A method for monitoring vegetation ecological environment in small watershed in arid region

Abstract

The application discloses a kind of arid region small watershed vegetation ecological environment monitoring method, comprising: obtaining arid region small watershed surface image, arid region small watershed surface image is handled, obtains first vegetation distribution image;According to the matching of first vegetation distribution image and spectral feature library, obtain second vegetation distribution image;From second vegetation distribution image, the spatial distribution characteristics of vegetation community are extracted, the dynamic change trend of spatial distribution characteristics is calculated, and community structure change dataset is obtained;Based on community structure change dataset, short-term evolution trend of vegetation community structure is predicted using long short-term memory network, and dynamic change prediction result of vegetation community structure is obtained;According to dynamic change prediction result, combined with community structure change dataset, quantify biodiversity level, and obtain biodiversity evaluation result;From biodiversity evaluation result, key indicators are extracted, and according to key indicators, real-time monitoring vegetation ecological environment state data.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of environmental monitoring, and particularly relates to a method for monitoring the vegetation ecological environment in a small watershed in a dry area. BACKGROUND

[0002] Vegetation ecological environment monitoring is a core field of ecological protection and sustainable development, and is of key significance to maintaining ecological balance and assessing environmental changes. In particular, in small watersheds in dry areas with scarce water resources and fragile ecology, the vegetation state directly reflects the health status of the ecological system. However, existing monitoring methods mostly rely on manual investigation or traditional remote sensing technology. Manual investigation is time-consuming and labor-intensive, and has limited coverage. Traditional remote sensing technology is difficult to accurately identify drought-tolerant vegetation types unique to dry areas due to insufficient resolution or confusion of vegetation spectral characteristics. These limitations result in low monitoring accuracy and low efficiency, which cannot meet the needs of dynamically tracking changes in vegetation community structure. SUMMARY

[0003] To solve the above technical problems, the present application provides a method for monitoring the vegetation ecological environment in a small watershed in a dry area, which can accurately identify, dynamically monitor and predict the vegetation ecological environment in a small watershed in a dry area, and provides a scientific basis for ecological protection and resource management.

[0004] The present application provides a method for monitoring the vegetation ecological environment in a small watershed in a dry area, comprising:

[0005] acquiring a surface image of a small watershed in a dry area, processing the surface image of the small watershed in the dry area, and acquiring a first vegetation distribution image;

[0006] According to the matching of the first vegetation distribution image and the spectral feature library, a second vegetation distribution image is obtained;

[0007] From the second vegetation distribution image, the spatial distribution characteristics of the vegetation community are extracted, the dynamic change trend of the spatial distribution characteristics is calculated, and a community structure change dataset is obtained;

[0008] Based on the community structure change dataset, a long short-term memory network is used to predict the short-term evolution trend of the vegetation community structure, and a dynamic change prediction result of the vegetation community structure is obtained;

[0009] According to the dynamic change prediction result, combined with the community structure change dataset, the biological diversity level is quantified, and a biological diversity evaluation result is obtained;

[0010] From the biological diversity evaluation result, key indicators are extracted, and according to the key indicators, real-time monitoring of vegetation ecological environment state data is performed.

[0011] Optionally, processing the surface image of the small watershed in the dry area to obtain the first vegetation distribution image comprises:

[0012] Preprocessing techniques were used to denoise and enhance surface images of small watersheds in arid areas to obtain the first surface image;

[0013] Based on the first surface image, a deep convolutional neural network is used to extract features from the complex terrain and sparse vegetation, segment the vegetation area, and obtain the first vegetation distribution image.

[0014] Optionally, obtaining a second vegetation distribution image based on the matching of the first vegetation distribution image with the spectral feature library includes:

[0015] The first vegetation image was used to extract features using a spectral feature library to obtain an initial matching feature set;

[0016] A drought-resistant vegetation classification model is constructed based on the initial matching feature set, and the classification model is obtained.

[0017] If the matching rate of the classification model for the first vegetation image is lower than the preset threshold, the gradient descent method is used to optimize the classification model to obtain an optimized classification model.

