Intelligent soilless culture environment monitoring system and method based on big data

Abstract

The invention discloses an intelligent soilless culture environment monitoring system and method based on big data, and relates to the technical field of soilless culture environment monitoring, and the method comprises the following steps: dividing a soilless culture environment region into a plurality of sub-regions according to the same area size, setting the area of each sub-region according to the specific culture environment and monitoring requirements, therefore, the plants in each sub-region have certain representativeness. According to the invention, the convolutional neural network is utilized to carry out feature extraction and pest prediction on the acquired image data, so that the health condition of each sub-region can be accurately judged, and the focus and focus details of the camera can be automatically adjusted and the focal length can be shortened when a potential pest region is found, so that early symptoms of pests can be clearly captured, and the detection accuracy is improved. The method greatly improves the accuracy of disease and pest identification, can timely discover the hidden danger of disease and pest, guarantees the early intervention of disease and pest, effectively prevents the spread of disease and pest, and guarantees the health of crops in the whole cultivation environment.

Description

Technical Field

[0001] The present invention relates to the technical field of soilless cultivation environment monitoring, and in particular to an intelligent soilless cultivation environment monitoring system and method based on big data. Background Art

[0002] Intelligent soilless cultivation environment monitoring based on big data uses advanced sensor technology, the Internet of Things (IoT) and big data analysis methods to monitor and adjust various environmental parameters in the soilless cultivation system in real time, such as temperature, humidity, light intensity, air circulation, nutrient solution concentration, etc. By collecting this data and combining it with machine learning algorithms and intelligent prediction models, the cultivation environment can be dynamically adjusted to achieve precise control and optimization to ensure that plants grow under optimal growth conditions. This intelligent monitoring system can not only improve crop yield and quality, but also discover potential problems through data analysis, issue early warnings, reduce resource waste, and improve the sustainability and efficiency of agricultural production.

[0003] The prior art has the following deficiencies: In the intelligent soilless cultivation environment monitoring, existing technologies usually rely on a single camera to monitor plant leaves in a large area. However, due to the limitations of the camera's viewing angle, its focal length is usually large, resulting in it only being able to capture images of a wide area, which may mask or ignore the early signs of pests and diseases. Especially at different growth stages of plants, the performance of leaves varies greatly, and images obtained from a single perspective make it difficult to accurately distinguish the specific diseases of each plant. Therefore, the early symptoms of pests and diseases are often not identified in time, delaying the time for intervention, which may eventually lead to the spread of pests and diseases and affect the health of crops in the entire cultivation area.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0005] The purpose of the present invention is to provide an intelligent soilless cultivation environment monitoring system and method based on big data, which ensures the representativeness of plant features in each sub-region by dividing the cultivation area into multiple sub-regions and collecting image data, thereby improving the accuracy of pest and disease monitoring. Combined with convolutional neural networks for feature extraction and pest and disease prediction, it is possible to accurately judge the health status of each sub-region. When identifying potential pest and disease areas, the camera focus is automatically adjusted to focus on details and shorten the focal length, thereby improving the accuracy of capturing early signs of pests and diseases. Through intelligent and dynamically adjusted monitoring methods, it is ensured that pests and diseases can be identified and intervened in time, prevent the spread of pests and diseases, and ensure crop health, so as to solve the problems in the above-mentioned background technology.

[0006] In order to achieve the above object, the present invention provides the following technical solution: an intelligent soilless cultivation environment monitoring method based on big data, comprising the following steps: The soilless cultivation environment area is divided into several sub-areas of equal size. The area of ​​each sub-area is set according to the specific cultivation environment and monitoring requirements to ensure that the plants in each sub-area are representative. In each sub-area, the camera automatically collects image data of plant leaves in the sub-area and transmits it to the data processing system in real time for subsequent pest and disease detection and analysis; Summarize the image data collected from each sub-area, extract the features reflecting potential pests and diseases from the summary, analyze the extracted features, input the analyzed features into a pre-trained convolutional neural network, and predict the pests and diseases in the sub-area through the pre-trained convolutional neural network; Based on the prediction results of the convolutional neural network, the health status of each sub-area is classified into "potential pest and disease area" and "normal growth area"; For normal growth areas, continue to maintain the initial status for pest and disease monitoring; For the sub-area identified as a potential pest and disease area, the camera control module automatically adjusts the focus of the camera so that the camera focuses on the sub-area. After the camera is focused, it will capture the details of the plants in the sub-area more accurately. After the camera focuses on the potential pest and disease area, the current focal length of the camera is shortened according to the prediction results of the convolutional neural network to capture the details of the target area more clearly and identify pests and diseases.

