Pest and disease probability prediction method and system based on multi-time sequence feature fusion

Abstract

The invention relates to the technical field of agricultural information, in particular to a pest and disease damage probability prediction method and system based on multi-time sequence feature fusion, and the method comprises the steps: collecting multi-source time sequence data in a preset time, and aligning the time granularity of different time sequence data through a time granularity alignment method; time sequence correlation features related to diseases and insect pests are mined to generate a fusion feature vector, the fusion feature vector is input into a prediction model, the model predicts the occurrence probability of the diseases and insect pests within expected time in combination with the weight of each dimension in the vector, and the judgment coefficient and the trend consistency proportion of the occurrence probability of the diseases and insect pests within the expected time are determined; and judging whether the prediction probability of the pest and disease damage meets the standard or not based on the judgment coefficient and the trend consistency proportion, and determining the weight coefficient of the core time sequence feature, the attention weight attenuation coefficient, the time step length, the movement interval and the time granularity. Accurate prediction of the pest and disease damage probability is effectively realized, and the prediction efficiency of the pest and disease damage probability is effectively improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural information technology, and in particular to a method and system for predicting the probability of pests and diseases based on the fusion of multiple temporal features. Background Technology

[0002] Multi-temporal feature fusion refers to the integration, correlation, and optimization of temporal features from multiple sources, multiple time dimensions, and various types to generate more representative fused features for subsequent prediction and analysis tasks. Temporal features refer to monitoring data that changes over time. Pests and diseases are a general term for harmful organisms that damage plants in agriculture and other fields. Pest and disease probability prediction refers to the technique of predicting the probability of a certain type of pest or disease occurring in a specific region and a specific crop within a certain time period based on data and models, providing a basis for precise pest and disease control. Pest and disease probability prediction based on multi-temporal feature fusion is a combined application of multi-temporal feature fusion technology and pest and disease probability prediction, which can achieve accurate and efficient pest and disease early warning. Traditional methods are mostly based on statistical analysis of a single factor, which has the limitation of ignoring the dynamic correlation of multiple factors. Therefore, research on pest and disease probability prediction based on multi-temporal feature fusion has important practical significance.

[0003] Chinese Patent Publication No. CN120832495A discloses an artificial intelligence-based method and system for predicting pests and diseases in salt-tolerant rice. The method includes rice data collection, rice data optimization, establishing a salt-tolerant rice pest and disease prediction model, and predicting pests and diseases in salt-tolerant rice. This invention belongs to the field of agricultural information technology, specifically referring to an artificial intelligence-based method and system for predicting pests and diseases in salt-tolerant rice. This solution dynamically adjusts the rhomboid plot segmentation based on the irregular terrain and salt-alkali accumulation zone boundaries of salt-tolerant rice fields. It calculates collaborative weights based on the degree of salt-alkali stress and the risk level of pests and diseases, and strengthens key area learning by combining the total plot score and selection probability formula. It introduces plot scores, the average density of neighboring pests and diseases, and the neighboring salt-alkali gradient into the spatial branch to improve the ability to capture spatial correlations. Based on dynamically adjusting the tolerance threshold, it limits the influence of extreme values ​​through a loss upper limit coefficient, thereby constructing a pest and disease prediction error control function to improve the accuracy of pest and disease prediction.

[0004] Therefore, the above scheme dynamically adjusts the diamond-shaped plot division, reduces interference from invalid pest and disease data, selects a probability formula, strengthens key area learning, generates intermediate state samples, improves the ability to capture spatial correlations, and constructs a pest and disease prediction error control function, thereby improving the reliability and accuracy of pest and disease prediction. However, the above scheme cannot achieve precise prediction of pest and disease probabilities, thus failing to guarantee the efficiency of pest and disease probability prediction. Summary of the Invention

[0005] To address this issue, the present invention provides a method and system for predicting the probability of pests and diseases based on the fusion of multiple temporal features, thereby overcoming the problem that existing technologies cannot achieve accurate prediction of the probability of pests and diseases, resulting in low prediction efficiency.

[0006] On the one hand, this invention provides a method for predicting the probability of pests and diseases based on multi-temporal feature fusion, including: Collect multi-source time-series data within a preset time period, including environmental time-series data, crop growth time-series data, historical time-series data of pests and diseases, and time-series data of agricultural operations; The temporal granularity of different time series data is aligned using a time granularity alignment method, and temporal correlation features related to pests and diseases are mined to generate a fused feature vector; The fused feature vector is input into the prediction model, and the model combines the weights of each dimension in the vector to predict the probability of pests and diseases occurring within the expected time. Determine the determination coefficient and trend consistency ratio of the probability of pest and disease occurrence within the expected time period; The predicted probability of pests and diseases is determined based on the determination coefficient to see if it meets the standard. Based on the determination coefficient, determine whether the predicted probability of pests and diseases based on the trend consistency ratio meets the standard. Alternatively, generate corresponding processing instructions; Based on the trend consistency ratio determination results, the probability prediction of pests and diseases is completed. Alternatively, generate corresponding processing instructions; Adjust the weight coefficient of the core time series feature based on the determination coefficient, adjust the attention weight decay coefficient based on the trend consistency ratio, increase the time step based on the increase of the weight coefficient of the core time series feature, decrease the movement interval based on the increase of the time step, refine the time granularity based on the adjustment level of the attention weight decay coefficient, and issue a technical optimization failure notification, or issue a consistency ratio optimization failure notification.

