A big data-based seed industry pest and disease early warning method and system

Abstract

This invention relates to the field of pest and disease early warning technology, and discloses a method and system for early warning of seed industry pests and diseases based on big data. The method includes acquiring environmental sound data and leaf image data; extracting lesions from the leaf image data to obtain lesion morphological features; extracting features from the environmental sound data to obtain pest sound features; matching the lesion morphological features and pest sound features with a pre-stored pest and disease database to obtain pest type and disease type; calculating the matching degree between the pest type and the disease type to obtain the pest-disease matching degree; and assessing the damage based on the pest-disease matching degree and a preset matching degree threshold to obtain the disease assessment result. This method has the following advantages: it can improve the accuracy of early warning of seed industry pests and diseases.

Description

Technical Field

[0001] This invention relates to the field of pest and disease early warning technology, and in particular to a method and system for early warning of pests and diseases in the seed industry based on big data. Background Technology

[0002] Currently, with the development of agricultural technology, big data-based seed disease and pest early warning methods and systems are gradually becoming important means to improve crop yield and quality. By integrating multi-source data such as meteorological data, soil information, crop growth status, and historical disease and pest occurrence records, this system can accurately predict and warn of disease and pest risks respectively. Especially for large-scale planting areas, big data technology can not only promptly detect potential outbreak trends of diseases and dynamic occurrences of pests, but also formulate targeted prevention and control strategies in advance, reduce the use of chemical pesticides, and protect the ecological environment. In one existing technology, the sound monitoring technology used in the big data-based seed disease and pest early warning system mainly relies on the unique sound characteristics emitted by pests for identification. The specific process is as follows: First, multiple high-sensitivity microphones are set up in the field to capture various sound signals in the environment; then, these sound spectrum data are transmitted to a central processing unit for analysis. With the help of machine learning algorithms, the system can identify specific audio frequency bands produced by pests when they fly or move, and match them with known pest sound patterns. To enhance accuracy, the system also combines other data such as weather conditions and time points for comprehensive judgment. Once an abnormal increase in pest activity is detected, the system will automatically trigger an alarm and send detailed warning information to farmers, including suggested measures. This method greatly improves the efficiency of pest identification.

[0003] For disease early warning, the system mainly analyzes the correlation between environmental data (such as temperature, humidity, and rainfall) and crop growth status (such as leaf color and lesion distribution), combined with historical disease occurrence records, to establish a disease risk prediction model. For example, by monitoring the correlation between soil moisture and specific fungal diseases, or by using drone images to identify early disease symptoms, the system can provide early warnings of disease outbreak risks and offer non-chemical control suggestions (such as adjusting irrigation strategies or biological control programs).

[0004] While the aforementioned pest early warning methods have numerous theoretical advantages, they face a key challenge in practical applications: the sound characteristics of pests often overlap with the frequencies of environmental noise, making accurate separation during feature extraction difficult. This is because the frequency range of sounds produced by many pests is similar to that of wind, rain, or even distant machinery in the natural environment, making pest identification based solely on frequency characteristics extremely difficult in complex acoustic environments. Furthermore, attempting to remove background noise by adding filters simultaneously weakens the sound signal intensity of the target pest, thus affecting the final identification result. In addition, background noise characteristics change due to variations in different geographical regions and seasons, rendering fixed filtering schemes unsuitable for all situations. This means that existing systems need more intelligent and adaptive methods to distinguish between pest sounds and environmental noise. This limitation restricts the widespread application and effectiveness of sound-based pest early warning systems, especially in open farmland environments with high noise levels. In summary, the accuracy of existing seed industry pest early warning methods in the pest sound extraction process still needs improvement. Summary of the Invention

[0005] This invention provides a method and system for early warning of seed pests and diseases based on big data, in order to improve the accuracy of sound feature extraction in the early warning process of seed pests and diseases.

[0006] Firstly, in order to solve the above-mentioned technical problems, the present invention provides a method for early warning of seed pests and diseases based on big data, comprising:

[0007] Acquire environmental sound data and blade image data;

[0008] Based on the leaf image data, lesion morphology characteristics were extracted to obtain the lesion morphology features.

[0009] Based on the environmental sound data, feature extraction is performed to obtain the sound characteristics of pests;

[0010] The pest type and disease type are obtained by matching the morphological characteristics of the lesions and the sound characteristics of the pests with a pre-stored pest and disease database.

[0011] The matching degree is calculated based on the pest type and the disease type to obtain the pest-disease matching degree.

[0012] Damage assessment is performed based on the pest and disease matching degree and the preset matching degree threshold to obtain the disease assessment result.

[0013] In one optional implementation, the step of extracting lesion morphology features from the leaf image data includes:

[0014] Image segmentation is performed based on the leaf image data to obtain images of lesion areas;

[0015] The lesion region image is input into a pre-trained lesion extraction model to obtain lesion morphological features;

[0016] The training process of the lesion extraction model includes:

[0017] The lesion extraction model is trained based on a convolutional neural network, with the input layer being historical lesion images and the output layer being texture shape feature vectors;

[0018] Training is completed when the model's loss function meets the conditions or the number of training iterations reaches the preset limit, and the trained model is obtained.

