Natural reserve intelligent ecological protection and management method, equipment and medium

Abstract

The invention discloses an intelligent ecological protection and management method for a natural reserve, and the method comprises a satellite remote sensing ecological environment monitoring method which achieves the remote sensing image segmentation based on a satellite remote sensing image and a neural network model, is used for accurately distinguishing ecological systems, and accurately calculates the area of each ecological system through two-dimensional integral operation. Visual display is realized in a thermodynamic diagram form, and real-time monitoring of ecological system changes is realized; according to the species quantity change prediction and decision-making method, a neural network model is used for extracting a nonlinear feature relationship between species quantity change and each environmental factor in a species quantity change prediction data set, and accurate prediction of the species quantity change trend on the premise of known environmental factor change trends is realized; according to the species identification method, accurate identification and distinguishing of similar species are realized through big data and a neural network model. The method has the advantages of being large in monitoring range, automatic in monitoring, high in species identification and quantity statistics accuracy and high in population quantity change trend prediction accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of ecological monitoring of nature reserves, and specifically to a method, device and storage medium for intelligent ecological protection and management of nature reserves based on deep learning. Background Art

[0002] A nature reserve refers to a specific area designated for the purpose of protecting biodiversity, maintaining ecosystem functions and providing ecological services, including nature reserves, scenic spots, forest parks, geological parks and other types of protected areas; ecological protection refers to a series of measures to maintain the integrity, stability and sustainability of natural ecosystems and prevent damage to the ecological environment caused by human activities; ecological management refers to the planning, monitoring, evaluation, law enforcement and other activities of nature reserves to ensure the achievement of protection goals.

[0003] The ecological protection and management of nature reserves emphasize the scientific protection of natural resources and ecosystems, and pay attention to the coordination of ecological balance and sustainable development. Its goals include protecting rare and endangered species, maintaining biodiversity, improving the quality of the ecological environment, and promoting harmonious coexistence between man and nature. To achieve these goals, it is necessary to adopt a variety of measures, such as formulating relevant laws and regulations, implementing ecological protection projects, strengthening scientific research monitoring and public education, etc.

[0004] The population size and richness, as well as the continuous changing trend of the population size over a period of time can effectively reflect the ecological environmental status of nature reserves, and assist the ecological protection and management personnel of nature reserves to take effective measures to maintain the stability of the ecological environment of nature reserves; at the same time, based on the specific parameters of the ecological environment of nature reserves, such as temperature, humidity, drought, water quality, soil pH, etc., the future changes in the population size of specific species can be effectively predicted, and timely early warnings can be issued for negative impact factors such as endangered species, invasion of alien species, ecological imbalance, and increased environmental pollution.

[0005] The existing ecological protection and management methods for nature reserves have problems such as small monitoring range, high monitoring manpower costs, low accuracy of species identification and quantity statistics, and large deviations in the prediction of population change trends.

[0006] To this end, this application specifically proposes a method for intelligent ecological protection and management of nature reserves to solve the technical problems raised above. Summary of the invention

[0007] The main purpose of the present invention is to provide an intelligent ecological protection and management method for nature reserves to solve the technical problems existing in the prior art of background technology, such as small monitoring range, high monitoring manpower cost, low accuracy of species identification and quantity statistics, and large deviation in prediction of population quantity change trend.

[0008] The present invention adopts the following technical solutions to solve the above technical problems:

[0009] A method for intelligent ecological protection and management of nature reserves, applied to the ecological protection and management system of nature reserves, the system includes an intelligent perception subsystem, an intelligent prediction subsystem and an intelligent species identification subsystem, including a satellite remote sensing ecological environment monitoring method, a species quantity change prediction and decision-making method and a species identification method, wherein:

[0010] The satellite remote sensing ecological environment monitoring method is applied to the intelligent perception subsystem. The remote sensing image segmentation is realized based on the satellite remote sensing image and the neural network model. It is used to accurately distinguish the mountain, lake, wetland, grassland, and forest ecosystems, and accurately calculate the area of ​​each ecosystem through two-dimensional integral operation. It is intuitively displayed in the form of a heat map to realize real-time monitoring of ecosystem changes.

[0011] The species population change prediction and decision-making method is applied to the intelligent prediction subsystem, which is used to extract the nonlinear characteristic relationship between species population change and various environmental factors in the species population change prediction data set using a neural network model, and to achieve accurate prediction of species population change trends under the premise of known environmental factor change trends;

[0012] The species identification method is applied to the intelligent species identification subsystem, which is used to accurately identify and distinguish similar species through big data and neural network models, and to assist in the accurate statistics of species populations in natural ecological reserves.