[0007] Preferably, the soilless cultivation environment area is divided into several sub-areas of equal size, and the specific steps are as follows: It is necessary to clarify the total area of ​​the monitoring area of ​​the entire soilless cultivation environment and the area ratio required for dividing the sub-areas. The size of each sub-area should be determined according to the shape of the environment, the layout of the cultivation system, the type of plants, and the utilization requirements of the cultivation area. After determining the area of ​​the sub-region, select the region division method to divide the entire cultivation environment area into a number of small squares according to an equal grid, and each small square is a sub-region.

[0008] Preferably, features reflecting potential pests and diseases are extracted from the summary, wherein the extracted features include the degree of contrast between leaf veins and surrounding leaf tissues and the progressive color deviation of leaf color. After analyzing the degree of contrast between leaf veins and surrounding leaf tissues and the progressive color deviation of leaf color in each sub-region, a vein highlighting reference value and a color deviation progressive reference value are generated respectively. The intensity of the contrast between leaf veins and surrounding leaf tissues is quantified by the vein highlighting reference value, and the degree of progressive deviation of leaf color in time or space is quantified by the color deviation progressive reference value.

[0009] Preferably, the vein highlighting reference value and the color deviation progressive reference value generated after analyzing the contrast between the leaf veins and the surrounding leaf tissues and the progressive color deviation change of the leaf color are input into a pre-trained convolutional neural network, and a pest and disease index is generated based on the convolutional neural network, and the pest and disease index is used to predict the pests and diseases of the sub-area.

[0010] Preferably, the specific steps of analyzing the contrast between the leaf veins and the surrounding leaf tissues in each sub-region and generating the leaf vein highlight reference value are as follows: The plant images of each sub-region are processed, and the spectral reflectance characteristics of the leaves are extracted using multispectral imaging technology. The spectral contrast matrix between the vein area and the surrounding leaf tissue is constructed through the spectral contrast value. The construction expression of the spectral contrast matrix is: ,in: The pixel points of the leaf veins and surrounding leaf tissues The spectral contrast value, Pixel The reflectance value of the near infrared band in the vein area, Pixels The reflectance value of the green light band in the tissue around the leaf veins, The near-infrared band, It is the green light band; Based on the spectral contrast matrix, the position weight function and dynamic threshold function are introduced to normalize the matrix and calculate the reference value of leaf vein highlighting. The calculation expression is: ,in: It is the reference value for leaf vein highlighting, used to quantify the intensity of leaf vein contrast in the entire sub-region. H , W are the height and width of the image, is the position weight function, which weights the pixel points according to the importance of the leaf area. Assign different weights, is a dynamic threshold function used to remove the background noise of normal leaf vein contrast, defined as: ,in Pixel The distance to the center of the blade, and is a parameter used to control the threshold attenuation.

[0011] Preferably, the specific steps of analyzing the progressive color deviation change of the leaf color in each sub-region and generating the progressive color deviation reference value are as follows: By performing multi-dimensional color analysis on the leaf images in each sub-region, a color shift feature matrix is ​​extracted. The color shift feature matrix is ​​based on the leaf's position at time t and space The color change is calculated, and the calculation expression is: ,in, is the color shift feature matrix, indicating the spatial position at time t The degree of color shift, Indicated in wavelength Lower space position The color reflectance value, Indicates the last time point The color reflectance value, is the band weight coefficient; The color shift feature matrix and spatial shift weight are used to calculate the color shift gradual reference value of the sub-region and quantify the degree of gradual shift of the leaf color in time or space. The calculation expression is: ,in, is the color shift gradual reference value, which quantifies the overall gradual shift degree of leaf color in the sub-region. The spatial shift weight is set based on the leaf density and the spatial concentration of color shift in the sub-region. is the smoothing factor to prevent The denominator is zero.