[0019] In one optional implementation, the step of extracting features from the environmental sound data to obtain the sound features of pests includes:

[0020] Noise reduction filtering is performed based on the ambient sound data and preset filtering parameters to obtain noise-reduced sound data;

[0021] Based on the noise reduction audio data, time-frequency analysis is performed to obtain frequency domain characteristics and time domain characteristics;

[0022] Blind source separation is performed on the noise-reduced sound data that overlaps with the frequency domain features and environmental noise to obtain pure pest sounds;

[0023] Time-frequency analysis was performed on the pure insect sounds to obtain the characteristics of the insect sounds.

[0024] In one optional implementation, the step of matching the morphological characteristics of the lesions and the sound characteristics of the pests with a pre-stored pest and disease database to obtain the pest type and disease type includes:

[0025] The similarity between the lesion morphology and the disease feature data in the pest and disease database is calculated to obtain the disease similarity.

[0026] When the disease similarity is greater than a preset first similarity threshold, it is determined to be the disease type corresponding to the disease feature data;

[0027] The similarity between the insect sound characteristics and the insect characteristic data in the insect disease database is calculated to obtain the insect similarity.

[0028] When the similarity of the pests is greater than a preset second similarity threshold, it is determined to be the pest type corresponding to the pest feature data.

[0029] In one optional implementation, the step of calculating the matching degree based on the pest type and the disease type to obtain the pest-disease matching degree includes:

[0030] Obtain the infringement vector database;

[0031] Based on the pest type, the pest infestation vector is retrieved from the infestation vector database to obtain the pest infestation vector.

[0032] Based on the disease type, the disease damage vector is retrieved from the damage vector database.

[0033] The pest-disease matching degree is obtained by calculating the cosine correlation between the pest infestation vector and the disease infestation vector.

[0034] In one optional implementation, the step of assessing the damage based on the pest / disease matching degree and a preset matching degree threshold to obtain a disease assessment result includes:

[0035] When the matching degree of the pests and diseases is lower than the preset matching degree threshold, the determination results of the pest type and the disease type are marked as unreliable, and data acquisition is performed again.

[0036] When the matching degree of the pest disease is higher than the preset matching degree threshold, the distribution of pests is analyzed based on the disease type, the sound characteristics of the pests, and the pest type to obtain a pest distribution map.

[0037] Based on the pest and disease distribution map, density statistics were performed to obtain the pest and disease distribution density.

[0038] Areas where the distribution density of pests and diseases is greater than a preset density threshold are identified as high-density areas;

[0039] Areas where the distribution density of pests and diseases is less than or equal to a preset density threshold are identified as low-density areas.

[0040] The disease assessment results include disease type, pest type, and pest distribution density.

[0041] In an optional implementation, after matching the lesion morphology characteristics and the insect sound characteristics with a pre-stored pest and disease database to obtain the pest type and disease type, the method further includes:

[0042] The morphological features of the lesions and the sound features of the pests are fused to obtain comprehensive feature data.

[0043] The comprehensive feature data is added to the pest and disease database.

[0044] Secondly, the present invention provides a seed industry pest and disease early warning system based on big data, comprising:

[0045] The data acquisition module is used to acquire environmental sound data and blade image data;

[0046] The lesion extraction module is used to extract lesions based on the leaf image data to obtain the lesion morphology characteristics;

[0047] The pest sound module is used to extract features from the environmental sound data to obtain pest sound features.

[0048] The data retrieval module is used to match the morphological characteristics of the lesions and the sound characteristics of the pests with a pre-stored pest and disease database to obtain the pest type and disease type.

[0049] The type matching module is used to calculate the matching degree based on the pest type and the disease type to obtain the pest-disease matching degree.

[0050] The assessment results module is used to assess the damage based on the pest and disease matching degree and a preset matching degree threshold, and obtain the disease assessment results.

[0051] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the seed industry pest and disease early warning method based on big data as described above.

[0052] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute any of the above-described big data-based early warning methods for seed pests and diseases.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] (1) The process of acquiring environmental sound data and leaf image data ensures comprehensive monitoring of plant health. Acquiring ambient sound information and detailed images of the leaf surface using devices such as high-sensitivity microphones and high-definition cameras provides abundant raw data for subsequent analysis. This process not only improves the quality and completeness of the data but also lays a solid foundation for accurate identification of pests and diseases.

[0055] (2) Based on the leaf image data, lesion morphology features are extracted. Advanced image processing techniques, such as edge detection, color segmentation, and texture analysis, are used to accurately extract lesion areas from the complex leaf background and quantify their shape, size, color, and distribution. These features provide crucial information for identifying different types of diseases and help improve diagnostic accuracy.

[0056] (3) Feature extraction is performed based on the environmental sound data to obtain the sound characteristics of pests. Acoustic analysis tools are used to perform spectral analysis on the sound signals in the environment to identify and separate the unique audio signals emitted by potential pests. Through detailed analysis of these sound characteristics, different types of pests and their activity patterns can be distinguished, providing strong support for subsequent pest identification.

[0057] (4) The morphological characteristics of the lesions and the sound characteristics of the pests are matched with a pre-stored pest and disease database to obtain the pest type and disease type. The extracted morphological characteristics of the lesions and the sound characteristics of the pests are compared with known samples in the database to find the closest match. This method can quickly and accurately determine the specific type of pests and diseases, significantly improving diagnostic efficiency and reliability.

[0058] (5) Calculate the matching degree based on the pest type and the disease type to obtain the pest-disease matching degree. After determining the pest and disease types, further calculate the matching degree of each combination. This step considers multiple factors, including the cosine similarity of features, constructs an infestation vector, and compares and matches it with the template library to obtain the most corresponding pest and disease type.