[0018] The first vegetation image is classified using the optimized classification model to obtain the second vegetation image.

[0019] Optionally, spatial distribution features of vegetation communities are extracted from the second vegetation distribution image, and the dynamic change trend of spatial distribution features is calculated to obtain a community structure change dataset, including:

[0020] Spatial distribution features of vegetation communities are extracted from the second vegetation distribution image using an image segmentation algorithm to generate a spatial feature dataset.

[0021] If the pixel value distribution of the spatial feature dataset meets the preset threshold, then the spatial features are classified by a clustering algorithm to obtain the community structure classification result.

[0022] Based on the community structure classification results, a spatiotemporal analysis algorithm is used to calculate the dynamic changes in the community structure and generate a dynamic change dataset.

[0023] The dynamic trend dataset is obtained by analyzing the time trend of the dynamic change dataset using a linear regression algorithm.

[0024] Extract significantly changing community structure features from the aforementioned trend dataset to generate a structural change feature set;

[0025] Based on the structural change feature set, spatial features and change trends are integrated using data fusion technology to generate the community structure change dataset.

[0026] Optionally, based on the community structure change dataset, a long short-term memory network is used to predict the short-term evolution trend of the vegetation community structure, and a dynamic change prediction result of the vegetation community structure is obtained, comprising:

[0027] A vegetation community structure change dataset is obtained, and data characteristics are determined through time series analysis to obtain a standardized dataset;

[0028] Data processing technology is used to pre-process the standardized dataset to obtain a pre-processed dataset;

[0029] A long short-term memory network is used to train the pre-processed dataset to obtain a prediction model;

[0030] Based on the prediction result, the prediction result is obtained.

[0031] Optionally, according to the dynamic change prediction result, the community structure change dataset is combined to quantify the biodiversity level, and a biodiversity evaluation result is obtained, comprising:

[0032] A species diversity index calculation method is used to combine the community dynamic trend to calculate the species diversity level to obtain a preliminary quantification result;

[0033] If the preliminary quantification result is inconsistent with a preset ecosystem dynamic threshold, a random forest algorithm is used to adjust the species distribution feature weight to obtain an optimized quantification result;

[0034] According to the optimized quantification result, environmental factor influence is analyzed, key environmental driving factors are extracted, and biodiversity change characteristics are determined;

[0035] Through the biodiversity change characteristics, a time series analysis method is used to predict the community dynamic trend to obtain the biodiversity evaluation result.

[0036] Optionally, key indicators are extracted from the biodiversity evaluation result, and according to the key indicators, real-time monitoring of vegetation ecological environment state data is performed, comprising:

[0037] Key indicators in the biodiversity evaluation result are extracted through data cleaning to obtain a standardized indicator set;

[0038] The standardized indicator set is analyzed in real time to obtain a real-time data stream;

[0039] By comparing the real-time data stream with historical data, a change amplitude is calculated to determine whether it exceeds a preset threshold;

[0040] If the change amplitude exceeds the preset threshold, a dynamic update mechanism is triggered to generate updated vegetation ecological environment state data;

[0041] According to the updated vegetation ecological environment state data, the time series analysis is adopted to obtain the environmental state change trend;

[0042] The change trend is classified through the clustering algorithm to obtain the vegetation ecological characteristics of different regions;

[0043] The visual technology is adopted to dynamically present the vegetation ecological characteristics to obtain the real-time monitoring data.