[0007] Furthermore, the process of determining whether the predicted probability of pests and diseases meets the standard based on the determination coefficient includes: Determine the linear correlation between the predicted probability and the actual probability, and record the obtained linear correlation as the determination coefficient; When the determination coefficient is greater than or equal to the preset determination coefficient, the predicted probability of pests and diseases is determined based on the trend consistency ratio to determine whether it meets the standard. When the determination coefficient is less than the preset determination coefficient, the predicted probability of pests and diseases is determined to be inconsistent with the standard, and the weight coefficient of the core time series feature is adjusted based on the determination coefficient.

[0008] Furthermore, the process of determining whether the predicted probability of pests and diseases meets the standard based on the trend consistency ratio includes: The number of days in which the predicted probability rises and falls in line with the actual probability rises and falls is recorded as the number of days in which the trend is consistent. Calculate the ratio of the number of days with consistent trend to the total number of days predicted within the expected time period, and record the obtained ratio as the trend consistency ratio; When the trend consistency ratio is greater than or equal to the preset trend consistency ratio, it is determined that the predicted probability of pests and diseases meets the standard, the probability prediction of pests and diseases is completed, and the probability prediction of pests and diseases with prediction needs is performed. When the trend consistency ratio is less than the preset trend consistency ratio, it is determined that the predicted probability of pests and diseases does not meet the standard, and the attention weight decay coefficient is adjusted based on the trend consistency ratio.

[0009] Furthermore, the process of increasing the weight coefficient of the core time-series feature based on the determination coefficient includes: Calculate the difference between the preset decision coefficient and the decision coefficient, and record the obtained difference as the decision coefficient difference; The weight coefficient of the core time series feature is increased based on the difference in the determination coefficients, and the increase in the weight coefficient of the core time series feature is proportional to the difference in the determination coefficients.

[0010] Furthermore, after the weight coefficients of the core time series features are increased, the process of determining the time step based on the increase in the weight coefficients of the core time series features includes: The time step is determined by the increase in the weight coefficients of the core time series features, and the increase in the time step is proportional to the increase in the weight coefficients of the core time series features. After the time step is increased, the moving interval is reduced based on the increase in the time step, and the reduction in the moving interval is proportional to the increase in the time step.

[0011] Furthermore, the process of determining whether the predicted probability of pests and diseases meets the standard based on the determination coefficient after the movement interval has been reduced includes: Determine the linear correlation between the predicted probability and the actual probability, and record the obtained linear correlation as the determination coefficient; When the determination coefficient is greater than or equal to the preset determination coefficient, the predicted probability of pests and diseases is determined based on the trend consistency ratio to determine whether it meets the standard. When the determination coefficient is less than the preset determination coefficient, the predicted probability of pests and diseases is determined to be non-compliant with the standard, and a technical optimization failure notification is issued.

[0012] Furthermore, the process of adjusting the attention weight decay coefficient based on the trend consistency ratio includes: Calculate the difference between the preset trend consistency ratio and the trend consistency ratio, and record the obtained difference as the consistency difference. The attention weight decay coefficient is adjusted based on the consistency difference, and the adjustment range of the attention weight decay coefficient is proportional to the consistency difference.

[0013] Furthermore, the process of refining the time granularity based on the adjustment level of the attention weight decay coefficient after the attention weight decay coefficient adjustment is completed includes: The time granularity is refined based on the adjustment level of the attention weight decay coefficient, and the degree of refinement of the time granularity is proportional to the adjustment level of the attention weight decay coefficient.

[0014] Furthermore, the process of determining whether the predicted probability of pests and diseases meets the standard based on the trend consistency ratio after the time granularity refinement includes: When the trend consistency ratio is greater than or equal to the preset trend consistency ratio, it is determined that the predicted probability of pests and diseases meets the standard, the probability prediction of pests and diseases is completed, and the probability prediction of pests and diseases with prediction needs is performed. When the trend consistency ratio is less than the preset trend consistency ratio, it is determined that the predicted probability of pests and diseases does not meet the standard, and a consistency ratio optimization failure notification is issued.