[0012] Preferably, the pest index generated when predicting the pests and diseases of the sub-regions by the pre-trained convolutional neural network is compared and analyzed with the pre-set pest index reference threshold, and the health status of each sub-region is classified. The specific classification is as follows: If the pest index generated under a sub-region is greater than the pest index reference threshold, the health status of the sub-region is classified as a potential pest area; If the pest index generated under the sub-region is less than or equal to the pest index reference threshold, the health status of the sub-region is classified as a normal growth area.

[0013] Preferably, for the sub-area identified as a potential pest area, automatic focusing is achieved through the control module of the camera, specifically: First, the pest index generated by the convolutional neural network determines the location coordinates of the target sub-area, and the coordinates are used as the target point of the camera; Next, the camera's autofocus function is used in combination with a depth calculation algorithm to dynamically adjust the camera's focal plane so that the camera can accurately focus on the target sub-area.

[0014] Preferably, after the camera focuses on the potential pest area, the current focal length of the camera is shortened according to the result predicted by the convolutional neural network. The specific steps are as follows: The pest index generated by the convolutional neural network in the sub-area is collected, and the pest index is compared with the preset pest index reference threshold. According to the relationship between the two, the focal length difference factor is calculated. The calculation expression is: ,in, is the focal length difference factor, which is used to calculate the adjusted focal length later. represents the pest index of the sub-region predicted by the convolutional neural network. is the reference threshold of pest index, is the adjustment factor used to quantify the effect of the degree to which the pest index exceeds the reference threshold of the pest index on the shortening of the focal length; According to the current focal length and the calculated focal length difference factor, the adjusted target focal length is calculated and ensured to be within the focal length adjustment range of the camera. The expression for target focal length adjustment is: ,in, The adjusted target focal length is used to achieve clearer image capture. is the current focal length of the camera, and is the physical adjustment range of the camera focal length, ensuring that the calculated target focal length does not exceed the camera hardware capability, where: and They are the maximum and minimum focal lengths of the camera respectively; The camera’s focal length is automatically adjusted based on the calculated target focal length. The focus result is verified in real time through the image clarity index. If the focus effect does not reach the expected setting value, it is fine-tuned again based on the adjustment range. The clarity index formula is as follows: ,in, is the image clarity index, which indicates the clarity of the image in the target area. Current focal length F Lower pixel The image intensity value at is the image gradient, which is used to quantify the change in pixel intensity, N is the total number of pixels in the target area.

[0015] Intelligent soilless cultivation environment monitoring system based on big data, including sub-area division module, image data acquisition module, feature extraction and pest prediction module, health status classification module, normal area monitoring module, potential pest focus module and focal length optimization and fine recognition module; The sub-area division module divides the soilless cultivation environment area into several sub-areas of equal size. The area of ​​each sub-area is set according to the specific cultivation environment and monitoring requirements to ensure that the plants in each sub-area are representative. The image data acquisition module automatically collects the image data of the plant leaves in each sub-area through the camera, and transmits it to the data processing system in real time for subsequent pest and disease detection and analysis; The feature extraction and pest prediction module summarizes the image data collected from each sub-area, extracts the features reflecting potential pests and diseases from the summary, analyzes the extracted features, inputs the analyzed features into a pre-trained convolutional neural network, and predicts the pests and diseases of the sub-area through the pre-trained convolutional neural network; The health status classification module classifies the health status of each sub-area into "potential pest and disease area" and "normal growth area" based on the prediction results of the convolutional neural network; Normal area monitoring module, for normal growth areas, continues to maintain the initial state for pest and disease monitoring; The potential pest and disease focusing module automatically adjusts the focus of the camera through the camera control module for the sub-area identified as the potential pest and disease area, so that the camera focuses on the sub-area. After the camera is focused, it will capture the details of the plants in the sub-area more accurately. The focal length optimization and fine recognition module shortens the current focal length of the camera based on the prediction results of the convolutional neural network after the camera focuses on the potential pest and disease area, so as to capture the details of the target area more clearly and identify pests and diseases.