[0013] Preferably, the specific operation steps of the method include:

[0014] S1. Obtain real-time satellite remote sensing images of ecological protection areas through geosynchronous orbit satellites;

[0015] S2. Send the remote sensing image to the neural network model to distinguish the lake, grassland, mountain, jungle, and wetland ecological environment contained in the ecological protection area in the satellite remote sensing image through the remote sensing image segmentation technology based on the neural network model, circle the range in the remote sensing image with different colored curve frames, and mark the corresponding names;

[0016] S3. Use satellite remote sensing scale to calculate the real length and width of each ecological environment at this moment, obtain the real area through irregular two-dimensional integral operation, and display it in real time using a heat map;

[0017] S4. Use the heat map to determine whether the lake area continues to decrease and whether the forest and grassland greening areas continue to decrease. If so, it means that the drought in the nature reserve is gradually getting worse and the ecological environment is continuously deteriorating, requiring human intervention and protection;

[0018] S5. Managers use sensors to obtain relevant environmental parameters of nature reserves and input them into the species population change prediction and decision-making neural network to obtain the changes in the number of species in nature reserves over a period of time in the future and determine whether there is a risk of reduced species diversity;

[0019] S6. If there is a small decline in the number of species in a nature reserve, the cause of the deterioration of the ecological environment in the nature reserve is inferred based on the neural network model and corresponding operations are taken to curb the continued deterioration of the ecological environment;

[0020] S7. At the same time, an ecological archive of species in nature reserves should be established, and a species identification method based on a neural network should be built into portable devices. When the portable devices are aimed at the species to be identified in the wild, they can accurately identify birds of similar appearance but different types through neural network calculations, so that managers of ecological reserves can understand and grasp the species situation in nature reserves, classify the rarity of species, and take necessary protection measures for rare and endangered species in a timely manner.

[0021] Preferably, the specific operation steps of the satellite remote sensing ecological environment monitoring method include:

[0022] a1. Collect satellite remote sensing images of ecological protection areas through geosynchronous orbit satellites, and form remote sensing data sets through preprocessing and data annotation;

[0023] a2. Divide the prepared remote sensing image dataset into training set and test set in a ratio of 8:2;

[0024] a3. Construct a gated residual neural network model to complete model training on the training set, and verify that the performance reaches the preset standard on the test set, and then use the trained model to perform real-time remote sensing image segmentation tasks;

[0025] a4. After the neural network completes the segmentation of satellite remote sensing images, the area of ​​each ecosystem is calculated through irregular two-dimensional integration and displayed in the form of a heat map. The irregular two-dimensional integration formula is: ∫∫Df(x, y)dxdy, where D is the two-dimensional integration area, f(x, y) is the integrand, and dxdy represents the integration of x and y.

[0026] Preferably, the step of forming the remote sensing data set in step a1 includes:

[0027] a11. Use high-precision remote sensing cameras to continuously collect images of specific areas at a certain period;

[0028] a12. Perform resizing, cropping, normalization, data enhancement and filtering operations on the collected images;

[0029] a13. Perform image preprocessing by denoising, radiation correction, orthorectification, atmospheric correction, image enhancement, impact superposition, and coordinate transformation;

[0030] a14. Data annotation involves interpreting remote sensing images, identifying various types of land features, and labeling them accordingly.

[0031] Preferably, the main structure of the gated residual neural network model in step a3 is a convolution layer, and the convolution layer of the gated residual neural network model adopts a 3×3 convolution kernel, and the gated residual neural network model is connected to a batch normalization operation and a nonlinear activation function after the convolution operation;

[0032] The output layer of the gated residual neural network model is designed as an activation gate structure, which is composed of a ReLU nonlinear activation function and a Sigmoid nonlinear activation function, and is used to take into account the context feature relationship between images captured in consecutive time periods and improve the continuous processing capability of the neural network for continuous data, wherein:

[0033] The calculation formula of the ReLU nonlinear activation function is:

[0034] f(x) = max(0, x)

[0035] The calculation formula of Sigmoid nonlinear activation function is:

[0036]

[0037] The model training uses the cross entropy loss function, and the calculation formula is:

[0038]

[0039] Where N is the number of pixels in the image, c is the number of categories, and y ic is an indicator variable. If the true label of sample i is category c, then y ic =1, otherwise y ic =0, p ic is the probability that the model predicts sample i as category c, calculated by the softmax function.

[0040] Preferably, the species population change prediction and decision-making method includes:

[0041] b1. Collect data on lake area, grassland area, forest area, temperature, humidity, soil pH, rainfall, snowfall, average daily light intensity, light duration and corresponding species quantity within a specified time period through sensors;

[0042] b2. Set the "corresponding number of species" as the predicted label data, and the remaining data as the factors that affect the change in the number of species, which are also the input data of the neural network model, using α1~α 13 The influence weight of each influencing factor is represented by β1~β 13 Indicates that the data set structure can be expressed as [α1β1…α 13 β 13 ], the constructed data set is divided into training set and test set in a ratio of 8:2;

[0043] b3. As the ecological environment of natural ecological protection areas is relatively closed, the number of species is relatively stable without special geological disasters and human destruction, so the natural birth and death rates of species are relatively balanced. Under the premise of known influencing factor values, the corresponding species number prediction value is:

[0044]

[0045] Among them, number t+1 is the predicted value of population size at the next moment, α it is the influencing factor value at the current moment, α i(t+1) is the influencing factor value at the next moment, γ t is the current population size. At this time, the species population prediction problem is transformed into the reasonable weight β of each influencing factor;

[0046] b4. Construct a bidirectional long short-term memory neural network model, and train the constructed neural network model on the training set. The MSE loss function is used in the model training, and the expression is as follows:

[0047]

[0048] Among them, y i For actual results, Predict results for the neural network;

[0049] b5. After the model training is completed, save the corresponding model weight file, load the weight file and test set data, and test the model performance;

[0050] b6. Send the constructed data set into the bidirectional long short-term memory neural network model and give a reasonable weight configuration to perform population prediction in continuous time. The neural network model gives human intervention decision recommendations based on the prediction results and whether the environmental factors reach the ideal threshold.