[0044] Compared with the prior art, the present application has the following advantages and technical effects: the present application obtains the vegetation spectral characteristics through high-resolution multispectral remote sensing data, constructs the drought-tolerant vegetation spectral feature library, and establishes a classification model by using support vector machine and random forest algorithm. Combined with the high-resolution image of the unmanned aerial vehicle, the deep convolutional neural network and the transfer learning technology are used to realize the vegetation type recognition and the distribution image generation. Further, the spatio-temporal analysis algorithm and the long short-term memory network are used to predict the change trend of the vegetation community structure and calculate the biodiversity index. Finally, the real-time data stream processing and the visualization technology are used to dynamically update and output the vegetation ecological environment monitoring results. The present application realizes the accurate identification, dynamic monitoring and prediction of the vegetation ecological environment of the small watershed in the arid region, and provides a scientific basis for ecological protection and resource management. BRIEF DESCRIPTION OF DRAWINGS

[0045] The drawings constituting a part of this application are used to provide a further understanding of the application, the illustrative embodiments of the application and the description thereof are used to explain the application, and do not constitute an improper limitation on the application. In the drawings:

[0046] Figure 1 is a flow chart of a vegetation ecological environment monitoring method of a small watershed in an arid region according to an embodiment of the present application. DETAILED DESCRIPTION

[0047] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the drawings and in combination with the embodiments.

[0048] It should be noted that the steps shown in the flow chart of the drawings can be executed in a computer system such as a group of computer executable instructions, and although the logical order is shown in the flow chart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0049] The present embodiment proposes a vegetation ecological environment monitoring method of a small watershed in an arid region, as shown in Figure 1 , specifically comprising the following steps:

[0050] Obtaining a small watershed surface image in an arid region, processing the small watershed surface image in the arid region, and obtaining a first vegetation distribution image;

[0051] According to the matching of the first vegetation distribution image and the spectral feature library, a second vegetation distribution image is obtained;

[0052] Spatial distribution characteristics of the vegetation community are extracted from the second vegetation distribution image, a dynamic change trend of the spatial distribution characteristics is calculated, and a community structure change dataset is obtained;

[0053] Based on the community structure change dataset, a long short-term memory network is used to predict a short-term evolution trend of the vegetation community structure, and a dynamic change prediction result of the vegetation community structure is obtained;

[0054] According to the dynamic change prediction result, in combination with the community structure change dataset, a biological diversity level is quantified, and a biological diversity evaluation result is obtained;

[0055] Key indicators are extracted from the biological diversity evaluation result, and according to the key indicators, vegetation ecological environment state data is monitored in real time.

[0056] Specifically, the present embodiment obtains vegetation spectral features through high-resolution multispectral remote sensing data, constructs a drought-tolerant vegetation spectral feature library, and establishes a classification model using support vector machine and random forest algorithm. In combination with high-resolution images of unmanned aerial vehicles, deep convolutional neural network and transfer learning technology are used to realize vegetation type recognition and distribution image generation. Further, spatio-temporal analysis algorithm and long short-term memory network are used to predict the change trend of the vegetation community structure, and the biological diversity index is calculated. Finally, through real-time data stream processing and visualization technology, the vegetation ecological environment monitoring result is dynamically updated and output. The present application realizes accurate identification, dynamic monitoring and prediction of the vegetation ecological environment of the small watershed in the arid region, and provides a scientific basis for ecological protection and resource management.

[0057] Further, the first vegetation distribution image is obtained by processing the surface image of the small watershed in the arid region, including:

[0058] The surface image of the small watershed in the arid region is denoised and enhanced by using a preprocessing technique, and a first surface image is obtained;

[0059] According to the first surface image, a deep convolutional neural network is used to extract features of complex terrain and sparse vegetation, and the vegetation area is segmented to obtain the first vegetation distribution image.

