Forestry pest and disease damage prediction method and system based on artificial intelligence

Abstract

The invention provides a forestry pest and disease prediction method and system based on artificial intelligence, and relates to the technical field of artificial intelligence, and the method comprises the steps: firstly obtaining a dynamic monitoring data set of a target forest region, covering an environment and vegetation physiological parameter sequence collected by a plurality of monitoring nodes, and a space coordinate region corresponding to each node; extracting features of the dynamic monitoring data set to obtain a forestry environment feature set containing vegetation growth, environment fluctuation and potential identification features of plant diseases and insect pests, and analyzing the forestry environment feature set by using a pre-trained plant disease and insect pest prediction model to generate a prediction parameter set representing the occurrence probability and influence range grade of the plant diseases and insect pests; executing dynamic prediction matching, determining a disease and insect pest distribution thermodynamic map and generating a control priority strategy, and finally fusing the disease and insect pest distribution thermodynamic map and the disease and insect pest control priority strategy to generate a forest region control optimization instruction set, and feeding back the instruction set to a forestry monitoring platform to trigger resource scheduling so as to realize efficient disease and insect pest control.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology. Specifically, it relates to a method and system for predicting forestry pests and diseases based on artificial intelligence. Background Art

[0002] Forestry is of great significance to ecological balance and resource supply. However, pests and diseases are important factors threatening its healthy development. Traditional prevention and control rely on manual work and experience, with low efficiency, difficulty in early detection, missed best opportunities, high costs and poor effects. There is an urgent need for accurate prediction of pests and diseases.

[0003] There are many deficiencies in existing forestry pest and disease prediction technologies. In terms of data collection, there are few monitoring points, limited scope, and single data types, mostly common environmental parameters, lacking data on vegetation physiology and the relationship between the environment and pests and diseases, which affects the accuracy of prediction. When processing data, the ability to extract and analyze features is weak, mostly extracting surface features, difficult to mine deep information, unable to capture early signals and potential laws. Prediction models are mostly based on traditional statistics or simple machine learning algorithms, difficult to process complex forestry environment data, and rely on a large amount of historical data. Due to the accidental and sudden occurrence of pests and diseases, it is difficult to ensure the integrity and accuracy of the data, and the generalization ability and prediction credibility of the models are low. In addition, existing technologies focus on prediction results and neglect decision-making support for prevention and control, unable to provide specific prevention and control strategies and resource scheduling plans, increasing the difficulty and uncertainty of prevention and control, resulting in waste of resources and poor effects. Summary of the Invention

[0004] In view of the problems mentioned above, in combination with the first aspect of the present invention, embodiments of the present invention provide a method for predicting forestry pests and diseases based on artificial intelligence, the method comprising: Obtain a set of dynamic monitoring data of the target forest area, the set of dynamic monitoring data including sequences of environmental parameters and sequences of vegetation physiological parameters collected by a plurality of monitoring nodes, wherein each monitoring node corresponds to a spatial coordinate area; Extract features from the set of dynamic monitoring data to obtain a set of forestry environment features for each monitoring node, the set of forestry environment features including vegetation growth state features, environmental fluctuation correlation features, and potential pest and disease identification features; Based on a pre-trained pest and disease prediction model, perform abnormal parameter analysis and processing on the set of forestry environment features to generate a set of pest and disease prediction parameters for the monitoring nodes, the set of pest and disease prediction parameters being used to characterize the probability of occurrence of pests and diseases and the level of the affected range; Perform dynamic prediction matching operations according to the set of pest and disease prediction parameters to determine the heat map of the distribution of pests and diseases in the target forest area, and generate a priority strategy for pest and disease prevention and control; Fuse the pest and disease distribution heat map with the pest and disease control priority strategy to generate an optimized forest area control instruction set, and feedback the optimized forest area control instruction set to the forestry monitoring platform to trigger the control resource scheduling operation.

[0005] In another aspect, an embodiment of the present invention further provides an artificial intelligence-based forest pest and disease prediction system, including a processor and a machine-readable storage medium. The machine-readable storage medium is connected to the processor. The machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0006] Based on the above aspects, the embodiment of the present invention realizes comprehensive, accurate and efficient prediction and control guidance for the pest and disease situation in the target forest area. First, obtain the dynamic monitoring data set of the target forest area, covering the environmental parameter sequence and vegetation physiological parameter sequence of multiple monitoring nodes, and each monitoring node corresponds to a specific spatial coordinate area. When extracting features from the dynamic monitoring data set to obtain the forestry environment feature set, not only focus on the vegetation growth state features, but also deeply explore the environmental fluctuation correlation features and potential pest and disease identification features, comprehensively describe the forestry environment situation from multiple dimensions, and perform abnormal parameter analysis on the forestry environment feature set based on the pre-trained pest and disease prediction model to generate a pest and disease prediction parameter set. This pest and disease prediction parameter set can accurately represent the occurrence probability and influence range level of pests and diseases. Through in-depth analysis and abnormal identification of features by model algorithms, pest and disease prediction is no longer limited to simple qualitative judgments, but realizes quantitative and accurate assessments. Further, perform dynamic prediction matching operations according to the pest and disease prediction parameter set to determine the pest and disease distribution heat map of the target forest area, and generate a pest and disease control priority strategy, converting discrete prediction parameters into an intuitive distribution heat map, clearly showing the distribution of pests and diseases in the forest area. At the same time, combined with the control priority strategy, the control work can be targeted, improving the control efficiency and effect. Finally, fuse the pest and disease distribution heat map with the pest and disease control priority strategy to generate an optimized forest area control instruction set, and feedback it to the forestry monitoring platform to trigger the control resource scheduling operation, realizing the full process automation and intelligence from prediction to control decision-making and then to resource scheduling. This not only improves the accuracy and timeliness of forest pest and disease prediction, but also provides scientific and accurate decision-making support for forest pest control work, effectively improving the overall level of forest pest control and resource utilization efficiency. Description of the Drawings

[0007] Figure 1 It is a schematic execution flow diagram of the artificial intelligence-based forest pest and disease prediction method provided by the embodiment of the present invention.

[0008] Figure 2It is a schematic diagram of the forestry pest and disease prediction system based on artificial intelligence provided by an embodiment of the present invention. Detailed implementation manners

[0009] The present invention will be specifically described below in conjunction with the accompanying drawings of the specification. Figure 1 It is a schematic flowchart of the forestry pest and disease prediction method based on artificial intelligence provided by an embodiment of the present invention. The forestry pest and disease prediction method based on artificial intelligence will be introduced in detail below.

[0010] Step S110: Obtain a dynamic monitoring data set of the target forest area, where the dynamic monitoring data set includes an environmental parameter sequence and a vegetation physiological parameter sequence collected by multiple monitoring nodes, and each monitoring node corresponds to a spatial coordinate area.

[0011] In this embodiment, in order to effectively predict pests and diseases in the target forest area, it is first necessary to obtain the dynamic monitoring data set of the target forest area. In the target forest area, a plurality of monitoring nodes are pre-arranged. These monitoring nodes are distributed at different positions in the forest area, and each monitoring node corresponds to a specific spatial coordinate area. The monitoring nodes continuously collect the environmental parameter sequence and the vegetation physiological parameter sequence. The environmental parameter sequence includes data such as temperature, humidity, light intensity, and soil pH value, and the vegetation physiological parameter sequence contains information such as chlorophyll content, trunk water saturation, and canopy density. For example, in a large target forest area, 100 monitoring nodes are evenly distributed, and each monitoring node collects environmental parameters and vegetation physiological parameters every 1 hour. Taking one of the monitoring nodes A as an example, the temperature sequence collected in a day is [20°C, 22°C, 23°C,..., 21°C], the humidity sequence is [60%, 62%, 61%,..., 63%], and the chlorophyll content sequence is [0.5mg / g, 0.52mg / g, 0.51mg / g,..., 0.53mg / g], etc. By collecting these data, a dynamic monitoring data set containing data of multiple monitoring nodes is formed.

[0012] Step S120: Extract features from the dynamic monitoring data set to obtain a forestry environment feature set for each monitoring node, where the forestry environment feature set includes vegetation growth state features, environmental fluctuation correlation features, and potential pest and disease identification features.

[0013] In this embodiment, feature extraction operations are performed on the obtained dynamic monitoring data set to obtain a forestry environment feature set for each monitoring node. This process can be further divided into the following sub-steps.

[0014] Step S121: Align the environmental parameter sequence in the time dimension to generate a standardized environmental parameter sequence.

[0015] In this embodiment, since there may be certain differences in the data collection times of different monitoring nodes, in order to facilitate subsequent analysis and processing, it is necessary to align the environmental parameter sequences in the time dimension. For example, the time intervals for monitoring node A and monitoring node B to collect data may not be exactly the same. Monitoring node A collects data every 1 hour, while monitoring node B collects data every 1.5 hours. Through time dimension alignment, the data of the two monitoring nodes are unified to the same time scale. The specific approach is to use the shorter time interval as the benchmark and perform interpolation processing on the data collected at the longer time interval. Assuming a time interval of 1 hour, for the data collected by monitoring node B every 1.5 hours, the data value at the 1-hour moment is calculated through linear interpolation. After such processing, a standardized environmental parameter sequence is generated, making the environmental parameter data of different monitoring nodes comparable in the time dimension.