[0051] Preferably, the specific operation steps of the species identification method include:

[0052] c1. Collect a large amount of image data of species in nature reserves through manual photography and hidden cameras in the wild, and build a species image reference database;

[0053] c2. Grayscale each image to increase the model training speed;

[0054] c3. Convert the grayscale image to a different size, remove the background or blank area that does not contain any information, and prevent the image from being too large or having too many pixels;

[0055] c4. Mark the name and location of the species in each image, build a species identification dataset, and divide the species identification dataset into a training dataset and a test dataset in a ratio of 8:2;

[0056] c5. Build the SE-Inceptionv4 neural network model, complete the model training on the training data set, and use it to perform species recognition tasks after verifying its good performance on the test data set.

[0057] Preferably, the SE-Inceptionv4 neural network model in step c5 is mainly composed of a convolution layer and a maximum pooling layer. The convolution layer of the SE-Inceptionv4 neural network model uses four convolution kernels of different sizes: 1*1, 3*3, 1*7 and 7*1, and uses a large number of feature splicing operations to fully integrate the multi-scale features extracted by convolution kernels of different sizes.

[0058] The SE attention mechanism module is introduced after the first feature concatenation operation to improve the model's feature extraction capability by scaling the features at the channel scale.

[0059] The SE-Inceptionv4 neural network model adopts a maximum pooling layer operation in the model architecture, so that the model can obtain and retain the most prominent image features as much as possible under the constraint of a limited number of layers.

[0060] On the other hand, the present invention further discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the above method.

[0061] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0062] It can be seen from the above technical solutions that the present invention provides a method for intelligent ecological protection and management of nature reserves. Compared with the prior art, the present invention has the following advantages:

[0063] 1. The method of the present invention is not limited by ecological protection areas and species types. By adopting a satellite remote sensing image segmentation method based on a neural network, it can distinguish different ecological environments and landforms in nature reserves in real time, quickly and accurately, and accurately calculate the area parameters of each ecological environment and landform. Compared with the traditional method that relies on manual statistical measurement, it greatly reduces the statistical cost, and has higher statistical accuracy and faster speed, and can achieve real-time data update.

[0064] 2. The method of the present invention can accurately and effectively explore the mapping relationship between environmental influencing factors and changes in landforms and species numbers, accurately predict the changing trends of the ecological environment, landforms and species numbers of nature reserves in combination with current environmental parameters, and provide guidance and strategies for artificial intervention, providing a reference for artificial intervention protection.

[0065] 3. The satellite remote sensing ecological environment detection method based on neural network proposed in the present invention can distinguish different ecological environments and landforms in nature reserves in real time, quickly and accurately by adopting a satellite remote sensing image segmentation method based on neural network, and accurately calculate the area parameters of each ecological environment and landform, and intuitively display them in the form of a heat map. Compared with the traditional method that relies on manual statistical measurement, it greatly reduces the statistical cost, and has higher statistical accuracy and faster speed, and can achieve real-time data update.

[0066] 4. The neural network-based species population change prediction and decision-making method proposed in the present invention can accurately and effectively explore the mapping relationship between environmental influencing factors and landforms and species population changes, accurately predict the changing trends of the ecological environment, landforms and species populations of nature reserves in combination with current environmental parameters, and provide guidance and strategies for artificial intervention, providing a reference for artificial intervention protection.

[0067] 5. The species identification method proposed in the present invention, through the support of big data and computing resources, can achieve more accurate species identification, especially for similar species that are difficult to distinguish with the naked eye, the species identification accuracy is higher and the efficiency is higher.

[0068] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become easy to understand through the following description. Of course, it is not necessary to achieve all of the advantages described above simultaneously for any product implementing the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The drawings constituting a part of the present application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0070] Figure 1 It is a schematic diagram of the overall process of applying the method of the present invention to the ecological protection and management system of nature reserves;

[0071] Figure 2 The overall process step diagram of the method of the present invention is

[0072] Figure 3 It is a schematic diagram of the structure of the improved residual network model of the present invention;

[0073] Figure 4 A flow chart of the method for predicting and making decisions on species population changes of the present invention;

[0074] Figure 5 Schematic diagram of the SE-Inceptionv4 model structure of the present invention. DETAILED DESCRIPTION

[0075] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. In the absence of conflict, the embodiments in this application and the features in the embodiments can be combined with each other. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0076] In the embodiment, see Figures 1 to 5 .