[0060] Specifically, when acquiring the image of the surface of the small watershed in the arid region captured by the high-resolution camera carried by the unmanned aerial vehicle, preferably, a multispectral camera is used to capture the reflection characteristics of the surface in different wavebands. For example, the visible light and near-infrared waveband can be selected, with a resolution of 0.1 meters / pixel, to ensure the capture of details of sparse vegetation and complex terrain. The sun elevation angle needs to be considered during acquisition to avoid shadow interference, and preferably the shooting is performed from 10 am to 2 pm. For the preprocessing technology of image denoising and enhancement, specifically, median filtering can be used to remove salt and pepper noise and retain vegetation edge information; then histogram equalization is used to enhance the contrast of the image. For example, for an image with uneven gray value distribution, the difference in pixel values between vegetation and bare soil is made more obvious after adjustment, facilitating subsequent segmentation. This method can improve the image quality and provide reliable data for feature extraction. When using a deep convolutional neural network for feature extraction and vegetation region segmentation, it can be understood that the U-Net model is often chosen due to its adaptability to small sample data. For example, during training, a data set containing 1000 labeled images is input, and the network extracts texture and spectral features of vegetation through multiple layers of convolution, and outputs the first vegetation distribution image. During segmentation, the model can distinguish between sparse shrubs and bare soil to generate a pixel-level vegetation distribution map. For pixels that meet the preset vegetation coverage threshold, for example, the area with normalized difference vegetation index (NDVI) greater than 0.3 can be marked as a vegetation area. For example, if the NDVI value of a certain area is 0.4, it is marked as vegetation; if it is less than 0.3, it is marked as non-vegetation.

[0061] Further, according to the matching between the first vegetation distribution image and the spectral feature library, the second vegetation distribution image is obtained, comprising:

[0062] extracting features from the first vegetation image using the spectral feature library to obtain an initial matching feature set;

[0063] constructing a drought-tolerant vegetation classification model based on the initial matching feature set to obtain a classification model;

[0064] if the matching rate of the classification model to the first vegetation image is lower than a preset threshold, optimizing the classification model using the gradient descent method to obtain an optimized classification model;

[0065] classifying the first vegetation image through the optimized classification model to obtain the second vegetation image.

[0066] Specifically, the spectral feature library is used to extract features of the first vegetation image, aiming to match with known vegetation spectral feature templates. For example, in a small watershed in the arid region, the spectral feature library can contain spectral data of various drought-tolerant vegetation such as Cistanche salsa and Hippophae rhamnoides, covering the visible light to near-infrared range. In one embodiment, the system extracts features with a matching degree higher than 80% by comparing the spectral curve of the image pixels with the templates in the library, forming an initial matching feature set. This ensures that the subsequent classification model can focus on the spectral characteristics of the typical vegetation in the region. The transfer learning technique is used to build a drought-tolerant vegetation classification model, which is trained based on the initial matching feature set. It can be understood that transfer learning reduces the demand for a large amount of labeled data by reusing the deep feature extraction capability of a pre-trained model such as ResNet. In one possible implementation, assuming that the initial matching feature set contains 1000 samples, transfer learning can migrate the convolutional layer parameters of the pre-trained model to the new model, only fine-tune the fully connected layer, and generate a classification model suitable for the specific classification task of drought-tolerant vegetation. If the matching rate of the classification model is lower than a preset threshold, such as 85%, the gradient descent method is used for fine-tuning to obtain an optimized classification model. Illustratively, the fine-tuning process can adjust the learning rate of the model to 0.001, and through multiple iterations, the loss function is optimized to improve the classification accuracy of the model for sparse vegetation. Specifically, fine-tuning can adjust the weight of the near-infrared band in the spectral features to better distinguish between healthy vegetation and non-vegetation areas. By optimizing the classification model, the first vegetation image is classified to obtain a second vegetation image. For example, the classification model can classify the image pixels into three categories: drought-tolerant vegetation, non-vegetation, and transition area, generating a second vegetation image containing class labels.

[0067] Further, the spatial distribution features of the vegetation community are extracted from the second vegetation distribution image, the dynamic change trend of the spatial distribution features is calculated, and a community structure change dataset is obtained, including:

[0068] The spatial distribution features of the vegetation community are extracted from the second vegetation distribution image using an image segmentation algorithm to generate a spatial feature dataset;

[0069] If the pixel value distribution of the spatial feature dataset meets a preset threshold, the spatial features are classified by a clustering algorithm to obtain a community structure classification result;

[0070] According to the community structure classification result, a spatiotemporal analysis algorithm is used to calculate the dynamic change of the community structure to generate a dynamic change dataset;

[0071] The time trend of the dynamic change of the dynamic change dataset is analyzed by a linear regression algorithm to obtain a change trend dataset;

[0072] Significant change features of the community structure are extracted from the change trend dataset to generate a structure change feature set;

[0073] According to the structural change feature set, the spatial features and change trends are integrated by using a data fusion technology to generate a community structure change data set.