[0016] Step S122: Invoke the pre-set vegetation physiological feature encoder to perform physiological state analysis processing on the vegetation physiological parameter sequence, generating a vegetation growth state feature vector, where the vegetation growth state feature vector includes the chlorophyll content change rate, trunk water saturation, and canopy density attenuation coefficient.

[0017] Then, invoke the pre-set vegetation physiological feature encoder to process the vegetation physiological parameter sequence. The vegetation physiological feature encoder is a trained model that can analyze the physiological state information of the vegetation based on the input vegetation physiological parameter sequence. For example, for the chlorophyll content sequence, by calculating the ratio of the difference in chlorophyll content between adjacent time points to the chlorophyll content at the previous moment, the chlorophyll content change rate is obtained. Assuming that within a certain time period, the chlorophyll content sequence of monitoring node A is [0.5mg / g, 0.52mg / g, 0.51mg / g], then the chlorophyll content change rate for the first time interval is (0.52 - 0.5) / 0.5 = 0.04. Similarly, for the trunk water saturation and canopy density data, after being processed by the encoder, the trunk water saturation and canopy density attenuation coefficient are obtained. Finally, these features are combined into a vegetation growth state feature vector, such as [0.04, 0.8, 0.02], where 0.04 represents the chlorophyll content change rate, 0.8 represents the trunk water saturation, and 0.02 represents the canopy density attenuation coefficient.

[0018] Step S123: Perform environmental correlation analysis processing on the standardized environmental parameter sequence to extract environmental fluctuation correlation features, where the environmental fluctuation correlation features include the temperature-humidity co-variation gradient, soil pH offset, and cumulative light intensity deviation value.

[0019] Further, perform an environmental correlation analysis on the standardized environmental parameter sequence to extract environmental fluctuation correlation features. This process can be carried out according to the following specific steps.

[0020] Step S1231: Obtain the temperature sequence, humidity sequence, and light intensity sequence in the standardized environmental parameter sequence.

[0021] In this embodiment, the temperature sequence, humidity sequence, and light intensity sequence are extracted from the standardized environmental parameter sequence. For example, in the standardized environmental parameter sequence of monitoring node A, the temperature sequence is [20°C, 22°C, 23°C, 21°C], the humidity sequence is [60%, 62%, 61%, 63%], and the light intensity sequence is [1000 lux, 1100 lux, 1050 lux, 1080 lux].

[0022] Step S1232: Calculate the dynamic covariance matrix between the temperature sequence and the humidity sequence, and determine the temperature-humidity co-variation gradient, where the temperature-humidity co-variation gradient is quantified by the covariance change rate within a sliding time window.

[0023] In this embodiment, the method of a sliding time window can be adopted. Assume the window size is 3 time points. For the temperature sequence [20°C, 22°C, 23°C, 21°C] and the humidity sequence [60%, 62%, 61%, 63%], the temperature data within the first window is [20°C, 22°C, 23°C], and the humidity data is [60%, 62%, 61%]. Calculate the covariance of these two sequences. Then, slide the window backward by one time point and calculate the covariance within the second window. By comparing the covariance values of adjacent windows, obtain the covariance change rate to quantify the temperature-humidity co-variation gradient. For example, the covariance of the first window is 0.5, and the covariance of the second window is 0.6. Then the covariance change rate is (0.6 - 0.5) / 0.5 = 0.2, and this value is the temperature-humidity co-variation gradient within this time period.

[0024] Step S1233: Perform a trend fitting process on the soil pH data in the standardized environmental parameter sequence, and extract the soil pH offset, where the soil pH offset is determined by the standard deviation multiple of the current pH value and the historical reference value.

[0025] In this embodiment, first, a historical reference value is established. This historical reference value can be the average value of the soil pH over a past period of time. For example, the average value of the soil pH of monitoring node A in the past month is 7.0. Then, the current soil pH data is fitted to obtain the current pH value. Assume the current pH value is 7.2. Calculate the difference between the current value and the historical reference value, that is, 7.2 - 7.0 = 0.2. Then calculate the standard deviation of the historical data. Assume the standard deviation is 0.1. Then the soil pH offset is 0.2 / 0.1 = 2 times the standard deviation.

[0026] Step S1234: Calculate the cumulative deviation of the light intensity sequence to generate a light intensity cumulative deviation value, where the cumulative deviation value is the cumulative difference between the measured light intensity value and the theoretical value within a continuous time interval.

[0027] In this embodiment, the theoretical light intensity value can be calculated based on factors such as the local geographical location and time. For example, according to the geographical location and time of monitoring node A, the theoretical light intensity sequence is calculated as [980 lux, 1020 lux, 1060 lux, 1040 lux], while the measured light intensity sequence is [1000 lux, 1100 lux, 1050 lux, 1080 lux]. Calculate the difference between the measured value and the theoretical value at each time point, which are (1000 - 980) = 20 lux, (1100 - 1020) = 80 lux, (1050 - 1060) = -10 lux, (1080 - 1040) = 40 lux. Then accumulate these differences to obtain the light intensity cumulative deviation value of 20 + 80 - 10 + 40 = 130 lux.

[0028] Step S1235: Normalize the temperature - humidity co - variation gradient, the soil pH offset, and the light intensity cumulative deviation value to obtain the environmental fluctuation correlation feature.

[0029] In this embodiment, the purpose of normalization is to map these feature values to a unified range for subsequent analysis and processing. For example, using the min - max normalization method, assume the value range of the temperature - humidity co - variation gradient is [0, 1], the value range of the soil pH offset is [0, 5], and the value range of the light intensity cumulative deviation value is [0, 200]. Normalize the temperature - humidity co - variation gradient of 0.2, the soil pH offset of 2, and the light intensity cumulative deviation value of 130 to obtain the normalized environmental fluctuation correlation feature vector, such as [0.2, 0.4, 0.65].

[0030] Step S124: Perform anomaly identification fusion processing on the vegetation growth state feature vector and the environmental fluctuation correlation feature to generate potential pest and disease identification features, where the potential pest and disease identification features include the estimated value of the egg distribution density, the fungal infection diffusion rate, and the disease symptom manifestation intensity.

[0031] Next, perform anomaly identification fusion processing on the vegetation growth state feature vector and the environmental fluctuation correlation feature to generate potential pest and disease identification features. This process can be carried out according to the following specific steps.

[0032] For example, step S1241: Perform discretization and segmentation processing on the chlorophyll content change rate, trunk moisture saturation, and canopy density attenuation coefficient in the vegetation growth state feature vector to generate the chlorophyll content change rate interval, the trunk moisture saturation interval, and the canopy density attenuation coefficient interval.

[0033] In this embodiment, discretization and segmentation processing are performed on each feature in the vegetation growth state feature vector. For example, for the chlorophyll content change rate, its value range is divided into multiple intervals, such as [0, 0.02), [0.02, 0.05), [0.05, +∞). Suppose the chlorophyll content change rate of monitoring node A is 0.04, then it falls within the interval [0.02, 0.05). Similarly, similar segmentation processing is performed on the trunk moisture saturation and the canopy density attenuation coefficient to obtain the trunk moisture saturation interval and the canopy density attenuation coefficient interval.

[0034] Step S1242: Determine the estimated value of the egg distribution density according to the mapping relationship between the temperature and humidity co-variation gradient in the environmental fluctuation correlation feature and the chlorophyll content change rate interval.

[0035] In this embodiment, a mapping table can be established in advance, which records the estimated values of the egg distribution density under different combinations of the temperature and humidity co-variation gradient and the chlorophyll content change rate interval. For example, when the temperature and humidity co-variation gradient is 0.2 and the chlorophyll content change rate interval is [0.02, 0.05), by querying the mapping table, the estimated value of the egg distribution density is obtained as 50 eggs / m².

[0036] Step S1243: Generate the fungal infection diffusion rate based on the soil acidity and alkalinity offset and the cumulative light intensity deviation value in the environmental fluctuation correlation feature, in combination with the predefined fungal growth threshold range.

[0037] In this embodiment, first, determine the suitable soil pH range and light intensity range for the growth of fungi. For example, the suitable soil pH range for the growth of fungi is [6.5, 7.5], and the suitable light intensity range is [800 lux, 1200 lux]. Then, judge whether the current environment is suitable for the growth of fungi according to the soil pH offset and the cumulative deviation value of the light intensity. If the soil pH offset is small and the cumulative deviation value of the light intensity is also within a certain range, it is considered that the environment is suitable for the growth of fungi, and calculate the fungal infection and diffusion rate according to the pre-established model. Suppose the soil pH offset is 2 times the standard deviation and the cumulative deviation value of the light intensity is 130 lux. The fungal infection and diffusion rate calculated by the model is 0.1 m² / day.

[0038] Step S1244: Determine the disease symptom manifestation intensity according to the overlapping area ratio of the trunk water saturation interval and the canopy density attenuation coefficient interval.