[0077] like Figure 1 As shown. The complete intelligent ecological protection and management method for nature reserves proposed in the embodiment of the present invention can be directly applied to the ecological protection and management system of nature reserves, and is not limited by ecological reserves and species types. By adopting a satellite remote sensing image segmentation method based on a neural network, it can distinguish different ecological environments and landforms in nature reserves in real time, quickly and accurately, and accurately calculate the area parameters of each ecological environment and landform. Compared with the traditional method that relies on manual statistical measurement, it greatly reduces the statistical cost, and has higher statistical accuracy and faster speed, and can achieve real-time data update.

[0078] The system includes an intelligent perception subsystem, an intelligent prediction subsystem and an intelligent species identification subsystem, including a satellite remote sensing ecological environment monitoring method, a species population change prediction and decision-making method and a species identification method, among which:

[0079] The satellite remote sensing ecological environment monitoring method is applied to the intelligent perception subsystem. The remote sensing image segmentation is realized based on the satellite remote sensing image and the neural network model. It is used to accurately distinguish the mountain, lake, wetland, grassland, and forest ecosystems, and accurately calculate the area of ​​each ecosystem through two-dimensional integral operation. It is intuitively displayed in the form of a heat map to realize real-time monitoring of ecosystem changes.

[0080] The species population change prediction and decision-making method is applied to the intelligent prediction subsystem, which is used to extract the nonlinear characteristic relationship between species population change and various environmental factors in the species population change prediction data set using a neural network model, to achieve accurate prediction of species population change trends under the premise of known environmental factor change trends, and to give targeted decision-making suggestions to guide managers to take protection and intervention measures on the ecological environment of nature reserves;

[0081] The species identification method is applied to the intelligent species identification subsystem, which is used to accurately identify and distinguish similar species through big data and neural network models, and to assist in the accurate statistics of species populations in natural ecological reserves.

[0082] In a specific embodiment, Figure 2 The specific operation steps of the intelligent ecological protection and management method of the nature reserve shown include:

[0083] S1. Nature reserve management and researchers obtain real-time satellite remote sensing images of ecological reserves through geosynchronous orbit satellites;

[0084] S2. Send the remote sensing image to the neural network model to distinguish the lake, grassland, mountain, jungle, and wetland ecological environment contained in the ecological protection area in the satellite remote sensing image through the remote sensing image segmentation technology based on the neural network model, circle the range in the remote sensing image with different colored curve frames, and mark the corresponding names;

[0085] S3. Use satellite remote sensing scale to calculate the real length and width of each ecological environment at this moment, obtain the real area through irregular two-dimensional integral operation, and display it in real time using a heat map;

[0086] S4. Through the heat map, researchers found that the area of ​​lakes is continuously decreasing, and the green area of ​​forests and grasslands is also continuously decreasing. It is speculated that the drought in nature reserves is gradually getting worse, and the ecological environment may be continuously deteriorating, requiring human intervention and protection;

[0087] S5. Managers use sensors to obtain relevant environmental parameters of nature reserves and input them into the species population change prediction and decision-making neural network to obtain the changes in the number of species in nature reserves over a period of time in the future and determine whether there is a risk of reduced species diversity;

[0088] S6. If there is a small decline in the number of species in a nature reserve, the cause of the deterioration of the ecological environment in the nature reserve is inferred based on the neural network model and corresponding operations are taken to curb the continued deterioration of the ecological environment;

[0089] In the specific implementation, when it was found that the number of species in nature reserves might decline slightly for a period of time in the future and there was a risk of reduced species diversity, the neural network model inferred that it was caused by drought and water shortage, and recommended the use of artificial rainfall to alleviate the degree of drought, and gave a reference value of 3mm for the thickness of artificial rainfall. The management and researchers applied to the relevant departments for artificial rainfall according to the given reference value, successfully curbing the continuous deterioration of the ecological environment and protecting the species diversity of nature reserves.

[0090] S7. At the same time, in order to effectively establish ecological archives of species in nature reserves, a species identification method based on neural networks is built into portable devices. When the portable devices are aimed at the species to be identified in the wild, they can accurately identify birds of similar appearance but different types through neural network calculations. This allows managers of ecological reserves to more clearly understand and grasp the species situation in nature reserves, facilitate the classification of species rarity, and take necessary protection measures for rare and endangered species in a timely manner.

[0091] In summary, this method can accurately and effectively explore the mapping relationship between environmental influencing factors and changes in landforms and species numbers. Combined with current environmental parameters, it can accurately predict the changing trends of the ecological environment, landforms and species numbers in nature reserves, and provide guidance and strategies for artificial intervention, providing a reference for artificial intervention protection.