[0015] On the other hand, the present invention also provides a pest and disease probability prediction system based on multi-temporal feature fusion using the above method, comprising: The data input layer is used to collect the multi-source time-series data within the preset time period, including the environmental time-series data, the crop growth time-series data, the historical time-series data of pests and diseases, and the time-series data of agricultural operations. A feature fusion layer, which is connected to the data input layer, is used to align the time granularity of different time series data through a time granularity alignment method, and to mine the time series correlation features related to pests and diseases to generate the fused feature vector. The prediction output layer, which is connected to the feature fusion layer, is used to input the fused feature vector into the prediction model. The model combines the weights of each dimension in the vector to predict the probability of the occurrence of pests and diseases within the expected time. The intelligent analysis layer, which is connected to the prediction output layer, is used to determine the determination coefficient and the trend consistency ratio of the probability of pests and diseases occurring within the expected time period; the intelligent analysis layer is also used to determine whether the predicted probability of pests and diseases meets the standard based on the determination coefficient. The intelligent analysis layer is also used to determine whether the predicted probability of pests and diseases based on the trend consistency ratio meets the standard, or to generate corresponding processing instructions, based on the determination result of the determination coefficient. The intelligent analysis layer is also used to determine the probability prediction of pests and diseases based on the trend consistency ratio judgment result, or to generate corresponding processing instructions. A parameter adjustment layer, which is connected to the data input layer, the feature fusion layer, the prediction output layer, and the intelligent analysis layer, is used to adjust the weight coefficient of the core time-series feature based on the determination coefficient, adjust the attention weight decay coefficient based on the trend consistency ratio, increase the time step based on the increase in the weight coefficient of the core time-series feature, decrease the movement interval based on the increase in the time step, refine the time granularity based on the adjustment level of the attention weight decay coefficient, issue a technical optimization failure notification, and issue a consistency ratio optimization failure notification.

[0016] Compared with the prior art, the beneficial effects of the present invention are that it determines whether the predicted probability of pests and diseases meets the standard based on the determination coefficient and the trend consistency ratio, and completes the determination of whether the predicted probability of pests and diseases meets the standard in a timely and accurate manner. When it does not meet the standard, the cause is identified and a corresponding processing instruction is generated. Based on the generated processing instruction, the weight coefficient, attention weight decay coefficient, time step, movement interval and time granularity of the core time series features are adjusted, and a corresponding notification is issued under the appropriate circumstances. While effectively realizing the accurate prediction of pests and diseases probability, the prediction efficiency of pests and diseases probability is also effectively improved.

[0017] Furthermore, this invention determines whether the predicted probability of pests and diseases meets the standard based on the decision coefficient, accurately determines whether the predicted probability of pests and diseases meets the standard based on the trend consistency ratio, or whether the weight coefficient of the core time series features needs to be adjusted based on the decision coefficient, thus avoiding misjudgment. While further realizing the accurate prediction of pests and diseases probability, it also further improves the prediction efficiency of pests and diseases probability.

[0018] Furthermore, this invention determines whether the predicted probability of pests and diseases meets the standard based on the trend consistency ratio, accurately determines whether the predicted probability meets the standard, and promptly determines whether the attention weight attenuation coefficient needs to be adjusted based on the trend consistency ratio to maintain the accuracy of the determination. While further realizing the accurate prediction of pests and diseases probability, it also further improves the prediction efficiency of pests and diseases probability.

[0019] Furthermore, this invention increases the weight coefficient of core time-series features based on the determination coefficient, effectively avoiding the situation where the predicted probability of pests and diseases does not meet the standard due to the weight coefficient of core time-series features not meeting the standard. This not only further achieves accurate prediction of pests and diseases probability, but also further improves the prediction efficiency of pests and diseases probability.

[0020] Furthermore, after the weight coefficients of the core time series features are increased, the present invention effectively avoids the situation where the predicted probability of pests and diseases does not meet the standard due to the time step not meeting the standard. After the time step is increased, the moving interval is reduced based on the increase of the time step, and the time moving distance between two adjacent segments is reasonably adjusted. This not only further realizes the accurate prediction of pests and diseases probability, but also further improves the prediction efficiency of pests and diseases probability.

[0021] Furthermore, after the movement interval is reduced, the present invention determines whether the predicted probability of pests and diseases meets the standard based on the determination coefficient, accurately determines whether the predicted probability of pests and diseases meets the standard based on the trend consistency ratio or whether a technical optimization failure notification needs to be issued, ensuring the system's determination accuracy. While further realizing the accurate prediction of pests and diseases probability, the present invention further improves the prediction efficiency of pests and diseases probability.

[0022] Furthermore, this invention adjusts the attention weight decay coefficient based on the trend consistency ratio, effectively avoiding the situation where the predicted probability of pests and diseases does not meet the standard due to the attention weight decay coefficient not meeting the standard. While further realizing the accurate prediction of pests and diseases probability, it also further improves the prediction efficiency of pests and diseases probability.

[0023] Furthermore, after the attention weight decay coefficient is adjusted, the present invention refines the time granularity based on the adjustment level of the attention weight decay coefficient, which can effectively improve the effectiveness and accuracy of temporal correlation features and effectively avoid the situation where the predicted probability of pests and diseases does not meet the standard due to the time granularity not meeting the standard. While further realizing the accurate prediction of pests and diseases probability, it further improves the prediction efficiency of pests and diseases probability.