[0039] In this embodiment, first, calculate the length of the overlapping area of the two intervals, and then divide it by the total length of the two intervals to obtain the overlapping area ratio. For example, the trunk water saturation interval is [0.7, 0.9], the canopy density attenuation coefficient interval is [0.01, 0.03], the overlapping area length is 0, and the total length is (0.9 - 0.7) + (0.03 - 0.01) = 0.22. Then the overlapping area ratio is 0 / 0.22 = 0. According to the pre-established mapping relationship, when the overlapping area ratio is 0, the disease symptom manifestation intensity is weak.

[0040] Step S1245: Normalize, weight, and splice the estimated value of the egg distribution density, the fungal infection and diffusion rate, and the disease symptom manifestation intensity to generate the potential identification feature of the pests and diseases.

[0041] In this embodiment, first, normalize the estimated value of the egg distribution density, the fungal infection and diffusion rate, and the disease symptom manifestation intensity, and map them to the range of [0, 1]. For example, the estimated value of the egg distribution density is 50 pieces / m², and after normalization, it is 0.5; the fungal infection and diffusion rate is 0.1 m² / day, and after normalization, it is 0.3; the disease symptom manifestation intensity is weak, which is represented by 0.1. Then, assign corresponding weights according to the importance of different features. Suppose the weight of the estimated value of the egg distribution density is 0.5, the weight of the fungal infection and diffusion rate is 0.3, and the weight of the disease symptom manifestation intensity is 0.2. Multiply the normalized feature values by the weights and then splice them to obtain the potential identification feature vector of the pests and diseases, such as [0.5×0.5, 0.3×0.3, 0.2×0.1] = [0.25, 0.09, 0.02].

[0042] Step S125: Integrate the vegetation growth state features, the environmental fluctuation correlation features, and the potential pest identification features to form the forestry environment feature set.

[0043] In this embodiment, for example, the vegetation growth state feature vector is [0.04, 0.8, 0.02], the environmental fluctuation correlation feature vector is [0.2, 0.4, 0.65], and the potential pest identification feature vector is [0.25, 0.09, 0.02]. Concatenate them together to form the forestry environment feature set, such as [0.04, 0.8, 0.02, 0.2, 0.4, 0.65, 0.25, 0.09, 0.02].

[0044] Step S130: Based on the pre-trained pest prediction model, perform anomaly parameter analysis on the forestry environment feature set to generate the pest prediction parameter set of the monitoring node, where the pest prediction parameter set is used to characterize the probability of pest occurrence and the level of the affected range.

[0045] In this embodiment, use the pre-trained pest prediction model to process the forestry environment feature set to generate the pest prediction parameter set of the monitoring node. This process can be further divided into the following sub-steps.

[0046] Step S131: Input the forestry environment feature set into the feature encoding layer of the pest prediction model to generate a high-dimensional fusion feature vector.

[0047] In this embodiment, the feature encoding layer is a component of the model, which can convert the input low-dimensional features into high-dimensional features to better capture the information in the data. For example, the forestry environment feature set is [0.04, 0.8, 0.02, 0.2, 0.4, 0.65, 0.25, 0.09, 0.02]. After being processed by the feature encoding layer, a high-dimensional fusion feature vector is generated, such as [0.1, 0.2, 0.3, …, 0.05], and the dimension of the vector may be much higher than that of the input forestry environment feature set.

[0048] Step S132: Through the anomaly detection layer of the pest prediction model, calculate the deviation coefficient between the high-dimensional fusion feature vector and the pre-stored healthy vegetation feature template, where the deviation coefficient is determined by the weighted sum of the Euclidean distance and the cosine similarity.

[0049] Next, calculate the deviation coefficient between the high-dimensional fusion feature vector and the pre-stored healthy vegetation feature template through the anomaly detection layer of the pest prediction model. The specific steps are as follows.

[0050] Step S1321: Obtain the standard chlorophyll content, the standard trunk moisture saturation, and the standard canopy density in the healthy vegetation feature template.

[0051] In this embodiment, the standard chlorophyll content, the standard trunk moisture saturation, and the standard canopy density are obtained from the pre-stored healthy vegetation feature template. For example, the standard chlorophyll content recorded in the healthy vegetation feature template is 0.6 mg / g, the standard trunk moisture saturation is 0.9, and the standard canopy density is 0.8.

[0052] Step S1322: Calculate the absolute value of the first difference between the chlorophyll content change rate in the high-dimensional fusion feature vector and the standard chlorophyll content, and the absolute value of the second difference between the trunk moisture saturation and the standard trunk moisture saturation.

[0053] In this embodiment, the chlorophyll content change rate and the trunk moisture saturation information are extracted from the high-dimensional fusion feature vector. Assume that the element value corresponding to the chlorophyll content change rate in the high-dimensional fusion feature vector is 0.04, and the element value corresponding to the trunk moisture saturation is 0.8. Calculate the absolute value of the first difference as |0.04 - 0.6| = 0.56, and the absolute value of the second difference as |0.8 - 0.9| = 0.1.

[0054] Step S1323: Normalize the absolute value of the first difference and the absolute value of the second difference to generate a primary deviation index.

[0055] Normalize the absolute value of the first difference and the absolute value of the second difference. Assume that the value range of the chlorophyll content change rate is [0, 0.1], and the value range of the trunk moisture saturation is [0, 1]. Normalize the absolute value of the first difference 0.56 and the absolute value of the second difference 0.1. Since 0.56 exceeds the range of the chlorophyll content change rate [0, 0.1], after normalization, it gets 1 (taking the maximum value). The absolute value of the second difference 0.1 is within the range of the trunk moisture saturation [0, 1], and after normalization, it remains 0.1. Then, perform a weighted average on these two normalized values. Assume that the weight of the chlorophyll content change rate is 0.6, and the weight of the trunk moisture saturation is 0.4. Then the primary deviation index is 1×0.6 + 0.1×0.4 = 0.64.

[0056] Step S1324: Extract the canopy density attenuation coefficient in the high-dimensional fusion feature vector, calculate the reciprocal of its ratio to the standard canopy density, and generate a secondary deviation index.

[0057] In this embodiment, assume that the element value corresponding to the canopy density attenuation coefficient in the high-dimensional fusion feature vector is 0.02, and the standard canopy density is 0.8. Calculate the ratio of the canopy density attenuation coefficient to the standard canopy density as 0.02 / 0.8 = 0.025, and the reciprocal of its ratio is 1 / 0.025 = 40. This 40 is the secondary deviation index.

[0058] Step S1325: Linearly combine the primary deviation index and the secondary deviation index to obtain a comprehensive deviation coefficient, where the combination weights of the linear combination are dynamically adjusted according to the feature importance in the historical pest and disease data.

[0059] Linearly combine the primary deviation index 0.64 and the secondary deviation index 40. Suppose, based on the analysis of the feature importance in the historical pest and disease data, it is determined that the weight of the primary deviation index is 0.3 and the weight of the secondary deviation index is 0.7. Then the comprehensive deviation coefficient is 0.64×0.3 + 40×0.7 = 0.192 + 28 = 28.192.

[0060] Step S133: Match the preset pest and disease type mapping table according to the deviation coefficient to determine the pest and disease occurrence probability, where the pest and disease occurrence probability has an exponential relationship with the deviation coefficient.

[0061] In this embodiment, the preset pest and disease type mapping table records the pest and disease occurrence probabilities corresponding to different deviation coefficient ranges. For example, when the deviation coefficient is between 0 - 10, the pest and disease occurrence probability is 10%; when the deviation coefficient is between 10 - 20, the pest and disease occurrence probability is 30%; when the deviation coefficient is between 20 - 30, the pest and disease occurrence probability is 60%; when the deviation coefficient is greater than 30, the pest and disease occurrence probability is 90%. Since the comprehensive deviation coefficient is 28.192, which falls within the range of 20 - 30, the pest and disease occurrence probability of this monitoring node is determined to be 60%.

[0062] Step S134: Invoke the impact range calculation module of the pest and disease prediction model, and calculate the pest and disease impact range level based on the spatial association parameters in the high - dimensional fusion feature vector, where the pest and disease impact range level is determined by the chain propagation rate of the feature similarity of adjacent monitoring nodes.

[0063] In this embodiment, first, extract the spatial association parameters from the high - dimensional fusion feature vector. These spatial association parameters may include information such as the spatial coordinates of the monitoring node and the distances to adjacent monitoring nodes. Suppose there are three adjacent monitoring nodes B, C, and D around monitoring node A, and calculate the feature similarity between monitoring node A and the adjacent monitoring nodes. The feature similarity can be obtained by calculating the cosine similarity between the high - dimensional fusion feature vectors. For example, the high - dimensional fusion feature vector of monitoring node A is [0.1, 0.2, 0.3, …, 0.05], and the high - dimensional fusion feature vector of monitoring node B is [0.12, 0.22, 0.28, …, 0.06]. Calculate the cosine similarity between them to be 0.9. Similarly, calculate the feature similarities between monitoring node A and monitoring nodes C and D to be 0.8 and 0.7 respectively.