[0074] Specifically, the spatial distribution features of the vegetation community are extracted from the second vegetation distribution image by using an image segmentation algorithm. For example, a region growing-based segmentation algorithm can be used. In principle, this algorithm starts from a seed point and expands by merging similar pixels to form a community region based on the gray value or color features of the pixels. For example, in a vegetation image of a dry area, set the seed point as a green pixel and the threshold as a gray difference of 10. The community region dominated by drought-resistant shrubs is segmented, and a spatial feature data set containing the community boundary and area is generated, with the area unit being square kilometers. In a possible implementation manner, if the pixel value distribution of the spatial feature data set meets a preset threshold, for example, the pixel value variance is less than 50, the spatial features can be classified by using a K-means clustering algorithm. Based on the spatial coordinates and color features of the pixels, the community is divided into a high-density shrub area and a sparse grassland area. For example, set the K value to 3, and after clustering, a classification result that the shrub area accounts for 60%, the grassland area accounts for 30%, and the bare land accounts for 10% is obtained, clearly reflecting the community structure. Specifically, based on the community structure classification result, a spatio-temporal analysis algorithm is used to calculate the dynamic change of the community structure. Spatio-temporal analysis quantifies the area change by comparing the community boundaries of multiple time-phase images. For example, by analyzing the images of 2023 and 2024, it is found that the area of the shrub area increases from 100 square kilometers to 120 square kilometers, and a dynamic change data set containing the change amplitude and time stamp is generated, revealing the expansion trend of the community. Preferably, if the time series of the dynamic change data set meets the continuity condition, for example, 6 months of data are continuously collected every month, the time trend can be analyzed by using a linear regression algorithm. The linear regression fitting area change curve obtains a change trend data set with a positive slope. For example, the area of the shrub area increases by 2 square kilometers per month, indicating that the community is stably expanding, which provides a basis for ecological monitoring. In an embodiment, the significant change features of the community structure are extracted from the change trend data set, and a change rate threshold of 5% can be set to extract the expansion features of the shrub area, to generate a structural change feature set. For example, the shrub area boundary expands northward by 10 kilometers, reflecting the influence of improved water sources or increased rainfall, and providing data support for vegetation restoration research. For example, the spatial features and change trends are integrated by using a data fusion technology to generate a community structure change data set.

[0075] Further, based on the community structure change data set, a long short-term memory network is used to predict the short-term evolution trend of the vegetation community structure, to obtain a dynamic change prediction result of the vegetation community structure, including:

[0076] A vegetation community structure change data set is obtained, time series analysis is performed to determine the data features, and a standardized data set is obtained;

[0077] The standardized data set is preprocessed by using a data processing technology to obtain a preprocessed data set;

[0078] A prediction model is obtained by training a long short-term memory network on the preprocessed data set;

[0079] Based on the prediction result, the prediction result is obtained.