[0092] In a specific embodiment, the specific operation steps of the satellite remote sensing ecological environment monitoring method include:

[0093] a1. Ecological management and research of nature reserves collect satellite remote sensing images of ecological reserves through geosynchronous orbit satellites, and form remote sensing data sets through preprocessing and data annotation. The steps for forming remote sensing data sets include:

[0094] a11. When collecting images, use a high-precision remote sensing camera to continuously collect images of a specific area at a certain period;

[0095] a12. Performing resizing, cropping, normalization, data enhancement and filtering operations on the collected images during image acquisition;

[0096] a13. Image preprocessing includes image denoising, radiation correction, orthorectification, atmospheric correction, image enhancement, impact superposition, and coordinate conversion operations;

[0097] a14. Data annotation involves interpreting remote sensing images, identifying various types of objects, and labeling them accordingly;

[0098] a2. Divide the prepared remote sensing image dataset into training set and test set in a ratio of 8:2;

[0099] a3. Construct a gated residual neural network model to complete model training on the training set, and verify that the performance reaches the preset standard on the test set, and then use the trained model to perform real-time remote sensing image segmentation tasks;

[0100] The gated residual neural network is improved on the basis of the residual neural network, retaining the skip connection structure of the residual network, alleviating the gradient vanishing and gradient exploding phenomena caused by the deepening of the model layers. Figure 3 As shown in the figure, the main architecture of the network model is a convolutional layer, and the convolutional layer of the gated residual neural network model uses a 3×3 convolution kernel. The convolution operation of the gated residual neural network model is connected to a batch normalization operation and a nonlinear activation function.

[0101] At the same time, inspired by the long short-term memory network, the output layer of the gated residual neural network model is designed as an activation gate structure. The activation gate structure is composed of the ReLU nonlinear activation function and the Sigmoid nonlinear activation function. It is used to take into account the contextual feature relationship between images captured in consecutive time periods and improve the continuous processing ability of the neural network for continuous data.

[0102] The calculation formula of the ReLU nonlinear activation function is:

[0103] f(x) = max(0, x)

[0104] The calculation formula of Sigmoid nonlinear activation function is:

[0105]

[0106] The model training uses the cross entropy loss function, and the calculation formula is:

[0107]

[0108] Where N is the number of pixels in the image, c is the number of categories, and y ic is an indicator variable. If the true label of sample i is category c, then y ic =1, otherwise y ic =0, p ic is the probability that the model predicts sample i as category c, calculated by the softmax function;

[0109] a4. After the neural network completes the segmentation of satellite remote sensing images, the area of ​​each ecosystem is calculated through irregular two-dimensional integration and displayed in the form of a heat map. The irregular two-dimensional integration formula is: ∫∫Df(x, y)dxdy, where D is the two-dimensional integration area, f(x, y) is the integrand, and dxdy represents the integration of x and y.

[0110] In summary, the satellite remote sensing ecological environment detection method based on neural network, by adopting the satellite remote sensing image segmentation method based on neural network, can distinguish the different ecological environments and landforms of nature reserves in real time, quickly and accurately, and accurately calculate the area parameters of each ecological environment and landform, and intuitively display them in the form of a heat map. Compared with the traditional method that relies on manual statistical measurement, it greatly reduces the statistical cost, and has higher statistical accuracy and faster speed, and can achieve real-time data update.

[0111] In a specific embodiment, if Figure 4 As shown in the figure, the prediction and decision-making methods for species population changes include:

[0112] b1. Ecological management and research personnel of nature reserves use sensors to collect data on lake area, grassland area, forest area, temperature, humidity, soil pH, rainfall, snowfall, average daily light intensity, light duration and corresponding species quantity within a specified time period;

[0113] b2. Set the "corresponding number of species" as the predicted label data, and the remaining data as the factors that affect the change in the number of species, which are also the input data of the neural network model, using α1~α 13 The influence weight of each influencing factor is represented by β1~β 13 Indicates that the data set structure can be expressed as [α1β1…α 13 β 13 ], the constructed data set is divided into training set and test set in a ratio of 8:2;

[0114] b3. As the ecological environment of natural ecological protection areas is relatively closed, the number of species is relatively stable without special geological disasters and human destruction, so the natural birth and death rates of species are relatively balanced. Under the premise of known influencing factor values, the corresponding species number prediction value is:

[0115]

[0116] Among them, number t+1 is the predicted value of population size at the next moment, α it is the influencing factor value at the current moment, α i(t+1) is the influencing factor value at the next moment, γ tis the current population size. At this time, the species population prediction problem is transformed into the reasonable weight β of each influencing factor;

[0117] b4. Since the population size is a process that changes over time, a bidirectional long short-term memory neural network model is constructed. This neural network is improved on the basis of the long short-term memory neural network, so that the input data flows in both the forward and reverse directions at the same time, fully extracting the correlation characteristics between continuous data in the time context and improving the prediction accuracy. After the bidirectional long short-term memory neural network model is constructed, the constructed neural network model is trained on the training set. The model training uses the MSE loss function, which is expressed as follows:

[0118]

[0119] Among them, y i For actual results, Predict results for the neural network;

[0120] b5. After the model training is completed, save the corresponding model weight file, load the weight file and test set data, and test the model performance;

[0121] b6. Send the constructed data set into the bidirectional long short-term memory neural network model and give a reasonable weight configuration to perform population prediction in continuous time. The neural network model gives human intervention decision recommendations based on the prediction results and whether the environmental factors reach the ideal threshold.

[0122] In summary, this neural network-based species population change prediction and decision-making method can accurately and effectively explore the mapping relationship between environmental influencing factors and landforms and species population changes. Combined with current environmental parameters, it can accurately predict the changing trends of the ecological environment, landforms and species populations in nature reserves, and provide guidance and strategies for artificial intervention, providing a reference for artificial intervention protection.