[0024] Furthermore, after the time granularity is refined, the present invention determines whether the predicted probability of pests and diseases meets the standard based on the trend consistency ratio. It can promptly and accurately determine whether the predicted probability of pests and diseases meets the standard or whether a consistency ratio optimization failure notification needs to be issued, thus avoiding misjudgment and ensuring the system's judgment accuracy. While further realizing the accurate prediction of pests and diseases probability, it also further improves the prediction efficiency of pests and diseases probability.

[0025] Furthermore, by setting up a data input layer, a feature fusion layer, a prediction output layer, an intelligent analysis layer, and a parameter adjustment layer, which are interconnected into a system, this invention effectively ensures the efficiency of the pest and disease probability prediction method based on multi-temporal feature fusion. While further realizing the accurate prediction of pest and disease probability, it also further improves the prediction efficiency of pest and disease probability. Attached Figure Description

[0026] Figure 1 This is a structural block diagram of the pest and disease probability prediction system based on multi-temporal feature fusion according to an embodiment of the present invention; Figure 2 This is a flowchart of a pest and disease probability prediction method based on multi-temporal feature fusion according to an embodiment of the present invention; Figure 3 This is a flowchart illustrating how an embodiment of the present invention determines whether the predicted probability of pests and diseases meets the standard and the reasons why it does not meet the standard; Figure 4 This is a flowchart illustrating the reasons why the predicted probability of pests and diseases does not meet the standard in an embodiment of the present invention. Detailed Implementation

[0027] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0028] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0029] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0030] Please see Figure 1 The diagram shown is a structural block diagram of a pest and disease probability prediction system based on multi-temporal feature fusion according to an embodiment of the present invention. The structure of this embodiment includes a data input layer, a feature fusion layer, a prediction output layer, an intelligent analysis layer, and a parameter adjustment layer; wherein, The data input layer is used to collect multi-source time-series data within a preset time period, including environmental time-series data, crop growth time-series data, historical time-series data of pests and diseases, and time-series data of agricultural operations. The feature fusion layer is connected to the data input layer to align the time granularity of different time series data through a time granularity alignment method, and to mine time series correlation features related to pests and diseases to generate a fused feature vector. The prediction output layer is connected to the feature fusion layer to input the fused feature vector into the prediction model. The model combines the weights of each dimension in the vector to predict the probability of pests and diseases occurring within the expected time. The intelligent analysis layer is connected to the prediction output layer and is used to determine the determination coefficient and trend consistency ratio of the probability of pest and disease occurrence within the expected time period. The intelligent analysis layer is also used to determine whether the predicted probability of pests and diseases meets the standard based on the determination coefficient. The intelligent analysis layer is also used to determine, based on the determination coefficient, whether the predicted probability of pests and diseases based on the trend consistency ratio meets the standard. Alternatively, generate corresponding processing instructions; The intelligent analysis layer is also used to determine the probability prediction of pests and diseases based on the trend consistency ratio judgment result. Alternatively, generate corresponding processing instructions; The parameter adjustment layer is connected to the data input layer, the feature fusion layer, the prediction output layer, and the intelligent analysis layer, respectively. It is used to adjust the weight coefficient of the core time series feature based on the determination coefficient, adjust the attention weight decay coefficient based on the trend consistency ratio, increase the time step based on the increase of the weight coefficient of the core time series feature, decrease the movement interval based on the increase of the time step, refine the time granularity based on the adjustment level of the attention weight decay coefficient, issue a technical optimization failure notification, and issue a consistency ratio optimization failure notification.

[0031] Specifically, the prediction time is several days prior to the date on which the probability of pests and diseases is predicted; The environmental time-series data includes temperature, humidity, rainfall, and sunshine duration; The crop growth time series data includes growth height and crop chlorophyll content; The historical time-series data on pests and diseases includes the probability of occurrence and the degree of danger of pests and diseases in the same period of several historical years. The agricultural operation timeline data includes fertilization time, fertilization frequency, irrigation time, and irrigation frequency; The purpose of aligning time granularity is to unify data from different time intervals into the same time unit; The expected time is several days after the day the probability of pests and diseases is predicted. The predicted time and the expected time are determined by the types of pests and diseases for which prediction is needed; The prediction model described in this embodiment uses the multi-source time series data from 2024, 2023, 2022, 2021, and 2020 as model training data, and uses the multi-source time series data from 2025 for backtracking verification.