[0059] (6) Damage assessment is performed based on the pest / disease matching degree and a preset matching degree threshold to obtain disease assessment results. Based on the calculated pest / disease matching degree and referring to the set matching degree threshold, pest / disease identification results and damage distribution are generated. This scientifically based assessment method not only improves the objectivity of decision-making but also helps farmers take timely and effective prevention and control measures to minimize losses and ensure crop yield and quality. In this way, intelligent monitoring and management of pests and diseases in the agricultural ecosystem are realized, promoting the development of precision agriculture. Attached Figure Description

[0060] Figure 1 This is a schematic diagram of a seed industry pest and disease early warning method based on big data provided in the first embodiment of the present invention;

[0061] Figure 2 This is a schematic diagram of a seed industry pest and disease early warning system based on big data, provided in the second embodiment of the present invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] Reference Figure 1 The first embodiment of the present invention provides a method for early warning of seed pests and diseases based on big data, including the following steps:

[0064] S11, acquire ambient sound data and blade image data;

[0065] S12, Extract lesions from the leaf image data to obtain lesion morphology characteristics;

[0066] S13, feature extraction is performed based on the environmental sound data to obtain the sound features of pests;

[0067] S14. Match the lesion morphology characteristics and the insect sound characteristics with a pre-stored pest and disease database to obtain the pest type and disease type.

[0068] S15, calculate the matching degree based on the pest type and the disease type to obtain the pest-disease matching degree;

[0069] S16, based on the pest and disease matching degree and the preset matching degree threshold, a damage assessment is performed to obtain the disease assessment result.

[0070] In step S11, environmental sound data and blade image data are acquired.

[0071] In one implementation, environmental sound data can be collected using distributed, highly sensitive acoustic sensors (such as MEMS microphone arrays or ultrasonic sensors) with a frequency response range covering 0-100kHz to capture the sound waves of pest activity. These sensors are deployed at a density of 3-5 nodes per hectare to evenly cover the farmland, focusing on monitoring the area 30-50cm below the crop canopy. Leaf image data is acquired using smart cameras with multispectral imaging capabilities (such as CMOS sensors with at least 5 megapixels and near-infrared band). The equipment is mounted on a field support frame to create a 45° overhead angle, and an automatic rotating gimbal enables six-directional scanning of each crop. Both types of data are transmitted in real-time to an edge computing gateway via the LoRaWAN IoT protocol. After H.265 encoding and compression, the data is stored in a local NAS storage system with RAID5 redundancy. Simultaneously, it is uploaded to a cloud-based time-series database via a 5G network. The sound data is stored in WAV format in 10-minute segments, and the image data is saved in JPEG2000 format containing EXIF ​​metadata. All data is supplemented with environmental parameters such as GPS positioning and temperature and humidity sensor readings to form a structured dataset.

[0072] In step S12, lesion morphology features are extracted based on the leaf image data.

[0073] Image segmentation is performed based on the leaf image data to obtain images of lesion areas;

[0074] The lesion region image is input into a pre-trained lesion extraction model to obtain lesion morphological features;

[0075] The training process of the lesion extraction model includes:

[0076] The lesion extraction model is trained based on a convolutional neural network, with the input layer being historical lesion images and the output layer being texture shape feature vectors;

[0077] Training is completed when the model's loss function meets the conditions or the number of training iterations reaches the preset limit, and the trained model is obtained.

[0078] It is worth noting that the specific steps in the image segmentation process are as follows: First, the leaf image data is preprocessed. A Gaussian filtering algorithm is used to smooth the image. By setting a 3×3 or 5×5 convolution kernel window, the neighborhood weighted average is calculated pixel by pixel to eliminate noise interference. Next, histogram equalization technology is used to stretch and adjust the pixel value distribution of each channel of the image, especially enhancing the contrast difference between diseased areas and healthy leaves in the red and blue channels. Then, the RGB color space is converted to HSV or Lab mode. By calculating the joint distribution histogram of hue and saturation components, the grayscale threshold that can best distinguish diseased areas from the background is found. At this time, the system will traverse the grayscale value range of 0-255 and automatically select the critical value that maximizes the inter-class variance between the foreground and background as the segmentation standard to generate a binary mask. After generating the binary mask, morphological processing is performed. First, a 3×3 circular structuring element is used to perform an erosion operation on the image to eliminate isolated noise points. Then, a dilation operation is used to repair broken edges. At the same time, a region filling algorithm is used to repair the internal holes of the diseased areas. Subsequently, edge detection technology is used to determine strong and weak edges by calculating the horizontal and vertical gradient magnitudes of pixels and combining this with a double thresholding method, ultimately generating continuous lesion contour lines. Region growing optimization is then applied to the initial segmentation results. Starting from seed points on the contour lines, regions are expanded based on whether the color difference between adjacent pixels is less than a preset threshold (e.g., the Euclidean distance in RGB space is less than 15), until all broken regions are connected. Finally, a connected component labeling algorithm is used to statistically analyze each independent lesion region, calculating parameters such as the pixel area, aspect ratio of the bounding rectangle, and edge perimeter of each region. After removing interference regions with fewer than 50 pixels, a lesion region image with coordinate localization and geometric feature annotations is generated.