[0064] Then, determine the chain propagation rate based on the feature similarity. Assume that when the feature similarity is greater than 0.8, the chain propagation rate is to spread to 1 adjacent monitoring node per day; when the feature similarity is between 0.6 and 0.8, the chain propagation rate is to spread to 1 adjacent monitoring node every two days; when the feature similarity is less than 0.6, the chain propagation rate is to spread to 1 adjacent monitoring node every three days. For the case of monitoring node A and its adjacent monitoring nodes, the chain propagation rate with node B is to spread to 1 adjacent monitoring node per day, the chain propagation rate with node C is to spread to 1 adjacent monitoring node every two days, and the chain propagation rate with node D is to spread to 1 adjacent monitoring node every two days.

[0065] Predict the spread range of the pests and diseases within a certain period according to the chain propagation rate and the distribution of adjacent monitoring nodes. Assume that with a one-week time period, node B will be infected within one week, and there is a 50% probability that nodes C and D will be infected within one week. Based on these prediction results, determine the level of the impact range of the pests and diseases. If the impact range is small and only involves a few adjacent monitoring nodes, the impact range level is low; if the impact range is large and involves more adjacent monitoring nodes, the impact range level is high. Assume that in this example, the impact range level of the pests and diseases is medium.

[0066] Step S135: Combine and package the occurrence probability of the pests and diseases and the impact range level to generate the set of pest and disease prediction parameters.

[0067] Combine and package the calculated pest and disease occurrence probability of 60% and the impact range level of "medium" to generate the set of pest and disease prediction parameters. For example, it can be expressed as [60%, "medium"], and this set of pest and disease prediction parameters is used to characterize the occurrence probability of the pests and diseases and the impact range level of this monitoring node.

[0068] Step S140: Perform a dynamic prediction matching operation according to the set of pest and disease prediction parameters, determine the heat map of the distribution of pests and diseases in the target forest area, and generate a pest and disease control priority strategy.

[0069] In this step, a dynamic prediction matching operation needs to be performed according to the set of pest and disease prediction parameters to determine the heat map of the distribution of pests and diseases in the target forest area and generate a pest and disease control priority strategy. This process can be divided into the following sub-steps.

[0070] Step S141: Extract the occurrence probability of the pests and diseases and the impact range level from the set of pest and disease prediction parameters.

[0071] In this embodiment, the occurrence probability and the influence range level of pests and diseases are extracted from the previously generated set of prediction parameters for pests and diseases. For example, for the set of prediction parameters for pests and diseases of monitoring node A [60%, "medium"], the occurrence probability of pests and diseases is extracted as 60%, and the influence range level is "medium". Such extraction operations are performed on the sets of prediction parameters for pests and diseases of all monitoring nodes in the target forest area to obtain the occurrence probability and influence range level data of each monitoring node.

[0072] Step S142: According to the spatial coordinate area of the monitoring node, a three-dimensional geographic information grid is constructed, and the occurrence probability of pests and diseases of each grid node is mapped to the corresponding spatial coordinates.

[0073] A three-dimensional geographic information grid is constructed according to the spatial coordinate area of the monitoring node. First, the geographical range of the target forest area is determined. Assume that the target forest area is a region with a length of 1000 meters, a width of 800 meters, and a height (considering a certain vertical space, such as the height of trees) of 50 meters. This area is divided into three-dimensional grids. For example, the size of each grid is 10 meters × 10 meters × 5 meters.

[0074] Then, the spatial coordinates of each monitoring node are corresponding to a certain grid node in the three-dimensional geographic information grid. Assume that the spatial coordinates of monitoring node A are (200, 300, 10), then it corresponds to the grid node in the 20th row, 30th column, and 2nd layer of the three-dimensional geographic information grid. The occurrence probability of pests and diseases of this monitoring node, 60%, is mapped to this corresponding grid node. Such mapping operations are performed on all monitoring nodes, so that each grid node has a corresponding occurrence probability value of pests and diseases.

[0075] Step S143: Based on the Kriging interpolation algorithm, spatial interpolation processing is performed on the discrete occurrence probability of pests and diseases to generate a continuous probability distribution surface.

[0076] In this step, the Kriging interpolation algorithm is used to perform spatial interpolation processing on the discrete occurrence probability of pests and diseases. The specific steps are as follows.

[0077] Step S1431: Based on the spatial coordinates of the monitoring node, a spatial distance matrix between the monitoring nodes is generated.

[0078] In this embodiment, the spatial distance between any two monitoring nodes is calculated according to the spatial coordinates of the monitoring nodes. Assume that the spatial coordinates of monitoring node A are (200, 300, 10), and the spatial coordinates of monitoring node B are (210, 310, 12). Their distance is calculated as the square root of ((210 - 200)^2 + (310 - 300)^2 + (12 - 10)^2) = the square root of (100 + 100 + 4) = the square root of 204 ≈ 14.28 meters according to the spatial distance calculation formula (such as the three-dimensional Euclidean distance formula).

[0079] Perform such distance calculations for all monitoring nodes to obtain a spatial distance matrix between the monitoring nodes. For example, there are three monitoring nodes A, B, and C, and the spatial distance matrix is as follows: [0, 14.28, 20.5], [14.28, 0, 18.3], [20.5, 18.3, 0] Step S1432: Calculate the semi-variogram model parameters based on the spatial distance matrix and construct a semi-variogram model.

[0080] The semi-variogram is used to describe the variability of spatial data. Based on the spatial distance matrix and the corresponding pest occurrence probability data, calculate the parameters of the semi-variogram model. First, group the monitoring nodes according to the distance and calculate the average value of the squared differences in the pest occurrence probabilities of the monitoring nodes within each group. For example, for the monitoring nodes within the distance range of 0 - 10 meters, calculate the average value of the squared differences in their pest occurrence probabilities.

[0081] Then, fit a semi-variogram model based on these calculation results. Common semi-variogram models include the spherical model, exponential model, etc. Assume the spherical model is used, and through fitting, obtain the parameters of the model, such as the nugget effect, sill, and range. Assume the parameters of the fitted spherical model are: the nugget effect is 0.05, the sill is 0.3, and the range is 50 meters. Construct a semi-variogram model based on these parameters.

[0082] Step S1433: Calculate the interpolation weight coefficients between the target grid points and the monitoring nodes based on the semi-variogram model parameters and the spatial distance matrix.

[0083] For each target grid point in the three-dimensional geographic information grid, calculate its spatial distance from all monitoring nodes. Then, calculate the interpolation weight coefficients based on the semi-variogram model and these distances. Assume the coordinates of the target grid point G are (205, 305, 11), its distance from monitoring node A is 5 meters, its distance from monitoring node B is 12 meters, and its distance from monitoring node C is 22 meters.

[0084] Substitute these distances into the semi-variogram model to obtain the corresponding semi-variogram values. Then, according to the principle of Kriging interpolation, solve a set of linear equations to calculate the interpolation weight coefficients between the target grid point G and the monitoring nodes A, B, and C. Assume the solved weight coefficients are 0.4, 0.3, and 0.3 respectively.

[0085] Step S1434: Perform a weighted sum of the pest occurrence probabilities of the monitoring nodes based on the interpolation weight coefficients to generate the interpolation probability value of the target grid point.

[0086] In this embodiment, it is assumed that the probability of pest and disease occurrence at monitoring node A is 60%, the probability of pest and disease occurrence at monitoring node B is 55%, and the probability of pest and disease occurrence at monitoring node C is 65%. The interpolation probability value of the target grid point G is 0.4×60% + 0.3×55% + 0.3×65% = 0.4×0.6 + 0.3×0.55 + 0.3×0.65 = 0.24 + 0.165 + 0.195 = 0.6.

[0087] Step S1435: Traverse all target grid points in the three-dimensional geographic information grid to generate a continuous probability distribution surface covering the target forest area.

[0088] In this embodiment, by traversing each target grid point, the interpolation probability value of each grid point is obtained. Visualizing these interpolation probability values in the three-dimensional geographic information grid can generate a continuous probability distribution surface covering the target forest area. This surface intuitively shows the probability of pest and disease occurrence at different positions within the target forest area.

[0089] Step S1436: Perform Gaussian filtering on the continuous probability distribution surface to eliminate the local mutation noise of the interpolation probability value, and output the smoothed continuous probability distribution surface.

[0090] To eliminate the possible local mutation noise in the continuous probability distribution surface, Gaussian filtering is performed on it. Gaussian filtering is a linear smoothing filtering method, and the surface can be smoothed through a convolution operation. Select appropriate Gaussian kernel size and standard deviation, for example, the Gaussian kernel size is 3×3×3 and the standard deviation is 1.

[0091] For each grid point in the continuous probability distribution surface, the interpolation probability values of the grid points in its neighborhood are weighted and averaged according to the weights of the Gaussian kernel. For example, for the target grid point G, there are 26 grid points in its neighborhood (in three-dimensional space), and the interpolation probability values of these 26 grid points are weighted and averaged according to the weights of the Gaussian kernel to obtain the smoothed interpolation probability value. Such processing is performed on all grid points, and the smoothed continuous probability distribution surface is output.

[0092] Step S144: Perform regional segmentation on the continuous probability distribution surface according to the influence range level, divide it into multiple partitions, and assign a color gradient identifier to each partition.