[0032] Please see Figure 2 The diagram shown is a flowchart of a pest and disease probability prediction method based on multi-temporal feature fusion according to an embodiment of the present invention. The method described in this embodiment includes: Collect the multi-source time-series data within the preset time period, including the environmental time-series data, the crop growth time-series data, the historical time-series data of pests and diseases, and the time-series data of agricultural operations; The time granularity of different time series data is aligned using a time granularity alignment method, and the time series correlation features related to pests and diseases are mined to generate the fused feature vector; The fused feature vector is input into the prediction model, and the model combines the weights of each dimension in the vector to predict the probability of pests and diseases occurring within the expected time. The determination coefficient and the trend consistency ratio are used to determine the probability of pest and disease occurrence within the expected time period. The predicted probability of pests and diseases is determined based on the determination coefficient to see if it meets the standard. Based on the determination coefficient, determine whether the predicted probability of pests and diseases based on the trend consistency ratio meets the standard. Alternatively, generate corresponding processing instructions; Based on the trend consistency ratio determination results, the probability prediction of pests and diseases is completed. Alternatively, generate corresponding processing instructions; Adjust the weight coefficient of the core time series feature based on the determination coefficient, adjust the attention weight decay coefficient based on the trend consistency ratio, increase the time step based on the increase of the weight coefficient of the core time series feature, decrease the movement interval based on the increase of the time step, refine the time granularity based on the adjustment level of the attention weight decay coefficient, and issue a technical optimization failure notification, or issue a consistency ratio optimization failure notification.

[0033] Please see Figure 3 The diagram shows a flowchart illustrating an embodiment of the present invention for determining whether the predicted probability of pests and diseases meets the standard and the reasons why it does not. The process of determining whether the predicted probability of pests and diseases meets the standard based on the determination coefficient in this embodiment includes: Determine the linear correlation between the predicted probability and the actual probability, and record the obtained linear correlation as the determination coefficient; When the determination coefficient is greater than or equal to the preset determination coefficient Y, the predicted probability of pests and diseases is determined based on the trend consistency ratio to determine whether it meets the standard. In this embodiment, the preset determination coefficient Y = 0.8. When the determination coefficient is less than the preset determination coefficient Y, the predicted probability of pests and diseases is determined to be non-compliant with the standard, and the weight coefficient of the core time series feature is adjusted based on the determination coefficient. Specifically, the preset judgment coefficient is set to 0.8, which is an empirical threshold, and the core time series features are determined by the types of pests and diseases that require prediction.

[0034] Please continue reading. Figure 3 As shown, the process of determining whether the predicted probability of pests and diseases meets the standard based on the trend consistency ratio in this embodiment of the invention includes: The number of days in which the predicted probability rises and falls in line with the actual probability rises and falls is recorded as the number of days in which the trend is consistent. Calculate the ratio of the number of days with consistent trend to the total number of days predicted within the expected time period, and record the obtained ratio as the trend consistency ratio; When the trend consistency ratio is greater than or equal to the preset trend consistency ratio B, it is determined that the predicted probability of pests and diseases meets the standard, the probability prediction of pests and diseases is completed, and the probability prediction of pests and diseases with prediction needs is performed. In this embodiment, the preset trend consistency ratio B = 90%; When the trend consistency ratio is less than the preset trend consistency ratio B, it is determined that the predicted probability of pests and diseases does not meet the standard, and the attention weight attenuation coefficient is adjusted based on the trend consistency ratio. Specifically, the preset trend consistency ratio is set to 90%, which is an empirical threshold.

[0035] Please continue reading. Figure 3 As shown, the process of increasing the weight coefficient of the core time-series feature based on the determination coefficient in this embodiment of the invention includes: Calculate the difference between the preset decision coefficient and the decision coefficient, and record the obtained difference as the decision coefficient difference; The weight coefficient of the core time series feature is increased based on the difference in the determination coefficient; When the difference in the determination coefficients is greater than the second preset difference in the determination coefficients At that time, the weight coefficient of the core time series feature is increased to 2.17 times the weight coefficient of the initial core time series feature, wherein, in this embodiment, the second preset judgment coefficient difference... ; When the difference in the determination coefficients is less than or equal to the second preset difference in the determination coefficients And greater than the difference of the first preset judgment coefficient At that time, the weight coefficient of the core time series feature is increased to 1.54 times the weight coefficient of the initial core time series feature, wherein, in this embodiment, the first preset judgment coefficient difference... ; When the difference in the determination coefficients is less than or equal to the first preset difference in the determination coefficients At that time, the weight coefficient of the core time series feature is increased to 1.21 times the weight coefficient of the initial core time series feature; Specifically, the values ​​of the difference in the determination coefficients and the weight coefficients of the core time series features are both derived from the actual debugging results; Increasing the weighting coefficients of the core time-series features can more closely link the predicted probability with the actual occurrence of pests and diseases; After the weight coefficients of the core time series features are increased, the following steps need to be performed: The difference between the weight coefficients of the core time-series features after the addition and the corresponding initial weight coefficients is calculated sequentially, and the sum of each difference is recorded as the total increment of the core features. Other features besides the core temporal features are denoted as the remaining features; Determine the initial weight coefficients of the remaining features respectively, and calculate the sum of the initial weight coefficients of the remaining features, which is denoted as the sum of the initial weights of the remaining features; The ratio of the initial weight coefficient of each of the remaining features to the sum of the initial weights of the remaining features is calculated and denoted as the initial weight ratio of the remaining features. Calculate the product of the initial weight percentage of each of the remaining features and the total increment of the core feature, and record it as the reduction amount of the remaining features; The difference between the initial weight coefficient and the reduction amount of each of the remaining features is calculated and recorded as the adjustment weight coefficient of the remaining features. The initial weight coefficients of the remaining weights are successively reduced to the corresponding adjustment weight coefficients of the remaining features.