[0079] It is worth noting that the training process of the lesion extraction model is as follows: First, a large number of labeled historical lesion images are collected as a training set. The images are then standardized, including unifying the resolution to 512×512 pixels, performing Gaussian blur denoising, and data augmentation through random rotation and mirror flipping. Subsequently, a deep convolutional neural network is constructed, with a multi-layered backbone structure. The first layer extracts edge texture features through 3×3 convolutional kernels, the second layer compresses the feature dimension through max pooling, and four sets of residual modules are then stacked to capture the complex morphological changes of lesions. An attention mechanism is embedded before the fully connected layer. The control module strengthens the weight of key regions; during training, a contrastive loss function is used, and the labeled texture shape feature vector (including 12-dimensional parameters such as area ratio, edge jaggedness, and shape irregularity index) is used as a supervision signal. The network parameters are optimized round by round through the backpropagation algorithm. After each iteration, the Euclidean distance between the predicted feature vector and the true label is calculated as the loss value. Training is automatically terminated when the loss value decreases by less than one-thousandth for 10 consecutive training cycles or when the total number of training cycles reaches 200. Finally, the model performance is evaluated through a validation set, and model versions with feature matching errors below 5% are selected for practical application.

[0080] In step S13, feature extraction is performed based on the environmental sound data to obtain the sound features of pests.

[0081] In one embodiment, noise reduction filtering is performed based on the ambient sound data and preset filtering parameters to obtain noise-reduced sound data;

[0082] Based on the noise reduction audio data, time-frequency analysis is performed to obtain frequency domain characteristics and time domain characteristics;

[0083] Blind source separation is performed on the noise-reduced sound data that overlaps with the frequency domain features and environmental noise to obtain pure pest sounds;

[0084] Time-frequency analysis was performed on the pure insect sounds to obtain the characteristics of the insect sounds.

[0085] It is worth noting that the extraction process of pest sound features is implemented as follows: First, based on preset filtering parameters (such as a passband frequency range of 200Hz-8kHz and a stopband attenuation of 40dB), adaptive noise reduction filtering is performed on the original environmental sound data, using a Butterworth bandpass filter bank to suppress low-frequency mechanical vibration noise and high-frequency electromagnetic interference; then, a short-time Fourier transform (STFT) is performed on the denoised sound signal, with a 25ms Hamming window and a 10ms step size, to extract 24-dimensional time-frequency features including Mel-frequency cepstral coefficients (MFCC), spectral centroid, and zero-crossing rate, while simultaneously calculating the root mean square energy, pulse duration, and repetition frequency in the time domain; for frequency... The mixed signal, which overlaps with domain features and environmental noise (such as wind and rain), is subjected to blind source separation using an improved Independent Component Analysis (ICA) algorithm. By maximizing the non-Gaussianity metric between signals, the pure acoustic signal of the target pest is separated from the multi-channel recording. Finally, the separated signal is subjected to refined time-frequency analysis. Continuous wavelet transform (CWT) is used to extract the sound features of pests feeding or flapping their wings in the range of 1-12kHz using Morlet wavelet basis functions. These features include key bioacoustic parameters such as the main frequency band energy distribution (e.g., lepidopteran pests are concentrated in 3-5kHz), pulse sequence time interval (approximately 50-200ms for coleopterans), and harmonic attenuation slope.

[0086] In one implementation, during the blind source separation stage, the system first performs multi-scale time-frequency analysis on the denoised mixed audio. A sliding window mechanism is used to segment the continuous audio into 50-millisecond segments and overlays a Hamming window. A three-dimensional spectrum (frequency-time-energy) is generated through short-time Fourier transform. Targeting the 3.5kHz dominant frequency pulse group characteristic unique to the rice leaf roller, the system constructs a dynamic mask template: in the spectral dimension, morphological dilation is performed on harmonic clusters exhibiting periodic abrupt changes in the 2-5kHz frequency band (e.g., five consecutive peaks appearing at 0.2-second intervals) to enhance the target signal; in the temporal dimension, a temporal energy baseline is established using the diurnal activity patterns of the pest (e.g., the rice stem borer's nighttime chirping intensity increases by 40%). Subsequently, non-negative matrix factorization is used to decompose the mixed spectrum into a basis matrix (representing insect chirping characteristics) and a coefficient matrix (representing environmental noise). Through iterative optimization, the cosine similarity between the decomposed insect chirping basis and a preset voiceprint feature library (e.g., containing standard chirping patterns of 20 Lepidoptera insects) reaches over 85%. For residual broadband noise (such as wind noise), the system will use an adaptive spectral subtraction algorithm: based on the median noise background energy of 10 adjacent time-frequency windows, the attenuation coefficient of each frequency point is dynamically adjusted, and finally the pure insect chirping signal is reconstructed through inverse Fourier transform, while preserving key biological characteristics (such as the 12Hz modulation ripple of the mating signal).

[0087] In step S14, the pest type and disease type are obtained by matching the morphological characteristics of the lesions and the sound characteristics of the pests with a pre-stored pest and disease database.

[0088] In one embodiment, a similarity calculation is performed based on the morphological characteristics of the lesions and the disease feature data in the pest disease database to obtain a disease similarity; when the disease similarity is greater than a preset first similarity threshold, it is determined to be the disease type corresponding to the disease feature data; a similarity calculation is performed based on the sound characteristics of the pests and the pest feature data in the pest disease database to obtain a pest similarity; when the pest similarity is greater than a preset second similarity threshold, it is determined to be the pest type corresponding to the pest feature data.