[0016] In the above technical solution, the technical effects and advantages provided by the present invention are: The present invention divides the cultivation area into multiple sub-areas and collects image data of each sub-area, ensuring that the plant features in each sub-area are representative, thereby improving the accuracy of pest and disease monitoring. Secondly, the convolutional neural network is used to extract features and predict pests and diseases from the collected image data, which can not only accurately judge the health status of each sub-area, but also automatically adjust the focus of the camera when a potential pest and disease area is found, focus on details and shorten the focal length, so that the early signs of pests and diseases can be clearly captured. This intelligent and dynamically adjusted monitoring method greatly improves the accuracy of pest and disease identification, can timely discover potential pest and disease risks, ensure that pests and diseases are intervened in the early stage, thereby effectively preventing the spread of pests and diseases and ensuring the health of crops in the entire cultivation environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0018] Figure 1 The present invention is a method flow chart of the intelligent soilless cultivation environment monitoring method based on big data.

[0019] Figure 2 It is a module schematic diagram of the intelligent soilless cultivation environment monitoring system based on big data of the present invention. DETAILED DESCRIPTION

[0020] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0021] The present invention provides Figure 1 The intelligent soilless cultivation environment monitoring method based on big data shown includes the following steps: The soilless cultivation environment area is divided into several sub-areas of equal size. The area of ​​each sub-area is set according to the specific cultivation environment and monitoring requirements to ensure that the plants in each sub-area are representative. Divide the soilless cultivation environment area into several sub-areas according to the size of the area. The specific steps are as follows: It is necessary to clarify the total area of ​​the monitoring area of ​​the entire soilless cultivation environment and the area ratio required for dividing the sub-areas. The size of each sub-area is determined based on the shape of the environment, the layout of the cultivation system, the type of plants, and the utilization requirements of the cultivation area. The total area can be obtained by measuring or obtaining a layout map of the environment, and then the standard area of ​​each sub-area can be calculated based on the requirements.

[0022] The purpose of this step is to determine the area of ​​each sub-area to ensure that the divided sub-areas can cover every part of the cultivation area. The appropriate sub-area size not only helps the efficiency of pest and disease monitoring, but also ensures the rational allocation and management of resources during the monitoring process, avoiding too large a monitoring area that makes it difficult to detect pest and disease signs, or too small a monitoring area that causes too much redundant data.

[0023] After determining the area of ​​the sub-region, choose a reasonable area division method. Common division methods include regular grid division, uniform area division, or division by crop type / growth stage. If the regular grid division method is used, the area of ​​the entire cultivation environment is divided into several small squares according to an equal grid, and each small square is a sub-region. If it is divided according to the growth stage of the crop, it can be adjusted according to the growth density and state of the plant to divide the area that needs to be focused on.

[0024] The purpose of this step is to ensure that the cultivation area is reasonably and evenly divided into multiple sub-areas so that each sub-area can be monitored independently. Through different division methods, customized regional divisions can be carried out according to different cultivation environments, crop types, cultivation stages and other factors. This process can ensure that the monitoring content of each sub-area is representative, avoid excessive or uneven division, and thus ensure the comprehensiveness and accuracy of subsequent pest and disease identification.

[0025] When monitoring the intelligent soilless cultivation environment, the area of ​​the sub-area is not fixed, but should be flexibly set according to the characteristics of the actual cultivation environment and the monitoring needs. Different cultivation environments may have different scales, layouts, and crop types, so the area of ​​each sub-area needs to be adjusted according to these factors. For example, if the cultivation environment is large, it may be necessary to divide the area into multiple smaller sub-areas to ensure that each sub-area can be effectively monitored; while in a smaller cultivation environment, it may be divided into fewer sub-areas. At the same time, the accuracy requirements of monitoring also affect the area of ​​the sub-area. If higher monitoring accuracy is required, each sub-area may need to be divided smaller so that the plant leaves in each area can be observed and analyzed more carefully. Conversely, if the monitoring requirements are low, the sub-area can be appropriately expanded, thereby reducing the number of monitoring equipment and resource consumption.