[0093] In this embodiment, the continuous probability distribution surface is segmented into regions according to the influence range level of each monitoring node. For the region with a low influence range level, a relatively small range is marked on the continuous probability distribution surface; for the region with a medium influence range level, a medium-sized range is marked; and for the region with a high influence range level, a large range is marked.

[0094] For example, for monitoring node A with a medium influence range level, a medium-sized region is divided on the continuous probability distribution surface with the grid point corresponding to this monitoring node as the center. Such regional division operations are performed on all monitoring nodes to obtain multiple partitions.

[0095] Then, a color gradient identifier is assigned to each partition. Different colors can be used to represent different ranges of pest occurrence probabilities. For example, the region with a pest occurrence probability of 0 - 20% is represented by green, 20% - 40% by yellow, 40% - 60% by orange, 60% - 80% by red, and 80% - 100% by dark red. In this way, the pest occurrence probability situation of each partition can be intuitively seen through the color gradient identifier.

[0096] Step S145: Superimpose the color gradient identifier on the three-dimensional geographic information grid to generate the heat map of the pest distribution, where the heat intensity is proportional to the product of the pest occurrence probability and the influence range level.

[0097] In this embodiment, the color gradient identifier of each partition is corresponding to the corresponding region in the three-dimensional geographic information grid, so that the entire three-dimensional geographic information grid presents a distribution of different colors, forming the heat map of the pest distribution.

[0098] The heat intensity is proportional to the product of the pest occurrence probability and the influence range level. For example, for a partition with an influence range level of "medium" and a pest occurrence probability of 60%, its heat intensity is 60% × medium (assuming the value corresponding to "medium" is 0.5) = 0.3. According to this heat intensity, the color display intensity of this partition in the heat map is adjusted to make the color more vivid to highlight the severity of pests in this region.

[0099] Step S146: Generate a pest control priority strategy, which specifically includes the following sub-steps.

[0100] Step S1461: Obtain the set of coordinates of the significant aggregation regions in the heat map of the pest distribution and the corresponding pest type identifiers.

[0101] In this embodiment, the significant aggregation area refers to the area where the occurrence probability of pests and diseases is relatively high and concentrated. By setting a threshold for the occurrence probability of pests and diseases, for example, 70%, the areas where the occurrence probability of pests and diseases is greater than 70% are marked as significant aggregation areas.

[0102] Obtain the coordinate sets of these significant aggregation areas. For example, there are three significant aggregation areas, and their coordinate sets are [(200 - 220, 300 - 320, 0 - 10), (400 - 420, 500 - 520, 0 - 10), (600 - 620, 700 - 720, 0 - 10)]. At the same time, according to the previous pest and disease prediction results, obtain the pest and disease type identifiers corresponding to these areas, such as "Pine wilt disease", "Poplar anthracnose", etc.

[0103] Step S1462: According to the historical control record database, match the control resource requirement list corresponding to each pest and disease type. The control resource requirement list includes the pesticide type, the number of equipment, and the manpower allocation.

[0104] Query the historical control record database, which records the control resource information used in the previous control processes of various pests and diseases. For each identified pest and disease type, match its corresponding control resource requirement list.

[0105] For example, for "Pine wilt disease", the historical control records show that the "nematicide" pesticide needs to be used, 10 sprayers and 5 drilling machines are required for equipment, and 20 professional control personnel are required for manpower allocation. For "Poplar anthracnose", the "carbendazim" pesticide needs to be used, 5 sprayers, and 10 control personnel. In this way, the corresponding control resource requirement list is determined for the pest and disease types in each significant aggregation area.

[0106] Step S1463: Calculate the urgency score for each area in the coordinate set of the significant aggregation area, which specifically includes the following sub-steps.

[0107] Step S14631: Obtain the pest and disease occurrence probability P and the influence range level R in the pest and disease prediction parameter set.

[0108] From the previously generated pest and disease prediction parameter set, obtain the corresponding pest and disease occurrence probability P and the influence range level R for each significant aggregation area. For example, for the first significant aggregation area, its pest and disease prediction parameter set is [80%, "high"], then the obtained pest and disease occurrence probability P is 80%, and the influence range level R is "high" (assuming the value corresponding to "high" is 0.8).

[0109] Step S14632: Query the vegetation economic value database of the target forest area, and extract the economic value coefficient V of the corresponding area. The economic value coefficient is comprehensively calculated from the vegetation type, tree age, and market unit price.

[0110] Query the vegetation economic value database of the target forest area, which records the vegetation economic value information of different areas within the target forest area. According to the coordinates of the significant aggregation area, extract the economic value coefficient V of the corresponding area.

[0111] The economic value coefficient V is comprehensively calculated from the vegetation type, tree age, and market unit price. For example, the vegetation in a certain area is mainly pine trees, with a tree age of 20 years and a market unit price of 1000 yuan per cubic meter. According to the pre-set calculation method, the economic value coefficient V of this area is calculated to be 0.6.

[0112] Step S14633: Calculate the initial emergency level score through the formula S = P × R × V.

[0113] Use the obtained pest occurrence probability P, influence range level R, and economic value coefficient V to calculate the initial emergency level score S. Taking the data in the previous example as an example, the pest occurrence probability P is 80% (i.e., 0.8), the value corresponding to the influence range level R is 0.8, and the economic value coefficient V is 0.6. Then the initial emergency level score S is equal to 0.8 multiplied by 0.8 and then multiplied by 0.6, and it is calculated that S is 0.384.

[0114] Step S14634: If the current inventory of prevention and control resources is lower than the demand threshold, increase the initial emergency level score by a preset proportion to obtain the corrected emergency level score.

[0115] Query the inventory of the current prevention and control resources and compare it with the demand threshold. The demand threshold is a resource quantity standard set based on historical prevention and control experience and the current pest situation. Assuming that for the prevention and control of "pine wilt disease", the demand threshold for the pesticide "nematicide" is 1000 liters, and the current inventory is 800 liters, which is lower than the demand threshold.

[0116] The preset increase proportion is comprehensively determined according to the degree of resource shortage and the severity of the pest. Assuming the preset proportion is 20%, then the corrected emergency level score is equal to the initial emergency level score 0.384 multiplied by (1 + 20%), that is, 0.384 multiplied by 1.2, and it is calculated that the corrected emergency level score is 0.4608.

[0117] Step S14635: Normalize the corrected emergency level score to obtain the final emergency level score and write it into the prevention and control task priority queue.

[0118] Normalization is to map the corrected urgency scores to a unified range for easy comparison and sorting. Assume the normalization range is [0, 1]. First, find the maximum and minimum values among the corrected urgency scores of all significant aggregation regions.

[0119] Suppose the corrected urgency scores of all regions are 0.4608, 0.35, 0.52, etc., the minimum value is 0.3, and the maximum value is 0.55. For the corrected urgency score 0.4608 calculated above, its normalized final urgency score is equal to (0.4608 - 0.3) divided by (0.55 - 0.3), that is, 0.1608 divided by 0.25, and the calculated final urgency score is 0.6432. Write this final urgency score into the prevention and control task priority queue.

[0120] Step S1464: Sort the set of coordinates of significant aggregation regions according to the urgency scores from high to low to generate a prevention and control task priority queue.

[0121] According to the final urgency scores of each significant aggregation region, sort the set of coordinates of significant aggregation regions. Arrange the regions with higher final urgency scores in the front and those with lower scores in the back. For example, after sorting, the prevention and control task priority queue may be [(600 - 620, 700 - 720, 0 - 10), (200 - 220, 300 - 320, 0 - 10), (400 - 420, 500 - 520, 0 - 10)], indicating that the prevention and control task of the first region has the highest priority, and so on.

[0122] Step S1465: Generate the pest control priority strategy including time nodes, resource allocation plans, and execution paths according to the prevention and control task priority queue and the prevention and control resource demand list.

[0123] According to the prevention and control task priority queue and the prevention and control resource demand list corresponding to each region, formulate specific time nodes, resource allocation plans, and execution paths.

[0124] For the region with the highest priority, such as the region with coordinates (600 - 620, 700 - 720, 0 - 10), assume that the disease occurring in this region is "pine wilt disease". According to the prevention and control resource demand list, it requires nematicide, 10 sprayers, 5 drilling machines, and 20 professional prevention and control personnel. Determine to start the prevention and control work within the next 2 days, which is the time node.

[0125] In terms of the resource allocation plan, allocate the corresponding quantities of pharmaceuticals, equipment, and manpower from the existing resource inventory to this area. Assume that the current inventory of the pharmaceutical "nematocide" is 800 liters, and 500 liters are allocated to this area; there are 15 sprayers in stock, and 10 are allocated; there are 8 drilling machines in stock, and 5 are allocated; in terms of manpower, 20 prevention and control personnel are allocated from the existing prevention and control staff to this area.

[0126] The execution path refers to the travel route of the prevention and control personnel and equipment from the resource storage point to the target area. According to the geographical information and traffic conditions of the target forest area, an optimal execution path is planned. For example, starting from the resource storage point, passing through specific roads and passages, and finally reaching the target area.