[0089] It is worth noting that when the similarity between diseases or pests does not reach the preset threshold, the system will take different strategies for subsequent processing: For diseases, if the similarity is lower than the threshold, it indicates that the current lesion morphology does not significantly match the known disease characteristics in the database. The system will mark it as "unknown disease type" and trigger a secondary verification process. For example, it will collect multi-angle images of lesions or combine them with environmental data (such as humidity and temperature) for comprehensive analysis. At the same time, it will push a prompt message to the farmer, suggesting manual re-examination or providing samples for testing. For pests, if the similarity does not reach the threshold, the system will determine that "the pest type cannot be identified". It will start the re-collection of sound signals or extend the monitoring time to obtain more complete data, and simultaneously combine other sensor information (such as crop damage degree and environmental vibration frequency) for cross-verification. If it still cannot be determined, it will send an early warning prompt to the farmer, suggesting strengthening field inspections or adopting broad-spectrum preventive measures. At the same time, the unidentified data will be stored in the database for subsequent model optimization reference.

[0090] It is worth noting that the pest and disease identification and matching process is based on a multimodal feature fusion mechanism: First, a pre-stored pest and disease database is invoked, which contains lesion morphology parameters of over 300 typical diseases (such as edge serration density, lesion expansion directionality, and color gradient patterns) and voiceprint feature parameters of over 200 pests (such as vibration fundamental frequency, pulse interval regularity, and harmonic attenuation characteristics). For disease identification, the system compares the real-time collected lesion morphology features with each disease template in the database item by item, and calculates the similarity using a weighted comprehensive scoring method (e.g., lesion area ratio accounts for 30%, texture complexity accounts for 25%, and edge sharpness accounts for 45%). When the comprehensive similarity of a disease exceeds the 85% threshold and is 5 percentage points higher than other candidate diseases, it is determined to be that disease type. In the insect identification stage, a dynamic time warping algorithm is used to align the collected voiceprint features with the standard acoustic features in the database in terms of time series. The system focuses on matching two core indicators: pulse interval stability (weight 40%) and spectral energy distribution (weight 35%). When the matching degree reaches 90%, a judgment is triggered, that is, the insect is identified as this type of pest. When the two identification results conflict (e.g., the disease points to fungal infection while the insect pest points to chewing mouthpart pests), the system will activate the multimodal decision module, giving priority to the judgment result with a confidence difference of more than 15%. If the difference is within 10%, a compound pest warning will be output. The entire matching process is equipped with a dual verification mechanism. When the matching degree of a single feature exceeds the threshold but the overall confidence degree is less than 75%, a secondary feature extraction process will be automatically triggered to eliminate misjudgments caused by environmental interference.

[0091] It is worth noting that the Dynamic Time Warping (DTW) algorithm performs time-series alignment processing between the collected voiceprint features and the standard acoustic features in the database through the following steps: First, the voiceprint signal is converted into time-series data, which includes two core features: pulse interval (such as the interval between wing flapping or biting by pests) and spectral energy distribution (the energy proportion of each frequency component). The DTW algorithm constructs a two-dimensional distance matrix, calculates the local distance (such as Euclidean distance) at each time point, and uses a dynamic programming strategy to find the globally optimal alignment path. This path allows the time series to be flexibly stretched on the time axis to adapt to the rhythm or frequency shifts caused by individual differences or environmental interference in the voiceprints of different pests. For example, if the standard voiceprint pulse interval of a certain pest is relatively stable (such as the regularity of wing flapping frequency), while the actual collected voiceprint has a slightly longer interval due to wind speed, DTW can adjust the path to align the pulse patterns of the two. The algorithm simultaneously assigns a weight of 40% to pulse interval stability and 35% to spectral energy distribution, and calculates the matching score of the two together. When the total matching degree (such as the weighted similarity percentage) reaches 90%, the system determines that the voiceprints of the two are highly matched, thereby confirming the pest type. This process solves the problem of temporal misalignment in time series through non-linear alignment, enabling the algorithm to more accurately identify the similarity of core features, rather than simply relying on strict time synchronization.

[0092] In a specific embodiment, a paddy field system achieves accurate identification through a multimodal feature fusion mechanism: when it detects brown lesions with serrated edges on leaves (lesion diameter 1.5cm, accounting for 28% of the leaf area, and edge serration density 8 teeth / cm), it identifies the lesions. 2 When the transition is steep and the center shows a grayish-white net-like mold layer, the system uses a weighted calculation based on the proportion of lesion area (30% weight), texture complexity (25% weight), and edge sharpness (45% weight) to obtain a similarity of 86.7% for rice blast, which is significantly higher than other candidate diseases. At the same time, after the collected 3.2kHz voiceprint signal is aligned by the dynamic time warping algorithm, it is matched with a pulse interval of 120±5ms (40% weight), an energy proportion of 65% in the 3.2kHz frequency band (35% weight), and a harmonic attenuation slope of 0.8dB / oct (25% weight). Finally, it is determined to be rice stem borer with a 90% matching degree. Since the confidence difference between the two is only 3.3% and the typical correlation between the pest type (chewing mouthparts) and the disease (rice blast) is weak, the system activates the multimodal decision module and issues a composite warning (the main warning is for rice blast, and the secondary warning is that the activity of rice stem borer will exacerbate the spread of fungi). In response to the interference of sound signature caused by 95% ambient humidity, the system conducts double verification by retesting the sound signal in the early morning and re-examining the spores in the lesion mold layer. Finally, the system maintains the judgment conclusion with a comprehensive confidence level of 78%.

