. The processor 420 provides computing and control capabilities. The memory includes a non-volatile storage medium 431 and internal memory 432. The non-volatile storage medium 431 stores an operating system, computer programs, and a database. The internal memory 432 provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium 431. The database stores data. The transceiver 440 communicates with external terminals via a network connection. The computer program is executed by the processor 420 to implement the aforementioned methods.

[0135] This application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor 420, implements the steps of the method provided in the above embodiments.

[0136] This application provides a computer program product containing instructions that, when run on a computer, cause the computer to perform the steps in the method provided in the above-described method embodiments.

[0137] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0138] In one possible implementation, the shooting prompting device provided in this application can be implemented as a computer program, which can be configured as follows: Figure 4 The device operates on the electronic device shown. The memory of the electronic device can store the various program modules that make up the above-described apparatus. The computer program composed of the various program modules causes the processor 420 to execute the steps of the methods in the various embodiments of this application described in this specification.

[0139] It should be noted that the descriptions of the storage medium and device embodiments above are similar to the descriptions of the method embodiments above, and have similar beneficial effects. For technical details not disclosed in the storage medium, storage medium, and device embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0140] It should be understood that the phrases "one embodiment," "an embodiment," or "some embodiments" mentioned throughout the specification mean that a specific feature, structure, or characteristic related to an embodiment is included in at least one embodiment of this application. Therefore, phrases such as "in one possible implementation," "in one embodiment," or "in some embodiments" appearing throughout the specification do not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the above-described embodiments are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. The descriptions of the various embodiments above tend to emphasize the differences between the various embodiments; their similarities or commonalities can be referred to mutually, and for the sake of brevity, they will not be repeated here.

[0141] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that there can be three kinds of relationships. For example, object A and / or object B can represent three situations: object A exists alone, object A and object B exist simultaneously, and object B exists alone.

[0142] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0143] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple modules or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or modules can be electrical, mechanical, or other forms.

[0144] The modules described above as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules. They may be located in one place or distributed across multiple network units. Some or all of the modules may be selected to achieve the purpose of this embodiment according to actual needs.

[0145] In addition, each functional module in the various embodiments of this application can be integrated into one processing unit, or each module can be a separate unit, or two or more modules can be integrated into one unit; the integrated modules can be implemented in hardware or in the form of hardware plus software functional units.

[0146] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, read-only memory (ROM), magnetic disks, or optical disks.

[0147] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0148] The methods disclosed in the several method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0149] The features disclosed in the several product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0150] The features disclosed in the several method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method or device embodiments.

[0151] The above description is merely an embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A regional ecological environment quality assessment method, characterized in that, The method comprises: determining a plurality of element images according to remote sensing data of a target region; determining a convolution scale corresponding to each pixel position in each element image according to a category image of the target region, wherein the category image is used to represent a landform category of each pixel position in the target region, and a convolution scale map of each element image is obtained; performing convolution processing on each element image according to the corresponding convolution scale map to obtain an element convolution result; generating an evaluation report of the target region according to each element convolution result; the method comprises: determining a convolution scale set corresponding to a landform category included in the category image; determining a landform category of each pixel position in each element image according to the category image; for each element image, determining a convolution scale corresponding to each pixel position according to a corresponding scale selection rule and a landform category of each pixel position in the convolution scale set, and obtaining a convolution scale map of the element image; performing consistency correction on the convolution scale map corresponding to each element image.

2. The method of claim 1, wherein, The method comprises: performing image preprocessing on the remote sensing data of the target region; determining an element image corresponding to a preset plurality of element dimensions according to the remote sensing data after image preprocessing.

3. The method of claim 1, wherein, The method comprises: for each element image, selecting a candidate scale of each pixel position in the convolution scale set corresponding to the pixel position; determining a convolution scale in the convolution scale set according to the candidate scale corresponding to each pixel position according to a scale selection rule corresponding to the element image, and obtaining a convolution scale map of the element image.

4. The method of claim 3, wherein, The element image comprises a greenness image, a dryness image, a wetness image and a heat image, and the method comprises: for the greenness image, the dryness image and the wetness image, selecting a preset reference scale as the candidate scale of each pixel position in the convolution scale set corresponding to the pixel position; for the heat image, selecting the candidate scale of each pixel position in the convolution scale set corresponding to the pixel position according to a preset spatial support condition.

5. The method of claim 1, wherein, The method comprises: performing principal component analysis based on each element convolution result to obtain a comprehensive ecological quality score; generating an evaluation report of the target region according to each element image, each element convolution result and each comprehensive ecological quality score.

6. A device for assessing the quality of the ecological environment of a region, characterized in that it comprises: The device comprises: an image acquisition module configured to determine a plurality of element images according to remote sensing data of a target region; a scale determination module, configured to determine a convolution scale corresponding to each pixel position in each of the element images according to a category image of the target region, to obtain a convolution scale map of each of the element images, the category image being used to represent a geomorphologic category of each pixel position in the target region; a convolution processing module, configured to perform convolution processing on each of the element images according to the corresponding convolution scale map, to obtain an element convolution result; a quality evaluation module, configured to generate an evaluation report of the target region according to each of the element convolution results; the scale determination module is further configured to: determine a convolution scale set corresponding to the geomorphologic categories included in the category image; determine the geomorphologic category of each pixel position in each of the element images according to the category image; for each of the element images, determine the convolution scale corresponding to each pixel position according to a corresponding scale selection rule and the geomorphologic category of the pixel position in the convolution scale set, to obtain the convolution scale map of the element image; perform consistency correction on the convolution scale map corresponding to each of the element images.

7. An electronic device comprising a memory and a processor, the memory storing a computer program executable on the processor, characterized in that, The processor executes the program to implement the steps of the method in any one of claims 1 to 5.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the method in any one of claims 1 to 5.

Images