[0127] For other areas with lower priorities, in the same way, according to their priority order and the list of prevention and control resource requirements, determine the time nodes, resource allocation plans, and execution paths in sequence, and finally form a complete pest control priority strategy.

[0128] Step S150: Integrate the pest distribution heat map with the pest control priority strategy to generate a forest area control optimization instruction set, and feedback the forest area control optimization instruction set to the forestry monitoring platform to trigger the prevention and control resource scheduling operation.

[0129] In this step, it is necessary to integrate the previously generated pest distribution heat map and the pest control priority strategy to generate a forest area control optimization instruction set, and then feedback this instruction set to the forestry monitoring platform.

[0130] First, perform the integration process. Integrate the pest distribution information in the pest distribution heat map with the information such as time nodes, resource allocation plans, and execution paths in the pest control priority strategy. For example, in the pest distribution heat map, mark the corresponding prevention and control time nodes, and the types and quantities of resources to be allocated for each significant aggregation area.

[0131] For the area with coordinates (600 - 620, 700 - 720, 0 - 10), mark at the corresponding position in the heat map that 500 liters of the "nematocide" pharmaceutical, 10 sprayers, 5 drilling machines, and 20 professional prevention and control personnel need to be allocated for prevention and control work within the next 2 days. Such marking and integration operations are carried out for all significant aggregation areas.

[0132] After the integration process, a forest area control optimization instruction set is generated. This instruction set contains detailed control information, such as the control time for each area, the required resources, and the execution path.

[0133] Then, the optimized forest area prevention and control instruction set is fed back to the forestry monitoring platform. After receiving the instruction set, the forestry monitoring platform triggers the prevention and control resource scheduling operation according to the information therein. The platform will allocate corresponding pesticides, equipment, and manpower from the resource inventory according to the resource allocation plan, and arrange personnel to transport these resources to the target area along the execution path. At the same time, the platform will monitor the progress of the prevention and control work in real time to ensure that the prevention and control work proceeds smoothly according to the time nodes.

[0134] For example, the platform will arrange vehicles to transport the prepared pesticides and equipment to the target area and notify the relevant prevention and control personnel to arrive at the designated location on time. During the prevention and control process, through the monitoring equipment installed in the forest area or the feedback from the prevention and control personnel, the platform can understand the progress of the prevention and control work in real time, such as whether the pesticides are sprayed in place and whether the equipment is operating normally. If problems occur, the platform can adjust the prevention and control strategy in a timely manner to ensure that the pests and diseases are effectively controlled.

[0135] Based on the above steps, starting from obtaining the dynamic monitoring data set of the target forest area, through feature extraction, pest and disease prediction, generating the heat map of pest and disease distribution and the prevention and control priority strategy, and finally fusing this information to generate the optimized forest area prevention and control instruction set and feeding it back to the forestry monitoring platform, it realizes the effective prediction of pests and diseases in the target forest area and the reasonable scheduling of prevention and control resources, improving the efficiency and effect of forestry pest and disease prevention and control.

[0136] Furthermore, for example, the method may further include the pre-training step of the pest and disease prediction model, which is specifically as follows.

[0137] Step S210: Obtain a historical monitoring sample data set, where the historical monitoring sample data set includes multiple healthy sample data subsets and pest and disease sample data subsets, and each sample data subset contains an environmental parameter sequence, a vegetation physiological parameter sequence, and a labeled pest and disease occurrence status label.

[0138] To pre-train the pest and disease prediction model, first, a historical monitoring sample data set needs to be obtained, which includes multiple healthy sample data subsets and pest and disease sample data subsets, and each sample data subset contains an environmental parameter sequence, a vegetation physiological parameter sequence, and a labeled pest and disease occurrence status label.

[0139] In this embodiment, historical monitoring data can be collected from multiple different forest areas, which cover different geographical environments, climatic conditions, and vegetation types to ensure the diversity and representativeness of the data. The environmental parameter sequence includes data such as temperature, humidity, light intensity, and soil pH, which reflect the environmental conditions of the forest area. The vegetation physiological parameter sequence contains information such as chlorophyll content, trunk water saturation, and canopy density, which are closely related to the growth state of the vegetation.

[0140] For each subset of sample data, it is also necessary to label the occurrence status of pests and diseases. This label can be determined through on-site investigations, expert evaluations, or long-term monitoring records. For example, for a subset of sample data, if pests and diseases are found in the area through on-site investigation, and according to the severity and scope of influence of the pests and diseases, it is labeled as different statuses such as "mild pests and diseases", "moderate pests and diseases", or "severe pests and diseases".

[0141] Step S220: Perform time window truncation and dimensional unit normalization on the historical monitoring sample data set to generate a standardized sample data set.

[0142] After obtaining the historical monitoring sample data set, it is necessary to perform time window truncation and dimensional unit normalization on it to generate a standardized sample data set.

[0143] Since environmental parameters and vegetation physiological parameters are time-varying sequence data, in order to better capture the characteristics of the data, time window truncation is required. Select a set time window size, for example, with one week as the time window. For each subset of sample data, divide it into multiple time windows in chronological order.

[0144] Suppose a subset of sample data contains one month of monitoring data. With one week as the time window, it can be divided into four time windows. The data within each time window contains the sequence of environmental parameters and the sequence of vegetation physiological parameters during that time period.

[0145] Different environmental parameters and vegetation physiological parameters have different dimensions and value ranges. In order to eliminate the influence of dimensions, it is necessary to normalize the data. Common normalization methods include min-max normalization and Z-score normalization.

[0146] Taking min-max normalization as an example, for each environmental parameter and vegetation physiological parameter, find its minimum and maximum values in the entire historical monitoring sample data set. Then, for each data point, subtract its minimum value and divide by the difference between the maximum value and the minimum value to obtain the normalized data.

[0147] For example, for the temperature parameter, in the historical monitoring sample data set, the minimum value is 10°C and the maximum value is 30°C. For a temperature data point of 20°C, the normalized value is (20 - 10) / (30 - 10) = 0.5.

[0148] Through time window truncation and dimensional unit normalization, a standardized sample data set is generated, making the data between different subsets of sample data comparable and facilitating subsequent model training.

[0149] Step S230: Input the standardized sample data set into the feature encoding layer of the initial pest and disease prediction model for iterative training to generate a set of high-dimensional feature vectors. The feature encoding layer extracts spatio-temporal correlation features through a parallel structure of a convolutional neural network and a long short-term memory network.

[0150] Input the standardized sample data set into the feature encoding layer of the initial pest and disease prediction model for iterative training to generate a set of high-dimensional feature vectors. The feature encoding layer extracts spatio-temporal correlation features through a parallel structure of a convolutional neural network (CNN) and a long short-term memory network (LSTM).

[0151] CNN is mainly used to extract the spatial features of data. The environmental parameter sequence and the vegetation physiological parameter sequence in the standardized sample data set are regarded as two-dimensional matrix data, where one dimension represents time and the other dimension represents different parameters.

[0152] In CNN, multiple convolutional kernels are used to perform convolutional operations on the data to extract features at different scales. Each convolutional kernel can be regarded as a small filter that slides on the data to calculate the convolution result. Through the combination of multiple convolutional layers and pooling layers, the high-level features of the data can be gradually extracted.

[0153] For example, use a 3×3 convolutional kernel to perform a convolutional operation on the data to obtain a new feature map. Then, downsample the feature map through a pooling layer (such as max pooling) to reduce the dimension of the data while retaining important feature information.

[0154] LSTM is mainly used to process sequence data and can capture the time-dependent relationship of the data. The environmental parameter sequence and the vegetation physiological parameter sequence in the standardized sample data set are input into the LSTM in chronological order.

[0155] LSTM controls the flow of information through a gating mechanism, including an input gate, a forget gate, and an output gate. The input gate determines whether new information should be added to the cell state, the forget gate determines which information should be forgotten, and the output gate determines which part of the cell state should be output.

[0156] Through the iterative training of LSTM, the time series features of the data can be learned, such as the changing trend of vegetation physiological parameters over time.

[0157] Concatenate the outputs of CNN and LSTM to obtain a high-dimensional feature vector. This high-dimensional feature vector contains the spatial and time features of the data and can describe the sample data more comprehensively.

[0158] During the iterative training process, the parameters of the CNN and LSTM are continuously adjusted so that the generated high-dimensional feature vectors can better reflect the characteristics of the sample data. Through multiple iterations, a set of high-dimensional feature vectors is finally generated.

[0159] Step S240: Screen out the high-dimensional feature vectors corresponding to the healthy sample data subset from the set of high-dimensional feature vectors, calculate the mean vector and covariance matrix for each feature dimension, and generate a standard healthy feature vector and store it in the healthy vegetation feature template library.

[0160] Screen out the high-dimensional feature vectors corresponding to the healthy sample data subset from the generated set of high-dimensional feature vectors. These healthy sample data subsets are the sample data marked as having no pests and diseases in step S210.

[0161] After screening out the high-dimensional feature vectors corresponding to the healthy sample data subset, further process them to generate a standard healthy feature vector.