[0036] Please continue reading. Figure 3 As shown, the process of determining the time step based on the increase in the weight coefficients of the core time series features after the weight coefficients of the core time series features have been increased in this embodiment of the invention includes: When the weight coefficient of the core time series feature is increased to 2.17 times the weight coefficient of the initial core time series feature, the time step is increased to 2.02 times the initial time step; When the weight coefficient of the core time series feature is increased to 1.54 times the weight coefficient of the initial core time series feature, the time step is increased to 1.61 times the initial time step; When the weight coefficient of the core time series feature is increased to 1.21 times the weight coefficient of the initial core time series feature, the time step is increased to 1.28 times the initial time step; After the time step is increased, the movement interval is reduced based on the increase in the time step. When the time step is increased to 2.02 times the initial time step, the movement interval is reduced to 0.54 times the initial movement interval; When the time step is increased to 1.61 times the initial time step, the movement interval is reduced to 0.67 times the initial movement interval; When the time step is increased to 1.28 times the initial time step, the movement interval is reduced to 0.88 times the initial movement interval; Specifically, the value of the time step is derived from the actual debugging results; The time step refers to the time span during which the prediction model reads and analyzes the continuous historical multi-source time series data, which determines the time accumulation pattern of the prediction model in capturing core features.

[0037] The moving interval is the time movement distance between two adjacent segments when the prediction model extracts continuous data segments of the multi-source time series data within the preset time period.

[0038] Please see Figure 4 The flowchart shown illustrates the reasons why the predicted probability of pests and diseases does not meet the standard in an embodiment of the present invention. The process of determining whether the predicted probability of pests and diseases meets the standard based on the determination coefficient after the moving interval has decreased in this embodiment of the present invention includes: Determine the linear correlation between the predicted probability and the actual probability, and record the obtained linear correlation as the determination coefficient; When the determination coefficient is greater than or equal to the preset determination coefficient Y, the predicted probability of pests and diseases is determined based on the trend consistency ratio to determine whether it meets the standard. When the determination coefficient is less than the preset determination coefficient Y, the predicted probability of pests and diseases is determined to be non-compliant with the standard, and a technical optimization failure notification is issued.

[0039] Please continue reading. Figure 4 As shown, the process of adjusting the attention weight decay coefficient based on the trend consistency ratio in this embodiment of the invention includes: Calculate the difference between the preset trend consistency ratio and the trend consistency ratio, and record the obtained difference as the consistency difference. The attention weight decay coefficient is adjusted based on the consistency difference. When the consistency difference is greater than the second preset consistency difference When this occurs, the attention weight attenuation coefficient is increased to 2.06 times the initial attention weight attenuation coefficient, wherein the second preset consistency difference in this embodiment... ; When the consistency difference is less than or equal to the second preset consistency difference When the difference is greater than the first preset consistency difference △G1, the attention weight attenuation coefficient is increased to 1.59 times the initial attention weight attenuation coefficient. In this embodiment, the first preset consistency difference... ; When the consistency difference is less than or equal to the first preset consistency difference At that time, the attention weight decay coefficient is increased to 1.27 times the initial attention weight decay coefficient; When the consistency difference is greater than the second preset consistency difference At that time, the attention weight decay coefficient is reduced to 0.54 times the initial attention weight decay coefficient; When the consistency difference is less than or equal to the second preset consistency difference And greater than the first preset consistency difference value At that time, the attention weight decay coefficient is reduced to 0.68 times the initial attention weight decay coefficient; When the consistency difference is less than or equal to the first preset consistency difference At that time, the attention weight decay coefficient is reduced to 0.83 times the initial attention weight decay coefficient; Specifically, the value of the attention weight attenuation coefficient is derived from the actual debugging results; When the predicted probability trend predicted by the prediction model is later than the actual probability, the attention weight decay coefficient is reduced. When the predicted probability trend predicted by the prediction model is earlier than the actual probability, the attention weight decay coefficient is increased. Adjusting the attention weight decay coefficient allows the prediction model to focus on the time segment that has the greatest impact on the trend.