[0080] Specifically, after obtaining the vegetation community structure change dataset, time series analysis is a key step to extract data features. For example, time series analysis can extract trends, seasonality, and random fluctuations through decomposition methods. For example, for vegetation coverage data in a certain area, the monthly coverage change in the past 10 years is analyzed, and it is found that the coverage changes periodically with the seasons, such as an increase of about 15% in spring and a decrease of about 10% in autumn each year. This analysis helps to understand the dynamic rules of the vegetation community and provides a basis for subsequent modeling. In one possible implementation, the generation of the standardized dataset requires converting the original data to a format with a mean of 0 and a standard deviation of 1. Specifically, assuming that the mean of the vegetation community coverage data is 50% and the standard deviation is 5%, the data is converted to a standard normal distribution through standardization, which facilitates model training. Standardization ensures that data of different dimensions is comparable, improving the convergence speed of the model. It should be noted that data cleaning is an important step to remove outliers. For example, the coverage recorded at a certain time point is abnormally high at 90%, far exceeding the normal range of 40%-60%. Through box plot analysis, it is determined that it is an outlier and is removed. The preprocessed dataset after cleaning can better reflect the true trend of the vegetation community, avoiding model bias due to noisy data. Preferably, a long short-term memory network is used to train the preprocessed dataset, as it is good at capturing long and short-term dependencies in time series. In one embodiment, for data of a certain forest community, the network is set to a 3-layer structure, and the last 5 years of coverage data are input to train the model to predict the trend for the next 12 months. The initial prediction model may predict a 2% decrease in coverage, but the actual data only decreases by 0.5%, with an error exceeding the threshold. Adjusting the network learning rate to 0.001 and increasing the hidden layer nodes, the optimized model prediction error is reduced to within 0.3%. For example, the dynamic change prediction result needs to be verified for consistency through time series analysis. Specifically, the prediction result shows that the coverage of a certain grassland community will decrease by 5% in the next 3 years. Through autoregressive model analysis of historical data, it is confirmed that the downward trend is consistent with the fluctuation pattern of the past 10 years, verifying the credibility of the prediction. This verification ensures that the prediction result is consistent with the actual ecological process. It can be understood that optimizing the prediction model for short-term evolution trend prediction is of great significance to ecological management. For example, predicting a 3% decrease in coverage of a certain wetland community in the short term can allow for the early adoption of protective measures, such as limiting development or increasing irrigation. This prediction capability supports precise decision-making and reduces ecological risks. In one embodiment, when adjusting the long short-term memory network parameters, the optimal parameter combination can be determined through grid search. For example, try learning rate 0.001-0.01, hidden layer nodes 50-200, and find the configuration with the smallest error. This method ensures that the model adapts to the data characteristics of different vegetation communities, improving prediction accuracy.

[0081] Further, according to the dynamic change prediction result, combined with the community structure change dataset, the biodiversity level is quantified, and the biodiversity evaluation result is obtained.

[0082] The species diversity index calculation method is combined with the community dynamic trend to calculate the species diversity level, and a preliminary quantitative result is obtained.

[0083] If the preliminary quantitative result is inconsistent with the preset ecosystem dynamic threshold, the species distribution characteristic weight is adjusted through the random forest algorithm to obtain an optimized quantitative result.

[0084] According to the optimized quantitative result, the key environmental driving factors are extracted, and the biodiversity change characteristics are determined.

[0085] Through the biodiversity change characteristics, a time series analysis method is used to predict the community dynamic trend, and a biodiversity evaluation result is obtained.

[0086] Specifically, in the field of vegetation community structure change analysis, data cleaning technology is a key step to ensure data quality. For example, for a community structure change dataset, abnormal values can be detected by statistical methods. For example, suppose a dataset records the species density of 100 sample plots in a certain area, and it is found that the species density of a certain sample plot is more than 3 times the standard deviation of the mean. It can be considered as noise data and removed. After cleaning, the data can better reflect the true community characteristics, which helps to improve the accuracy of subsequent analysis. Specifically, the standardized dataset needs to be obtained by unifying the data of different dimensions. For example, the species density, coverage and other indicators are standardized by Z-score, which is converted into a dataset with a mean of 0 and a standard deviation of 1. This can eliminate the difference in dimensions and facilitate time series analysis. In one embodiment, a forest plot data contains species density and soil moisture, which can be directly compared after standardization to reveal the community dynamic characteristics. Preferably, time series analysis is used to extract species distribution characteristics. For example, by analyzing the species abundance data of a grassland community for 5 years through an autoregressive model, it is found that the abundance of a dominant species shows periodic fluctuations, suggesting that the community dynamic trend is related to seasonal rainfall. In one possible implementation, the calculation of the species diversity index can use the Shannon index. Assuming that a wetland community contains 10 species of plants, after calculating the relative abundance of each species, the diversity index is 2.3, indicating that the community diversity is high. If this value is lower than the preset threshold of 2.5, further analysis of community stability is required. This method quantitatively reflects the community structure and is convenient for ecological assessment.