[0162] For the high-dimensional feature vectors corresponding to the screened healthy sample data subset, calculate the mean vector and covariance matrix for each feature dimension.

[0163] The mean vector represents the average value of each feature dimension in the healthy sample data. For a high-dimensional feature vector, assuming there are n feature dimensions, calculate the average value of each feature dimension in all healthy sample data to obtain an n-dimensional mean vector.

[0164] The covariance matrix describes the correlation between different feature dimensions. Calculate the covariance between any two feature dimensions to obtain an n×n covariance matrix.

[0165] Based on the calculated mean vector and covariance matrix, generate a standard healthy feature vector. The standard healthy feature vector can be regarded as the representative feature of the healthy sample data.

[0166] Store the generated standard healthy feature vector in the healthy vegetation feature template library as a reference template for subsequent anomaly detection. In the actual process of pest and disease prediction, by comparing the differences between the high-dimensional feature vectors of the input data and the standard healthy feature vectors, it is determined whether there are pest and disease anomalies.

[0167] Step S250: Based on the pest and disease sample data subset in the set of high-dimensional feature vectors, calculate the multi-dimensional deviation coefficient of each high-dimensional feature vector from the standard healthy feature vector, and optimize the combined weight of the Euclidean distance and cosine similarity of the anomaly detection layer through the backpropagation algorithm.

[0168] Based on the subset of pest and disease sample data in the high-dimensional feature vector set, calculate the multi-dimensional deviation coefficient of each high-dimensional feature vector from the standard healthy feature vector. The multi-dimensional deviation coefficient is used to measure the degree of difference between the pest and disease sample data and the healthy sample data.

[0169] For each high-dimensional feature vector corresponding to the subset of pest and disease sample data, calculate its Euclidean distance and cosine similarity to the standard healthy feature vector.

[0170] The Euclidean distance measures the distance between two vectors in space. The larger the distance, the greater the difference between the two vectors. The cosine similarity measures the directional similarity between two vectors. The closer the similarity is to 1, the more similar the directions of the two vectors are.

[0171] Combine the Euclidean distance and cosine similarity with weights to obtain the multi-dimensional deviation coefficient. The weights can be optimized by the backpropagation algorithm.

[0172] The backpropagation algorithm is an optimization algorithm for training neural networks. By continuously adjusting the weights, the error between the predicted result of the model and the true label is minimized. In this process, according to the multi-dimensional deviation coefficient and the labeled pest and disease occurrence status label, calculate the error function. Then, through backpropagating the error, adjust the combined weights of the Euclidean distance and cosine similarity so that the multi-dimensional deviation coefficient can more accurately reflect the difference between the pest and disease sample data and the healthy sample data.

[0173] Step S260: According to the labeled pest and disease occurrence status label and the preset true value of the influence range level, perform supervised training on the influence range calculation module of the pest and disease prediction model to generate a chain propagation rate matching function for spatial correlation parameters.

[0174] According to the labeled pest and disease occurrence status label and the preset true value of the influence range level, perform supervised training on the influence range calculation module of the pest and disease prediction model to generate a chain propagation rate matching function for spatial correlation parameters.

[0175] Extract spatial correlation parameters from the high-dimensional feature vector set. These spatial correlation parameters may include information such as the spatial coordinates of the monitoring nodes and the distances to adjacent monitoring nodes. The spatial correlation parameters are used to describe the spread of pests and diseases in space.

[0176] Take the labeled pest and disease occurrence status label and the preset true value of the influence range level as supervision signals and input them into the influence range calculation module. The influence range calculation module adjusts its own parameters by learning these supervision signals to generate a chain propagation rate matching function for spatial correlation parameters.

[0177] The chain propagation rate matching function is used to describe the propagation rate of pests and diseases between different spatial locations. Through training, the impact range calculation module can accurately predict the impact range level of pests and diseases according to the spatial correlation parameters.

[0178] Step S270: Adjust the hyperparameters of the feature encoding layer, anomaly detection layer, and impact range calculation module through cross-validation until the model prediction accuracy reaches the convergence threshold, and output the trained pest and disease prediction model.

[0179] Adjust the hyperparameters of the feature encoding layer, anomaly detection layer, and impact range calculation module through cross-validation until the model prediction accuracy reaches the convergence threshold, and output the trained pest and disease prediction model.

[0180] For example, divide the standardized sample data set into a training set and a validation set. For example, divide the data into a training set and a validation set according to the ratio of 80% and 20%.

[0181] Use the training set to train the model and use the validation set to evaluate the performance of the model. Through multiple cross-validations, continuously adjust the hyperparameters of the feature encoding layer, anomaly detection layer, and impact range calculation module, such as the learning rate, convolution kernel size, number of hidden layer neurons, etc.

[0182] In each cross-validation, calculate the prediction accuracy of the model. The prediction accuracy can be measured by comparing the difference between the prediction result of the model and the true label, such as using metrics like accuracy, recall rate, F1 value, etc.

[0183] Continuously adjust the hyperparameters until the prediction accuracy of the model reaches the convergence threshold. The convergence threshold is a pre-set accuracy value. When the prediction accuracy of the model reaches or exceeds this threshold, the model training is considered to converge.

[0184] When the model training converges, output the trained pest and disease prediction model. This model can be used for actual forest pest and disease prediction. By inputting new monitoring data, it outputs the occurrence probability of pests and diseases and the impact range level, providing decision-making support for forest pest and disease control.

[0185] During the actual pre-training process, attention should be paid to the dimensional consistency of the feature comparison and parameter calculation processes and the matching of feature dimensions. For example, when calculating the Euclidean distance and cosine similarity, ensure that the dimensions of the high-dimensional feature vector and the standard healthy feature vector are consistent. At the same time, when adjusting the hyperparameters, make reasonable selections according to the performance of the model and the characteristics of the data to improve the prediction accuracy and generalization ability of the model.

[0186] Figure 2FIG. shows a schematic diagram of exemplary hardware and software components of an artificial intelligence-based forest pest prediction system 100 provided by some embodiments of the present application, which can implement the idea of the present application. For example, the processor 120 can be used on the artificial intelligence-based forest pest prediction system 100 and is used to execute the functions in the present application.

[0187] The artificial intelligence-based forest pest prediction system 100 can be a general-purpose server or a special-purpose server, both of which can be used to implement the artificial intelligence-based forest pest prediction method of the present application. Although only one server is shown in the present application, for convenience, the functions described in the present application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0188] For example, the artificial intelligence-based forest pest prediction system 100 can include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as disks, ROM, or RAM, or any combination thereof. Exemplarily, the artificial intelligence-based forest pest prediction system 100 can also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application can be implemented according to these program instructions. The artificial intelligence-based forest pest prediction system 100 also includes an I / O interface 150 between the computer and other input / output devices.

[0189] For ease of explanation, only one processor is described in the artificial intelligence-based forest pest prediction system 100. However, it should be noted that the artificial intelligence-based forest pest prediction system 100 in the present application can also include multiple processors. Therefore, the steps performed by one processor described in the present application can also be jointly executed or separately executed by multiple processors. For example, if the processor of the artificial intelligence-based forest pest prediction system 100 executes step A and step B, it should be understood that step A and step B can also be jointly executed by two different processors or separately executed in one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor jointly execute steps A and B.

[0190] In addition, an embodiment of the present invention also provides a readable storage medium, in which computer-executable instructions are preset. When the processor executes the computer-executable instructions, the above-mentioned artificial intelligence-based forest pest prediction method is implemented.

[0191] It should be noted that, in order to simplify the description of the present invention disclosure and thus assist in the understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, sometimes multiple features are incorporated into one embodiment, drawing or description thereof.

Claims

1. An artificial intelligence-based forest pest and disease prediction method, characterized in that The method includes: Obtaining a dynamic monitoring data set of a target forest area, where the dynamic monitoring data set includes environmental parameter sequences and vegetation physiological parameter sequences collected by multiple monitoring nodes, and each monitoring node corresponds to a spatial coordinate area; Performing feature extraction on the dynamic monitoring data set to obtain a forestry environment feature set for each monitoring node, where the forestry environment feature set includes vegetation growth state features, environmental fluctuation correlation features, and potential pest and disease identification features; Based on a pre-trained pest and disease prediction model, performing abnormal parameter analysis processing on the forestry environment feature set to generate a pest and disease prediction parameter set for the monitoring node, where the pest and disease prediction parameter set is used to characterize the occurrence probability and influence range level of pests and diseases; Performing a dynamic prediction matching operation according to the pest and disease prediction parameter set to determine a pest and disease distribution heat map of the target forest area and generate a pest and disease control priority strategy; Performing fusion processing on the pest and disease distribution heat map and the pest and disease control priority strategy to generate a forest area control optimization instruction set, and feeding back the forest area control optimization instruction set to a forestry monitoring platform to trigger a control resource scheduling operation.