[0123] In a specific embodiment, the specific operation steps of the species identification method include:

[0124] c1. Ecological management and research of nature reserves. Through manual photography and field hidden camera photography, researchers collect a large amount of image data of species in nature reserves and build a species image reference database.

[0125] c2. Grayscale each image to increase the model training speed;

[0126] c3. Convert the grayscale image to a different size, remove the background or blank area that does not contain any information, and prevent the image from being too large or having too many pixels;

[0127] c4. Mark the name and location of the species in each image, build a species identification dataset, and divide the species identification dataset into a training dataset and a test dataset in a ratio of 8:2;

[0128] c5. Build the SE-Inceptionv4 neural network model. This model is based on the Inceptionv4 model and adds the SE attention mechanism module. The feature extraction capability is better than that of the Inceptionv4 model, and the recognition accuracy is higher. The model structure is referenced Figure 5 , the SE-Inceptionv4 neural network model was trained on the training data set, and after verifying its good performance on the test data set, it was used to perform the species identification task.

[0129] Furthermore, the SE-Inceptionv4 neural network model uses convolutional layers and maximum pooling layers as its main architecture. The convolutional layers of the SE-Inceptionv4 neural network model use four sizes of convolutional kernels, namely 1*1, 3*3, 1*7 and 7*1, and make extensive use of feature splicing operations to fully integrate the multi-scale features extracted by convolutional kernels of different sizes. At this time, the SE attention mechanism module is introduced after the first feature splicing operation to improve the feature extraction capability of the model by scaling the features at the channel scale. In addition, the SE-Inceptionv4 neural network model also uses the maximum pooling layer operation in its model architecture, so that the model can obtain and retain the most prominent image features as much as possible under the constraint of a limited number of layers.

[0130] In summary, this species identification method, through the support of big data and computing resources, can achieve more accurate species identification, especially for similar species that are difficult to distinguish with the naked eye. The species identification is more accurate and efficient.

[0131] In a specific embodiment, the designated nature reserve "A" has a variety of ecosystems, including wetlands, forests, grasslands and lakes. The management personnel need to monitor the health status of these ecosystems so as to take protective measures in time. Therefore, the present invention has the following specific execution steps:

[0132] Step 1: Satellite image acquisition

[0133] Periodic image acquisition of Nature Reserve “A” using high-precision remote sensing cameras on geosynchronous orbit satellites;

[0134] Specifically, images will be collected once a week, and the images will include key landscapes such as lakes, forests, and grasslands in the ecological protection area "A";

[0135] Step 2: Image preprocessing and data annotation

[0136] The quality of the obtained raw image is improved through preprocessing steps such as denoising;

[0137] The denoising process may remove high-frequency noise through a low-pass filter, while the radiometric correction ensures that changes in lighting conditions do not affect the analysis results;

[0138] The processed image is then resized and annotated, wherein the annotation may include distinguishing wetland areas from forest areas to facilitate subsequent identification;

[0139] Step 3: Building and training the neural network model

[0140] Create a gated residual neural network (GRNN) model to process these remote sensing images;

[0141] The model is trained to recognize and distinguish different ecosystems in images, which includes extracting features and identifying objects such as forests and lakes and their boundaries;

[0142] In order to train the model, the labeled data is divided into a training set and a test set in a ratio of 8:2. The neural network model learns how to correctly segment and identify ecosystems on the training set;

[0143] Step 4: Remote sensing image segmentation and ecosystem area calculation

[0144] Use the trained GRNN model to analyze and segment the newly collected remote sensing images;

[0145] For a given remote sensing image, which shows that there is a lake in the area, the model can identify the boundary of the lake and calculate the exact area of ​​the lake using the irregular two-dimensional integral method;

[0146] Area results can be displayed as heat maps, providing researchers with visual evidence of lake size reduction or other ecological changes;

[0147] When data such as multiple images analyzed by the model indicate that the lake area is decreasing, managers can study the reasons in more detail and take timely protective measures, such as water replenishment and ecological restoration projects.

[0148] Furthermore, during the monitoring of the designated nature reserve “A”, it was noticed that the lake area monitored using remote sensing images had decreased. Further analysis by the system revealed that this might be due to continued drought and reduced rainfall. Therefore, the management team was able to adopt artificial rainfall and successfully alleviate the drought based on the rainfall reference value provided by the prediction model, thereby curbing the decline in species numbers and protecting the ecological balance.

[0149] On the other hand, the present invention further discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the above method.

[0150] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0151] In another embodiment provided in the present application, a computer program product comprising instructions is also provided, which, when executed on a computer, enables the computer to execute any of the intelligent ecological protection and management methods for nature reserves in the above-mentioned embodiments.

[0152] It is understandable that the system provided by the embodiment of the present invention corresponds to the method provided by the embodiment of the present invention, and the explanation, examples and beneficial effects of the relevant contents can refer to the corresponding parts in the above method.

[0153] The embodiment of the present application also provides an electronic device, including a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus.