[0087] Further, key indicators are extracted from the biodiversity evaluation result, and the key indicators are used to monitor the vegetation ecological environment state data in real time, including:

[0088] The key indicators in the biodiversity evaluation result are extracted through data cleaning to obtain a standardized index set.

[0089] The real-time data stream is obtained by real-time analysis of the standardized index set.

[0090] By comparing real-time data stream with historical data, the change amplitude is calculated to determine whether it exceeds the preset threshold;

[0091] If the change amplitude exceeds the preset threshold, a dynamic updating mechanism is triggered to generate updated vegetation ecological environment state data;

[0092] According to the updated vegetation ecological environment state data, time series analysis is used to obtain the environmental state change trend;

[0093] Through clustering algorithm, the change trend is classified to obtain the vegetation ecological characteristics of different regions;

[0094] The visualization technology is used to dynamically present the vegetation ecological characteristics to obtain real-time monitoring data.

[0095] Specifically, missing data is filled by the average value of adjacent sampling points, and abnormal high-density records caused by human interference are removed. The cleaned data forms a standardized index set, such as species richness, evenness, and community density, providing a reliable basis for subsequent analysis. In one possible implementation, stream processing technology is used to analyze the standardized index set in real time. Stream processing can handle continuous input data streams, making it suitable for dynamic monitoring of ecosystem changes. Compared with traditional batch processing, stream processing can respond more quickly to ecological abnormalities and improve monitoring efficiency. Specifically, the change amplitude is calculated by comparing real-time data stream with historical data. Historical data can come from monthly records over the past five years, while real-time data stream is the current hourly monitoring value. For example, if the species richness of a certain forest area decreases from the historical average of 50 species to the real-time value of 40 species, the change amplitude is 20%. If the preset threshold is 15%, the dynamic updating mechanism is triggered. This comparison can quickly detect ecosystem abnormalities and provide a basis for protection measures. Preferably, the dynamic updating mechanism generates updated vegetation ecological environment state data. For example, when the vegetation coverage of a certain grassland area decreases due to drought, the system will update the state based on real-time data and generate comprehensive state data including coverage, species diversity, and soil moisture. These data provide accurate guidance for ecological restoration and avoid resource waste. In one embodiment, time series analysis is used to predict the environmental state change trend. For example, based on vegetation coverage data over the past three years, seasonal fluctuations and long-term downward trends are analyzed to predict that the coverage may decrease by 5% in the next year. This prediction helps to develop protection plans in advance and optimize resource allocation. For example, clustering algorithm is used to classify the change trend to identify regional characteristics.

[0096] The above merely provides the preferred embodiments of the present application, and the protection scope of the present application is not limited thereto, and any changes or substitutions within the technical scope disclosed by the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for monitoring the vegetation ecological environment in a small watershed in arid areas, characterized in that, include: Obtain surface images of a small watershed in an arid region, process the surface images of the small watershed in the arid region, and obtain a first vegetation distribution image; Based on the matching of the first vegetation distribution image with the spectral feature library, the second vegetation distribution image is obtained; The spatial distribution features of vegetation communities are extracted from the second vegetation distribution image, the dynamic change trend of spatial distribution features is calculated, and a dataset of community structure change is obtained. Spatial distribution features of vegetation communities are extracted from the second vegetation distribution image, and the dynamic change trend of spatial distribution features is calculated to obtain a community structure change dataset, including: Spatial distribution features of vegetation communities are extracted from the second vegetation distribution image using an image segmentation algorithm to generate a spatial feature dataset. If the pixel value distribution of the spatial feature dataset meets the preset threshold, then the spatial features are classified by a clustering algorithm to obtain the community structure classification result. Based on the community structure classification results, a spatiotemporal analysis algorithm is used to calculate the dynamic changes in the community structure and generate a dynamic change dataset. The dynamic trend dataset is obtained by analyzing the time trend of the dynamic change dataset using a linear regression algorithm. Extract significantly changing community structure features from the aforementioned trend dataset to generate a structural change feature set; Based on the structural change feature set, spatial features and change trends are integrated using data fusion technology to generate the community structure change dataset. Based on the community structure change dataset, a long short-term memory network is used to predict the short-term evolution trend of vegetation community structure and obtain the dynamic change prediction results of vegetation community structure. Based on the predicted dynamic changes, combined with the community structure change dataset, the level of biodiversity is quantified to obtain the biodiversity assessment results. Key indicators are extracted from the biodiversity assessment results, and vegetation ecological environment status data are monitored in real time based on these key indicators.