2. The method for predicting forestry pests and diseases based on artificial intelligence according to claim 1, wherein The performing feature extraction on the dynamic monitoring data set to obtain a forestry environment feature set for each monitoring node includes: Aligning the environmental parameter sequences in the time dimension to generate a standardized environmental parameter sequence; Invoking a preset vegetation physiological feature encoder to perform physiological state analysis processing on the vegetation physiological parameter sequence to generate a vegetation growth state feature vector, where the vegetation growth state feature vector includes a chlorophyll content change rate, a trunk water saturation degree, and a canopy density attenuation coefficient; Performing environmental correlation analysis processing on the standardized environmental parameter sequence to extract environmental fluctuation correlation features, where the environmental fluctuation correlation features include a temperature-humidity co-variation gradient, a soil pH offset, and a cumulative light intensity deviation value; Performing abnormal identification fusion processing on the vegetation growth state feature vector and the environmental fluctuation correlation features to generate potential pest and disease identification features, where the potential pest and disease identification features include an estimated egg distribution density, a fungal infection diffusion rate, and a disease symptom appearance intensity; Integrating the vegetation growth state features, the environmental fluctuation correlation features, and the potential pest and disease identification features to form the forestry environment feature set.

3. The method for predicting forestry pests and diseases based on artificial intelligence according to claim 2, wherein, The performing environmental correlation analysis processing on the standardized environmental parameter sequence to extract environmental fluctuation correlation features includes: Obtaining a temperature sequence, a humidity sequence, and a light intensity sequence in the standardized environmental parameter sequence; Calculating a dynamic covariance matrix between the temperature sequence and the humidity sequence to determine a temperature-humidity co-variation gradient, where the temperature-humidity co-variation gradient is quantified by a covariance change rate within a sliding time window; Performing trend fitting processing on the soil pH data in the standardized environmental parameter sequence to extract a soil pH offset, where the soil pH offset is determined by a standard deviation multiple of the current pH value and a historical reference value; Calculate the cumulative deviation of the light intensity sequence to generate a light intensity cumulative deviation value, where the cumulative deviation value is the cumulative difference between the measured value and the theoretical value of the light intensity within a continuous time interval; Normalize the temperature-humidity co-variation gradient, soil pH offset, and light intensity cumulative deviation value to obtain the environmental fluctuation correlation feature.

4. The method for predicting forestry pests and diseases based on artificial intelligence according to claim 2, characterized in that, The abnormal identification fusion process of the vegetation growth state feature vector and the environmental fluctuation correlation feature to generate a potential pest and disease identification feature includes: Perform discretized segmentation processing on the vegetation growth state feature vector to generate an interval of chlorophyll content change rate, an interval of trunk water saturation, and an interval of canopy density attenuation coefficient; Determine the suitability level for egg hatching based on the temperature-humidity co-variation gradient in the environmental fluctuation correlation feature, and calculate the estimated value of egg distribution density in combination with the interval of chlorophyll content change rate; Based on the soil pH offset and the light intensity cumulative deviation value, construct a fungal growth probability model and output the fungal infection and diffusion rate, which is calculated by the linear combination coefficient of the pH offset and the light deviation; Determine the intensity of disease symptom manifestation according to the cross-validation result of the interval of trunk water saturation and the interval of canopy density attenuation coefficient, where the intensity of disease symptom manifestation is inversely proportional to the product of water saturation and canopy density; Perform weighted splicing on the estimated value of egg distribution density, the fungal infection and diffusion rate, and the intensity of disease symptom manifestation to generate the potential pest and disease identification feature.

5. The method for predicting forestry pests and diseases based on artificial intelligence according to claim 1, wherein, Based on the pre-trained pest and disease prediction model, perform abnormal parameter analysis on the forestry environment feature set to generate a pest and disease prediction parameter set for the monitoring node, including: Input the forestry environment feature set into the feature encoding layer of the pest and disease prediction model to generate a high-dimensional fusion feature vector; Through the abnormal detection layer of the pest and disease prediction model, calculate the deviation coefficient between the high-dimensional fusion feature vector and the pre-stored healthy vegetation feature template, where the deviation coefficient is determined by the weighted sum of the Euclidean distance and the cosine similarity; Match the preset pest and disease type mapping table according to the deviation coefficient to determine the pest and disease occurrence probability, where the pest and disease occurrence probability has an exponential relationship with the deviation coefficient; Call the impact range calculation module of the pest and disease prediction model, and based on the spatial correlation parameters in the high-dimensional fusion feature vector, calculate the pest and disease impact range level, where the pest and disease impact range level is determined by the chain propagation rate of the feature similarity of adjacent monitoring nodes; Combine and encapsulate the pest and disease occurrence probability and the impact range level to generate the pest and disease prediction parameter set.

6. The method for predicting forestry pests and diseases based on artificial intelligence according to claim 5, wherein, The calculation of the deviation coefficient between the high-dimensional fusion feature vector and the pre-stored healthy vegetation feature template includes: Obtain the standard chlorophyll content, standard trunk water saturation, and standard canopy density in the healthy vegetation feature template; Calculate the absolute value of the first difference between the chlorophyll content change rate in the high-dimensional fusion feature vector and the standard chlorophyll content, and the absolute value of the second difference between the trunk water saturation and the standard trunk water saturation; Normalize the absolute value of the first difference and the absolute value of the second difference to generate a primary deviation degree; Extract the canopy density attenuation coefficient from the high-dimensional fusion feature vector, calculate the reciprocal of its ratio to the standard canopy density, and generate a secondary deviation degree index; Linearly combine the primary deviation degree index and the secondary deviation degree index to obtain a comprehensive deviation degree coefficient, where the combination weights of the linear combination are dynamically adjusted according to the feature importance in the historical pest and disease data.

7. The method for predicting forestry pests and diseases based on artificial intelligence according to claim 1, characterized in that The dynamic prediction matching operation performed according to the pest and disease prediction parameter set to determine the pest and disease distribution heat map of the target forest area includes: Extract the pest and disease occurrence probability and the influence range level from the pest and disease prediction parameter set; Construct a three-dimensional geographic information grid according to the spatial coordinate area of the monitoring node, and map the pest and disease occurrence probability of each grid node to the corresponding spatial coordinate; Perform spatial interpolation processing on the discrete pest and disease occurrence probability based on the Kriging interpolation algorithm to generate a continuous probability distribution surface; Perform regional segmentation on the continuous probability distribution surface according to the influence range level, divide it into multiple partitions, and assign a color gradient identifier to each partition; Overlay the color gradient identifier with the three-dimensional geographic information grid to generate the pest and disease distribution heat map, where the heat intensity is proportional to the product of the pest and disease occurrence probability and the influence range level.

8. The method for predicting forestry pests and diseases based on artificial intelligence according to claim 7, characterized in that, The performing spatial interpolation processing on the discrete pest and disease occurrence probability based on the Kriging interpolation algorithm to generate a continuous probability distribution surface includes: Generate a spatial distance matrix between the monitoring nodes based on the spatial coordinates of the monitoring nodes; Calculate the semi-variogram model parameters according to the spatial distance matrix and construct a semi-variogram model; Calculate the interpolation weight coefficient between the target grid point and the monitoring node according to the semi-variogram model parameters and the spatial distance matrix; Perform weighted summation on the pest and disease occurrence probability of the monitoring nodes according to the interpolation weight coefficient to generate the interpolation probability value of the target grid point; Traverse all target grid points in the three-dimensional geographic information grid to generate a continuous probability distribution surface covering the target forest area; Perform Gaussian filtering processing on the continuous probability distribution surface to eliminate the local mutation noise of the interpolation probability value and output the smoothed continuous probability distribution surface.

9. The method for predicting forestry pests and diseases based on artificial intelligence according to claim 1, characterized in that, The generating the pest and disease control priority strategy includes: Obtain the coordinate set of the significant aggregation areas in the pest and disease distribution heat map and the corresponding pest and disease type identifiers; Match the control resource demand list corresponding to each pest and disease type according to the historical control record database, and the control resource demand list includes the pesticide type, the number of equipment and the manpower allocation; Calculate the emergency score of each area in the coordinate set of the significant aggregation areas, and the emergency score is determined by the product of the pest and disease occurrence probability, the influence range level and the vegetation economic value; Sort the coordinate set of the significant aggregation areas from high to low according to the emergency score to generate a control task priority queue; Generate the pest and disease control priority strategy including time nodes, resource allocation plans and execution paths according to the control task priority queue and the control resource demand list; Among them, calculating the urgency score of each region in the set of coordinates of the significant aggregation region includes: Obtaining the pest occurrence probability P and the influence range level R in the set of pest prediction parameters; Querying the vegetation economic value database of the target forest area and extracting the economic value coefficient V of the corresponding region, where the economic value coefficient is comprehensively calculated from the vegetation type, tree age, and market unit price; Calculating the initial urgency score through the formula S = P × R × V; If the current inventory of prevention and control resources is lower than the demand threshold, then increase the initial urgency score by a preset ratio to obtain the corrected urgency score; Normalize the corrected urgency score to obtain the final urgency score and write it into the prevention and control task priority queue.

10. An artificial intelligence-based forestry pest and disease prediction system, characterized in that, It includes a processor and a memory. The memory is connected to the processor. The memory is used to store programs, instructions, or codes, and the processor is used to execute the programs, instructions, or codes in the memory to implement the artificial intelligence-based forest pest prediction method described in any one of claims 1-9 above.

Images