[0154] Memory, used to store computer programs;

[0155] The processor is used to implement the above-mentioned intelligent ecological protection and management method for nature reserves when executing the program stored in the memory.

[0156] The communication bus mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc.

[0157] The communication interface is used for communication between the above electronic device and other devices.

[0158] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0159] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, and discrete hardware components.

[0160] It should also be noted that electronic devices also include terminal devices, which can also be called terminals, user equipment (UE), mobile stations (MS), mobile terminals (MT), etc. Terminal devices can be mobile phones, smart TVs, wearable devices, tablet computers (Pad), computers with wireless transceiver functions, virtual reality (VR) terminal devices, augmented reality (AR) terminal devices, wireless terminals in industrial control, wireless terminals in self-driving, wireless terminals in remote medical surgery, wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, etc. The embodiments of the present application do not limit the specific technology and specific device form adopted by the terminal devices.

[0161] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.

[0162] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

[0163] In addition, it should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components in a certain specific posture. If the specific posture changes, the directional indication will also change accordingly.

[0164] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present invention, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the meaning of "and / or" appearing in the full text includes three parallel schemes. Taking "A and / or B" as an example, it includes scheme A, or scheme B, or schemes that A and B meet at the same time. In addition, in the embodiments of the present invention, "multiple" refers to more than two. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

Claims

1. A method for intelligent ecological protection and management of nature reserves, applied to the ecological protection and management system of nature reserves, the system includes an intelligent perception subsystem, an intelligent prediction subsystem and an intelligent species identification subsystem, characterized in that: It includes satellite remote sensing ecological environment monitoring methods, species population change prediction and decision-making methods, and species identification methods, among which: The satellite remote sensing ecological environment monitoring method is applied to the intelligent perception subsystem. The remote sensing image segmentation is realized based on the satellite remote sensing image and the neural network model. It is used to accurately distinguish the mountain, lake, wetland, grassland, and forest ecosystems, and accurately calculate the area of ​​each ecosystem through two-dimensional integral operation. It is intuitively displayed in the form of a heat map to realize real-time monitoring of ecosystem changes. The species population change prediction and decision-making method is applied to the intelligent prediction subsystem, which is used to extract the nonlinear characteristic relationship between species population change and various environmental factors in the species population change prediction data set using a neural network model, and to achieve accurate prediction of species population change trends under the premise of known environmental factor change trends; The species identification method is applied to the intelligent species identification subsystem, which is used to accurately identify and distinguish similar species through big data and neural network models, and to assist in the accurate statistics of species populations in natural ecological reserves.

2. The method for intelligent ecological protection and management of nature reserves according to claim 1, characterized in that: The specific operation steps of the method include: S1. Obtain real-time satellite remote sensing images of ecological protection areas through geosynchronous orbit satellites; S2. Send the remote sensing image to the neural network model to distinguish the lake, grassland, mountain, jungle, and wetland ecological environment contained in the ecological protection area in the satellite remote sensing image through the remote sensing image segmentation technology based on the neural network model, circle the range in the remote sensing image with different colored curve frames, and mark the corresponding names; S3. Use satellite remote sensing scale to calculate the real length and width of each ecological environment at this moment, obtain the real area through irregular two-dimensional integral operation, and display it in real time using a heat map; S4. Use the heat map to determine whether the lake area continues to decrease and whether the forest and grassland greening areas continue to decrease. If so, it means that the drought in the nature reserve is gradually getting worse and the ecological environment is continuously deteriorating, requiring human intervention and protection; S5. Managers use sensors to obtain relevant environmental parameters of nature reserves and input them into the species population change prediction and decision-making neural network to obtain the changes in the number of species in nature reserves over a period of time in the future and determine whether there is a risk of reduced species diversity; S6. If there is a small decline in the number of species in a nature reserve, the cause of the deterioration of the ecological environment in the nature reserve is inferred based on the neural network model and corresponding operations are taken to curb the continued deterioration of the ecological environment; S7. At the same time, establish an ecological archive of species in nature reserves, embed a species identification method based on a neural network into a portable device, and aim it at the species to be identified in the wild. The portable device can accurately identify birds of similar appearance but different species through neural network calculations, so that managers of ecological reserves can understand and grasp the species situation in nature reserves, classify the rarity of species, and take necessary protection measures for rare and endangered species in a timely manner.

3. The method for intelligent ecological protection and management of nature reserves according to claim 1, characterized in that: The specific operation steps of the satellite remote sensing ecological environment monitoring method include: a1. Collect satellite remote sensing images of ecological protection areas through geosynchronous orbit satellites, and form remote sensing data sets through preprocessing and data annotation; a2. Divide the prepared remote sensing image dataset into training set and test set in a ratio of 8:2; a3. Construct a gated residual neural network model to complete model training on the training set, and verify that the performance reaches the preset standard on the test set, and then use the trained model to perform real-time remote sensing image segmentation tasks; a4. After the neural network completes the segmentation of satellite remote sensing images, the area of ​​each ecosystem is calculated through irregular two-dimensional integration and displayed in the form of a heat map. The irregular two-dimensional integration formula is: ∫∫Df(x, y)dxdy, where D is the two-dimensional integration area, f(x, y) is the integrand, and dxdy represents the integration of x and y.