[0093] In step S15, the matching degree is calculated based on the pest type and the disease type to obtain the pest-disease matching degree.

[0094] In one implementation, an infringement vector database is obtained;

[0095] Based on the pest type, the pest infestation vector is retrieved from the infestation vector database to obtain the pest infestation vector.

[0096] Based on the disease type, the disease damage vector is retrieved from the damage vector database.

[0097] The pest-disease matching degree is obtained by calculating the cosine correlation between the pest infestation vector and the disease infestation vector.

[0098] It's worth noting that a damage vector database containing various biological damage characteristics needs to be pre-built. Each vector in this database consists of multiple dimensions, such as the part of the crop affected (e.g., leaves, roots), the season of occurrence, the transmission route (e.g., insect-borne, soil-borne), and typical symptoms (e.g., spots, wilting). When the system receives the pest and disease types input by the user, it retrieves the corresponding pest and disease damage vectors from the damage vector database. These two vectors quantify the characteristics of the pest and disease in various attributes. Subsequently, the system calculates the similarity between the two vectors using a cosine correlation algorithm. This algorithm analyzes the numerical distribution trends of the two vectors across various dimensions to determine their directional consistency: the smaller the angle between the directions of the two vectors (i.e., the closer the cosine value is to 1), the higher the matching degree of the pest and disease in terms of damage characteristics; conversely, the larger the angle, the lower the matching degree. This calculation process ultimately outputs a pest-disease matching degree value to assess the potential correlation between the two.

[0099] It is worth noting that the formula for calculating cosine correlation is as follows:

[0100]

[0101] Among them, cos sim (A,B) represents the cosine correlation between vectors A and B, A·B represents the dot product of vectors A and B, ∥A∥ represents the magnitude of vector A, and ∥B∥ represents the magnitude of vector B.

[0102] In a specific embodiment, taking aphids and cucumber mosaic virus as examples, the infestation vector can be specifically represented as follows: the aphid infestation vector includes four dimensions: affected parts (leaves, buds), peak season (spring and summer), transmission route (insect vector), and typical symptoms (leaf curling), with each dimension assigned a different numerical weight; the cucumber mosaic virus infestation vector includes four dimensions: affected parts (leaves, flowers), peak season (late spring to summer), transmission route (insect vector), and typical symptoms (yellow spots, malformation). When calculating the matching degree, the system compares each dimension of the two vectors—for example, their transmission routes are completely consistent, their seasons highly overlap, their affected parts partially overlap, and their symptoms are correlated. By analyzing the degree of synergy in the numerical changes of each dimension (e.g., both reach a peak in the transmission route dimension, and their seasonal dimension curves are similar), the system judges the overall directional consistency and finally obtains a high matching degree value (e.g., 0.85), reflecting the high probability of aphids transmitting the virus.

[0103] In step S16, the damage assessment is performed based on the pest and disease matching degree and the preset matching degree threshold to obtain the disease assessment result.

[0104] In one embodiment, when the pest / disease matching degree is lower than a preset matching degree threshold, the determination results of the pest type and the disease type are marked as unreliable, and data acquisition is performed again.

[0105] When the matching degree of the pest disease is higher than the preset matching degree threshold, the distribution of pests is analyzed based on the disease type, the sound characteristics of the pests, and the pest type to obtain a pest distribution map.

[0106] Based on the pest and disease distribution map, density statistics were performed to obtain the pest and disease distribution density.

[0107] Areas where the distribution density of pests and diseases is greater than a preset density threshold are identified as high-density areas;

[0108] Areas where the distribution density of pests and diseases is less than or equal to a preset density threshold are identified as low-density areas.

[0109] The disease assessment results include disease type, pest type, and pest distribution density.

[0110] In the process of analyzing the distribution of pests and diseases, the system first spatially correlates the identified disease types (such as fungal diseases and viral diseases) and pest types (such as aphids and locusts) with their corresponding geographic coordinates (such as the location of sensors or monitoring points). Specifically, the system binds the disease type, pest type, and their acoustic characteristics (such as pest activity frequency and voiceprint intensity) of each monitoring point to geographic coordinates, forming a multidimensional spatial dataset. Subsequently, using spatial clustering algorithms (such as K-means or DBSCAN) or rasterization methods, the farmland area is divided into fixed or dynamic grid cells, and the overall occurrence density of diseases and pests within each cell is calculated. For example, if multiple diseases or pests are detected simultaneously in a certain grid cell, the system will calculate the overall density value based on their degree of damage; if only a single disease or pest exists, its occurrence frequency is directly calculated. Ultimately, the system generates a heatmap (i.e., a pest and disease distribution map) overlaid with disease type, pest type, and density value. Based on a preset density threshold (e.g., an average damage index exceeding 3.0 per square meter), it divides areas into high-density regions (requiring emergency treatment) and low-density regions (for routine monitoring). This process combines spatial statistics with machine learning models, considering both the differences in pest and disease types and quantifying their spatial aggregation, providing a visual basis for precise control. For example, if a high-density area simultaneously contains rice sheath blight (a disease) and rice planthoppers (a pest), the system will mark the area as a "dual high-risk area" and recommend the combined use of biological pesticides and physical control methods.