[0040] Please continue reading. Figure 4 As shown, the process of refining the time granularity based on the adjustment level of the attention weight decay coefficient after the attention weight decay coefficient adjustment is completed in this embodiment of the invention includes: When the adjustment level of the attention weight decay coefficient is the first adjustment level, the time granularity is refined to 0.81 times the initial time granularity; When the adjustment level of the attention weight decay coefficient is the second adjustment level, the time granularity is refined to 0.63 times the initial time granularity; When the adjustment level of the attention weight decay coefficient is the third adjustment level, the time granularity is refined to 0.47 times the initial time granularity; Specifically, increasing the attention weight decay coefficient to 2.06 times the initial attention weight decay coefficient constitutes the third adjustment level; The second adjustment level is achieved when the attention weight attenuation coefficient is increased to 1.59 times the initial attention weight attenuation coefficient. The first adjustment level is when the attention weight decay coefficient is increased to 1.27 times the initial attention weight decay coefficient; The third adjustment level is achieved when the attention weight attenuation coefficient is reduced to 0.54 times the initial attention weight attenuation coefficient. The second adjustment level is defined as reducing the attention weight decay coefficient to 0.68 times the initial attention weight decay coefficient. The first adjustment level is when the attention weight decay coefficient is reduced to 0.83 times the initial attention weight decay coefficient; Refining the time granularity can improve the effectiveness and accuracy of the temporal correlation features.

[0041] Please continue reading. Figure 4 As shown, the process of determining whether the predicted probability of pests and diseases meets the standard based on the trend consistency ratio after the time granularity refinement in this embodiment of the invention includes: When the trend consistency ratio is greater than or equal to the preset trend consistency ratio B, it is determined that the predicted probability of pests and diseases meets the standard, the probability prediction of pests and diseases is completed, and the probability prediction of pests and diseases with prediction needs is performed. When the trend consistency ratio is less than the preset trend consistency ratio B, it is determined that the predicted probability of pests and diseases does not meet the standard, and a consistency ratio optimization failure notification is issued.

[0042] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0043] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting the probability of pests and diseases based on multi-temporal feature fusion, characterized in that, include: Collect multi-source time-series data within a preset time period, including environmental time-series data, crop growth time-series data, historical time-series data of pests and diseases, and time-series data of agricultural operations; The temporal granularity of different time series data is aligned using a time granularity alignment method, and temporal correlation features related to pests and diseases are mined to generate a fused feature vector; The fused feature vector is input into the prediction model, and the model combines the weights of each dimension in the vector to predict the probability of pests and diseases occurring within the expected time. Determine the determination coefficient and trend consistency ratio of the probability of pest and disease occurrence within the expected time period; The predicted probability of pests and diseases is determined based on the determination coefficient to see if it meets the standard. Based on the determination coefficient, determine whether the predicted probability of pests and diseases based on the trend consistency ratio meets the standard. Alternatively, generate corresponding processing instructions; Based on the trend consistency ratio determination results, the probability prediction of pests and diseases is completed. Alternatively, generate corresponding processing instructions; Adjust the weight coefficient of the core time series feature based on the determination coefficient, adjust the attention weight decay coefficient based on the trend consistency ratio, increase the time step based on the increase of the weight coefficient of the core time series feature, decrease the movement interval based on the increase of the time step, refine the time granularity based on the adjustment level of the attention weight decay coefficient, and issue a technical optimization failure notification, or issue a consistency ratio optimization failure notification.

2. The method for predicting the probability of pests and diseases based on multi-temporal feature fusion according to claim 1, characterized in that, The process of determining whether the predicted probability of pests and diseases meets the standard based on the determination coefficient includes: Determine the linear correlation between the predicted probability and the actual probability, and record the obtained linear correlation as the determination coefficient; When the determination coefficient is greater than or equal to the preset determination coefficient, the predicted probability of pests and diseases is determined based on the trend consistency ratio to determine whether it meets the standard. When the determination coefficient is less than the preset determination coefficient, the predicted probability of pests and diseases is determined to be inconsistent with the standard, and the weight coefficient of the core time series feature is adjusted based on the determination coefficient.

3. The method for predicting the probability of pests and diseases based on multi-temporal feature fusion according to claim 2, characterized in that, The process of determining whether the predicted probability of pests and diseases meets the standard based on the trend consistency ratio includes: The number of days in which the predicted probability rises and falls in line with the actual probability rises and falls is recorded as the number of days in which the trend is consistent. Calculate the ratio of the number of days with consistent trend to the total number of days predicted within the expected time period, and record the obtained ratio as the trend consistency ratio; When the trend consistency ratio is greater than or equal to the preset trend consistency ratio, it is determined that the predicted probability of pests and diseases meets the standard, the probability prediction of pests and diseases is completed, and the probability prediction of pests and diseases with prediction needs is performed. When the trend consistency ratio is less than the preset trend consistency ratio, it is determined that the predicted probability of pests and diseases does not meet the standard, and the attention weight decay coefficient is adjusted based on the trend consistency ratio.