4. The method for intelligent ecological protection and management of nature reserves according to claim 3, characterized in that: The steps of forming the remote sensing data set in step a1 include: a11. Use high-precision remote sensing cameras to continuously collect images of specific areas at a certain period; a12. Perform resizing, cropping, normalization, data enhancement and filtering operations on the collected images; a13. Perform image preprocessing by denoising, radiation correction, orthorectification, atmospheric correction, image enhancement, impact superposition, and coordinate transformation; a14. Data annotation involves interpreting remote sensing images, identifying various types of land features, and labeling them accordingly.

5. The method for intelligent ecological protection and management of nature reserves as claimed in claim 3, characterized in that: In step a3, the main structure of the gated residual neural network model is a convolution layer, and the convolution layer of the gated residual neural network model adopts a 3×3 convolution kernel, and the gated residual neural network model is connected to a batch normalization operation and a nonlinear activation function after the convolution operation; The output layer of the gated residual neural network model is designed as an activation gate structure, which is composed of a ReLU nonlinear activation function and a Sigmoid nonlinear activation function, and is used to take into account the context feature relationship between images captured in consecutive time periods and improve the continuous processing capability of the neural network for continuous data, wherein: The calculation formula of the ReLU nonlinear activation function is: f(x) = max(0, x) The calculation formula of Sigmoid nonlinear activation function is: The model training uses the cross entropy loss function, and the calculation formula is: Where N is the number of pixels in the image, c is the number of categories, and y ic is an indicator variable. If the true label of sample i is category c, then y ic =1, otherwise y ic =0, p ic is the probability that the model predicts sample i as category c, calculated by the softmax function.

6. The method for intelligent ecological protection and management of nature reserves according to claim 1, characterized in that: The species population change prediction and decision-making method includes: b1. Collect data on lake area, grassland area, forest area, temperature, humidity, soil pH, rainfall, snowfall, average daily light intensity, light duration and corresponding species quantity within a specified time period through sensors; b2. Set "the corresponding number of species" as the predicted label data, and the remaining data as the factors that affect the change in the number of species, which are also the input data of the neural network model, using α1~α 13 The influence weight of each influencing factor is represented by β1~β 13 Indicates that the data set structure can be expressed as [α1β1…α 13 β 13 ], the constructed data set is divided into training set and test set in the ratio of 8:2; b3. As the ecological environment of natural ecological protection areas is relatively closed, the number of species is relatively stable without special geological disasters and human destruction, so the natural birth and death rates of species are relatively balanced. Under the premise of known influencing factor values, the corresponding species number prediction value is: Among them, number t+1 is the predicted value of the population size at the next moment, α it is the influencing factor value at the current moment, α i(t+1) is the influencing factor value at the next moment, γ t is the current population size. At this time, the species population prediction problem is transformed into the reasonable weight β of each influencing factor; b4. Construct a bidirectional long short-term memory neural network model, and train the constructed neural network model on the training set. The MSE loss function is used in the model training, and the expression is as follows: Among them, y i For actual results, Predict results for the neural network; b5. After the model training is completed, save the corresponding model weight file, load the weight file and test set data, and test the model performance; b6. Send the constructed data set into the bidirectional long short-term memory neural network model and give a reasonable weight configuration to perform population prediction in continuous time. The neural network model gives human intervention decision recommendations based on the prediction results and whether the environmental factors reach the ideal threshold.

7. The method for intelligent ecological protection and management of nature reserves according to claim 1, characterized in that: The specific operation steps of the species identification method include: c1. Collect a large amount of image data of species in nature reserves through manual photography and hidden cameras in the wild, and build a species image reference database; c2. Grayscale each image to increase the model training speed; c3. Convert the grayscale image to a different size, remove the background or blank area that does not contain any information, and prevent the image from being too large or having too many pixels; c4. Mark the name and location of the species in each image, build a species identification dataset, and divide the species identification dataset into a training dataset and a test dataset in a ratio of 8:2; c5. Construct an SE-Inceptionv4 neural network model, complete model training on the training data set, and use it to perform species identification tasks after verifying good performance on the test data set. The SE-Inceptionv4 neural network model uses convolution layer and maximum pooling layer as the main architecture. The convolution layer of the SE-Inceptionv4 neural network model uses four sizes of convolution kernels: 1*1, 3*3, 1*7 and 7*1, and uses a large number of feature splicing operations to fully integrate the multi-scale features extracted by convolution kernels of different sizes. The SE attention mechanism module is introduced after the first feature concatenation operation to improve the model feature extraction capability by scaling features at the channel scale.

8. The method for intelligent ecological protection and management of nature reserves based on deep learning as claimed in claim 7, characterized in that: The SE-Inceptionv4 neural network model adopts a maximum pooling layer operation in the model architecture, so that the model can obtain and retain the most prominent image features as much as possible under the constraint of a limited number of layers.

9. A computer-readable storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 8.

10. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method according to any one of claims 1 to 8.

Images