[0111] In one implementation, when the system determines that the pest / disease matching degree is higher than a preset threshold, it first uses an IoT acoustic sensor array deployed in the farmland to accurately locate the spatial coordinates of the sound source by employing a multi-node signal time difference analysis method. This generates a heat map that reflects density differences using a color gradient—dark red indicates the core insect swarm area where more than 100 sound signals are detected per square meter, while light yellow indicates the edge area where fewer than 20 triggers are detected. Next, the farmland is divided into 10m × 10m grid units, and the total number of sound signal triggers per hour in each grid is counted. The activity intensity is then calibrated by combining the pest size coefficient (e.g., a single call by a large locust can cover an area of ​​3 square meters), and finally, a standardized density index is calculated. For example, if a grid records 800 triggers and the calibration coefficient is 0.8, the density index is 640. When this significantly exceeds the preset threshold of 500, the system automatically marks the area as a high-density red warning zone. Grids with an index between 200 and 500 are marked as yellow monitoring zones, and those with an index below 200 are marked as green safe zones.

[0112] It's worth noting that when calculating the pest density index, the system first counts the total number of sound signal triggers per hour within each grid cell (e.g., the number of times the sound waves generated by pest chirping or activity trigger the sensor), reflecting the real-time activity intensity of pests in that area. However, the size differences among different pests lead to varying sound coverage areas for individual individuals (for example, a single chirp of a large locust can cover 3 square meters, while the sound wave coverage area of ​​a small pest like an aphid is only 0.1 square meters). Therefore, the sound signal data needs to be calibrated using a "size coefficient": specifically, the system divides the total number of sound signal triggers per hour within the grid by the single sound wave coverage area of ​​the pest (i.e., the size coefficient) to estimate the standardized density index of pests within that grid. For example, if a grid detects 100 locust sound signal triggers per hour, its coverage area is 3 square meters, resulting in a standardized density index of 33.3 (unit: pests / square meter), while the same number of triggers for aphids would result in 1000 (unit: pests / square meter). This process eliminates the influence of body size differences on the coverage of sound signals, and transforms the activity intensity of different pests into a comparable "standardized density index", thereby more accurately reflecting the actual pest density distribution.

[0113] In summary, this invention discloses a big data-based early warning method for crop pests and diseases, aiming to improve the accuracy of early warning by integrating multiple data sources and advanced analysis techniques. The method first involves acquiring environmental sound data and leaf image data. High-sensitivity microphones and high-definition cameras are used to collect ambient sound information and detailed images of the leaf surface, providing rich raw data for subsequent analysis. Next, based on the acquired leaf image data, lesion extraction is performed. Advanced image processing techniques such as edge detection, color segmentation, and texture analysis are used to accurately extract lesion areas from the complex leaf background, quantifying their shape, size, color, and distribution characteristics. These features provide crucial evidence for identifying different types of diseases. Simultaneously, after feature extraction from the environmental sound data, acoustic analysis tools are used to perform spectral analysis on the sound signals in the environment, identifying and separating the unique audio signals emitted by potential pests. Detailed analysis of these sound features allows for the differentiation of different types of pests and their activity patterns, providing strong support for subsequent pest identification.

[0114] Furthermore, the method proposed in this invention matches the extracted lesion morphological features and pest sound features with a pre-stored pest and disease database, using machine learning or pattern recognition algorithms to find the closest match, thereby quickly and accurately determining the specific type of pest and disease. This process considers not only the morphological features of the lesions but also the specific sound patterns produced by the pests; this dual verification mechanism greatly improves the accuracy of diagnosis. After determining the pest and disease type, the system further calculates the matching degree of each combination. This process considers multiple factors, including feature similarity and probability of occurrence, thus assigning a quantitative score to each matching result. This scoring mechanism helps assess the credibility of the diagnostic results and guides subsequent damage assessment. Based on the calculated pest and disease matching degree and referring to the set matching degree threshold, the system can automatically determine the degree of crop damage and generate a detailed disease assessment report.

[0115] Reference Figure 2 The second embodiment of the present invention provides a seed industry pest and disease early warning system based on big data, comprising:

[0116] The data acquisition module is used to acquire environmental sound data and blade image data;

[0117] The lesion extraction module is used to extract lesions based on the leaf image data to obtain the lesion morphology characteristics;

[0118] The pest sound module is used to extract features from the environmental sound data to obtain pest sound features.

[0119] The data retrieval module is used to match the morphological characteristics of the lesions and the sound characteristics of the pests with a pre-stored pest and disease database to obtain the pest type and disease type.

[0120] The type matching module is used to calculate the matching degree based on the pest type and the disease type to obtain the pest-disease matching degree.

[0121] The assessment results module is used to assess the damage based on the pest and disease matching degree and a preset matching degree threshold, and obtain the disease assessment results.

[0122] It should be noted that the big data-based early warning system for seed pests and diseases provided in this embodiment of the invention is used to execute all the process steps of the big data-based early warning method for seed pests and diseases in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0123] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a data acquisition program. When the processor executes the computer program, it implements the steps described in the various embodiments of the big data-based seed pest and disease early warning method, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the data acquisition module.

[0124] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0125] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0126] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0127] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0128] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0129] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0130] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that 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 for those skilled in the art.