2. The method for monitoring vegetation ecological environment in arid small watersheds according to claim 1, characterized in that, Processing the surface image of the arid small watershed to obtain the first vegetation distribution image includes: Preprocessing techniques were used to denoise and enhance surface images of small watersheds in arid areas to obtain the first surface image; Based on the first surface image, a deep convolutional neural network is used to extract features from the complex terrain and sparse vegetation, segment the vegetation area, and obtain the first vegetation distribution image.

3. The method for monitoring vegetation ecological environment in arid small watersheds according to claim 1, characterized in that, Based on the matching of the first vegetation distribution image with the spectral feature library, the second vegetation distribution image is obtained, including: The first vegetation distribution image was used to extract features using a spectral feature library to obtain an initial matching feature set; A drought-resistant vegetation classification model is constructed based on the initial matching feature set, and the classification model is obtained. If the matching rate of the classification model for the first vegetation distribution image is lower than the preset threshold, the gradient descent method is used to optimize the classification model to obtain an optimized classification model. The optimized classification model is used to classify the first vegetation distribution image to obtain the second vegetation distribution image.

4. The method for monitoring vegetation ecological environment in arid small watersheds according to claim 1, characterized in that, Based on the aforementioned community structure change dataset, a long short-term memory network is used to predict the short-term evolution trend of vegetation community structure, and the dynamic change prediction results of vegetation community structure are obtained, including: Obtain a dataset of vegetation community structure changes, determine data characteristics through time series analysis, and obtain a standardized dataset; Data processing techniques are used to preprocess the standardized dataset to obtain a preprocessed dataset; The preprocessed dataset is trained using a Long Short-Term Memory network to obtain a predictive model; Based on the prediction model, the prediction results are obtained.

5. The method for monitoring vegetation ecological environment in arid small watersheds according to claim 1, characterized in that, Based on the dynamic change prediction results and combined with the community structure change dataset, the biodiversity level is quantified, and the biodiversity assessment results include: The species diversity index method was used, combined with community dynamics, to calculate the level of species diversity and obtain preliminary quantitative results. If the preliminary quantification results are inconsistent with the preset ecosystem dynamic threshold, the species distribution feature weights are adjusted using the random forest algorithm to obtain optimized quantification results. Based on the optimized quantitative results, the impact of environmental factors is analyzed, key environmental driving factors are extracted, and the characteristics of biodiversity change are determined. By analyzing the characteristics of biodiversity changes and employing time series analysis methods, we can predict the dynamic trends of communities and obtain the biodiversity assessment results.

6. The method for monitoring vegetation ecological environment in arid small watersheds according to claim 1, characterized in that, Key indicators are extracted from the biodiversity assessment results, and real-time monitoring of vegetation ecological environment status data based on these key indicators includes: Key indicators were extracted from the biodiversity assessment results through data cleaning to obtain a standardized indicator set; Perform real-time analysis on the standardized indicator set to obtain real-time data streams; By comparing the real-time data stream with historical data, the magnitude of change is calculated to determine whether it exceeds a preset threshold. If the change exceeds the preset threshold, a dynamic update mechanism is triggered to generate updated vegetation ecological environment status data. Based on the updated vegetation ecological environment status data, time series analysis was used to obtain the trend of environmental status change; Clustering algorithms are used to classify the changing trends and obtain the vegetation ecological characteristics of different regions; Visualization technology is used to dynamically present the ecological characteristics of vegetation, thereby obtaining real-time monitoring data.

Images