4. The method for predicting the probability of pests and diseases based on multi-temporal feature fusion according to claim 2, characterized in that, The process of increasing the weight coefficient of the core time-series feature based on the determination coefficient includes: Calculate the difference between the preset decision coefficient and the decision coefficient, and record the obtained difference as the decision coefficient difference; The weight coefficient of the core time series feature is increased based on the difference in the determination coefficients, and the increase in the weight coefficient of the core time series feature is proportional to the difference in the determination coefficients.

5. The method for predicting the probability of pests and diseases based on multi-temporal feature fusion according to claim 4, characterized in that, After the weight coefficients of the core time series features are increased, the process of determining the time step based on the increase in the weight coefficients of the core time series features includes: The time step is determined by the increase in the weight coefficients of the core time series features, and the increase in the time step is proportional to the increase in the weight coefficients of the core time series features. After the time step is increased, the moving interval is reduced based on the increase in the time step, and the reduction in the moving interval is proportional to the increase in the time step.

6. The method for predicting the probability of pests and diseases based on multi-temporal feature fusion according to claim 5, characterized in that, The process of determining whether the predicted probability of pests and diseases meets the standard based on the determination coefficient after the movement interval has been reduced includes: Determine the linear correlation between the predicted probability and the actual probability, and record the obtained linear correlation as the determination coefficient; When the determination coefficient is greater than or equal to the preset determination coefficient, the predicted probability of pests and diseases is determined based on the trend consistency ratio to determine whether it meets the standard. When the determination coefficient is less than the preset determination coefficient, the predicted probability of pests and diseases is determined to be non-compliant with the standard, and a technical optimization failure notification is issued.

7. The method for predicting the probability of pests and diseases based on multi-temporal feature fusion according to claim 3, characterized in that, The process of adjusting the attention weight decay coefficient based on the trend consistency ratio includes: Calculate the difference between the preset trend consistency ratio and the trend consistency ratio, and record the obtained difference as the consistency difference. The attention weight decay coefficient is adjusted based on the consistency difference, and the adjustment range of the attention weight decay coefficient is proportional to the consistency difference.

8. The method for predicting the probability of pests and diseases based on multi-temporal feature fusion according to claim 7, characterized in that, The process of refining the time granularity based on the adjustment level of the attention weight decay coefficient after the attention weight decay coefficient has been adjusted includes: The time granularity is refined based on the adjustment level of the attention weight decay coefficient, and the degree of refinement of the time granularity is proportional to the adjustment level of the attention weight decay coefficient.

9. The method for predicting the probability of pests and diseases based on multi-temporal feature fusion according to claim 8, characterized in that, The process of determining whether the predicted probability of pests and diseases meets the standard based on the trend consistency ratio after the time granularity is refined includes: When the trend consistency ratio is greater than or equal to the preset trend consistency ratio, it is determined that the predicted probability of pests and diseases meets the standard, the probability prediction of pests and diseases is completed, and the probability prediction of pests and diseases with prediction needs is performed. When the trend consistency ratio is less than the preset trend consistency ratio, it is determined that the predicted probability of pests and diseases does not meet the standard, and a consistency ratio optimization failure notification is issued.

10. A pest and disease probability prediction system based on multi-temporal feature fusion using the method described in any one of claims 1-9, characterized in that, include: The data input layer is used to collect the multi-source time-series data within the preset time period, including the environmental time-series data, the crop growth time-series data, the historical time-series data of pests and diseases, and the time-series data of agricultural operations. A feature fusion layer, which is connected to the data input layer, is used to align the time granularity of different time series data through a time granularity alignment method, and to mine the time series correlation features related to pests and diseases to generate the fused feature vector. The prediction output layer, which is connected to the feature fusion layer, is used to input the fused feature vector into the prediction model. The model combines the weights of each dimension in the vector to predict the probability of the occurrence of pests and diseases within the expected time. The intelligent analysis layer, which is connected to the prediction output layer, is used to determine the determination coefficient and the trend consistency ratio of the probability of pests and diseases occurring within the expected time period; the intelligent analysis layer is also used to determine whether the predicted probability of pests and diseases meets the standard based on the determination coefficient. The intelligent analysis layer is also used to determine whether the predicted probability of pests and diseases based on the trend consistency ratio meets the standard, or to generate corresponding processing instructions, based on the determination result of the determination coefficient. The intelligent analysis layer is also used to determine the probability prediction of pests and diseases based on the trend consistency ratio judgment result, or to generate corresponding processing instructions. A parameter adjustment layer, which is connected to the data input layer, the feature fusion layer, the prediction output layer, and the intelligent analysis layer, is used to adjust the weight coefficient of the core time-series feature based on the determination coefficient, adjust the attention weight decay coefficient based on the trend consistency ratio, increase the time step based on the increase in the weight coefficient of the core time-series feature, decrease the movement interval based on the increase in the time step, refine the time granularity based on the adjustment level of the attention weight decay coefficient, issue a technical optimization failure notification, and issue a consistency ratio optimization failure notification.

Images