Claims

1. A method for early warning of seed pests and diseases based on big data, characterized in that, include: Acquire environmental sound data and blade image data; Based on the leaf image data, lesion morphology characteristics were extracted to obtain the lesion morphology features. Based on the environmental sound data, feature extraction is performed to obtain the sound characteristics of pests; The pest type and disease type are obtained by matching the morphological characteristics of the lesions and the sound characteristics of the pests with a pre-stored pest and disease database. The matching degree is calculated based on the pest type and the disease type to obtain the pest-disease matching degree. Damage assessment is performed based on the pest and disease matching degree and the preset matching degree threshold to obtain the disease assessment result; The step of calculating the matching degree based on the pest type and the disease type to obtain the pest-disease matching degree includes: Obtain the infringement vector database; Based on the pest type, the pest infestation vector is retrieved from the infestation vector database to obtain the pest infestation vector. Based on the disease type, the disease damage vector is retrieved from the damage vector database; The pest-disease matching degree is obtained by calculating the cosine correlation between the pest infestation vector and the disease infestation vector. The step of assessing the damage based on the pest / disease matching degree and a preset matching degree threshold to obtain the disease assessment result includes: When the matching degree of the pests and diseases is lower than the preset matching degree threshold, the determination results of the pest type and the disease type are marked as unreliable, and data acquisition is performed again. When the matching degree of the pest disease is higher than the preset matching degree threshold, the distribution of pests is analyzed based on the disease type, the sound characteristics of the pests, and the pest type to obtain a pest distribution map. Based on the pest and disease distribution map, density statistics were performed to obtain the pest and disease distribution density. Areas where the distribution density of pests and diseases is greater than a preset density threshold are identified as high-density areas; Areas where the distribution density of pests and diseases is less than or equal to a preset density threshold are identified as low-density areas. The disease assessment results include disease type, pest type, and pest distribution density; The step of analyzing the distribution of diseases and pests based on the disease type, the sound characteristics of the pests, and the pest type includes: binding the disease type, pest type, and sound characteristics of each monitoring point to geographical coordinates to form a multidimensional spatial dataset; dividing the farmland area into dynamic grid units using a spatial clustering algorithm, and calculating the comprehensive occurrence density of diseases and pests within each unit; if multiple diseases or pests are detected simultaneously in a certain grid unit, calculating the comprehensive density value based on their degree of damage; if only a single disease or pest exists, directly calculating its occurrence frequency.

2. The seed industry pest and disease early warning method based on big data according to claim 1, characterized in that, The step of extracting lesions from the leaf image data to obtain lesion morphological characteristics includes: Image segmentation is performed based on the leaf image data to obtain images of lesion areas; The lesion region image is input into a pre-trained lesion extraction model to obtain lesion morphological features; The training process of the lesion extraction model includes: The lesion extraction model is trained based on a convolutional neural network, with the input layer being historical lesion images and the output layer being texture shape feature vectors; Training is completed when the model's loss function meets the conditions or the number of training iterations reaches the preset limit, and the trained model is obtained.

3. The seed industry pest and disease early warning method based on big data according to claim 1, characterized in that, The step of extracting features from the environmental sound data to obtain the sound features of pests includes: Noise reduction filtering is performed based on the ambient sound data and preset filtering parameters to obtain noise-reduced sound data; Frequency domain analysis is performed on the noise reduction audio data to obtain frequency domain characteristics; Blind source separation is performed on the noise-reduced sound data that overlaps with the frequency domain features and environmental noise to obtain pure pest sounds; Time-frequency analysis was performed on the pure insect sounds to obtain the characteristics of the insect sounds.

4. The seed industry pest and disease early warning method based on big data according to claim 1, characterized in that, The process involves matching the morphological characteristics of the lesions and the sound characteristics of the pests with a pre-stored pest and disease database to obtain the pest type and disease type, including: The similarity between the lesion morphology and the disease feature data in the pest and disease database is calculated to obtain the disease similarity. When the disease similarity is greater than a preset first similarity threshold, it is determined to be the disease type corresponding to the disease feature data; The similarity between the insect sound characteristics and the insect characteristic data in the insect disease database is calculated to obtain the insect similarity. When the similarity of the pests is greater than a preset second similarity threshold, it is determined to be the pest type corresponding to the pest feature data.

5. The seed industry pest and disease early warning method based on big data according to claim 1, characterized in that, After matching the lesion morphology characteristics and the insect sound characteristics with a pre-stored pest and disease database to obtain the pest type and disease type, the process further includes: The morphological features of the lesions and the sound features of the pests are fused to obtain comprehensive feature data. The comprehensive feature data is added to the pest and disease database.

6. A seed industry pest and disease early warning system based on big data, characterized in that, The method for implementing the big data-based early warning system for seed pests and diseases as described in any one of claims 1 to 5 includes: The data acquisition module is used to acquire environmental sound data and blade image data; The lesion extraction module is used to extract lesions based on the leaf image data to obtain the lesion morphology characteristics; The pest sound module is used to extract features from the environmental sound data to obtain pest sound features. The data retrieval module is used to match the morphological characteristics of the lesions and the sound characteristics of the pests with a pre-stored pest and disease database to obtain the pest type and disease type. The type matching module is used to calculate the matching degree based on the pest type and the disease type to obtain the pest-disease matching degree. The assessment results module is used to assess the damage based on the pest and disease matching degree and a preset matching degree threshold, and obtain the disease assessment results.

7. An electronic device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the big data-based seed pest and disease early warning method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the seed industry pest and disease early warning method based on big data as described in any one of claims 1 to 5.

Images