[0026] By subdividing a large monitoring area into multiple smaller monitoring units, the limitation of not being able to clearly identify pests and diseases from a single perspective can be avoided. By dividing the sub-areas, plants in different areas can be independently monitored and analyzed, improving the accuracy of pest and disease monitoring. This subdivision can effectively solve the monitoring ambiguity caused by a single perspective and ensure that signs of pests and diseases in each area can be discovered in a timely manner.

[0027] In each sub-area, the camera automatically collects image data of plant leaves in the sub-area and transmits it to the data processing system in real time for subsequent pest and disease detection and analysis; This link is the basis of pest and disease detection. The camera captures images in real time and provides raw data for subsequent feature extraction and analysis. By collecting images at a high frequency, subtle changes in plant leaves can be captured more accurately, especially in the early stages of pests and diseases.

[0028] Summarize the image data collected from each sub-area, extract the features reflecting potential pests and diseases from the summary, analyze the extracted features, input the analyzed features into a pre-trained convolutional neural network, and predict the pests and diseases in the sub-area through the pre-trained convolutional neural network; Features reflecting potential diseases and insect pests are extracted from the summarized images, wherein the extracted features include the contrast between the leaf veins and the surrounding leaf tissues and the gradual color deviation of the leaf color. After analyzing the contrast between the leaf veins and the surrounding leaf tissues and the gradual color deviation of the leaf color in each sub-region, a vein highlighting reference value and a color deviation gradual reference value are generated respectively. The vein highlighting reference value is used to quantify the intensity of the contrast between the leaf veins and the surrounding leaf tissues, and the clarity, color contrast and texture difference of the leaf veins relative to the surrounding leaf tissues in the image are measured to reflect whether the plant has abnormally prominent vascular bundles or degeneration of surrounding mesophyll tissues due to diseases or stress. The color deviation gradual reference value is used to quantify the degree of gradual deviation of leaf color in time or space, and the subtle change trajectory of leaf color (such as gradual transition from normal green to yellow-green, brown, red, etc.) is analyzed to reflect whether the plant is affected by diseases and insect pests or environmental stress, resulting in chlorophyll degradation or abnormal accumulation of pigments.

[0029] When the contrast between the veins in a sub-region and the surrounding leaf tissue is large, it usually indicates that there may be potential risks of pests and diseases in the sub-region. The large contrast of veins may be because pathogens or pests have destroyed the normal structure of leaf tissue, causing the surrounding mesophyll tissue to degenerate, lose green or necrosis, making the veins appear more prominent. This phenomenon may reflect that the plant has been attacked by pathogens, viruses or pests, such as mesophyll cell collapse caused by fungal infection, chlorophyll degradation caused by viral infection, or local tissue damage caused by pests. This feature is the key to the early identification of pests and diseases. By quantifying the contrast of veins, it can provide preliminary risk warnings for the monitoring system, facilitating timely intervention and preventing the spread of pests and diseases.

[0030] The specific steps for generating the reference value of leaf vein prominence after analyzing the contrast between the leaf veins and the surrounding leaf tissues in each sub-region are as follows: The plant images of each sub-region are processed, and the spectral reflectance characteristics of the leaves are extracted using multispectral imaging technology. The spectral contrast matrix between the vein area and the surrounding leaf tissue is constructed through the spectral contrast value. The spectral contrast matrix is ​​based on the difference between each pixel in a specific band (such as near infrared and green light ) is calculated by the reflectance ratio, and the spectral contrast matrix is ​​constructed as follows: ,in: The pixel points of the leaf veins and surrounding leaf tissues The spectral contrast value, Pixel The reflectance value of the near infrared band in the vein area, Pixels The reflectance value of the green light band in the tissue around the leaf veins, It is a near-infrared band and is used to detect the moisture content or structural integrity of plant internal tissues. Green light band, used to measure the distribution and health status of chlorophyll; The main purpose of this step is to construct a basic matrix that can quantify the contrast intensity between leaf veins and surrounding leaf tissues through the differences in spectral characteristics. By selecting the band, abnormal changes in leaf veins and surrounding tissues caused by pests or stress can be more sensitively captured.

[0031] Based on the spectral contrast matrix, the position weight function and dynamic threshold function are introduced to normalize the matrix and calculate the reference value of leaf vein highlighting. The calculation expression is: ,in: It is the reference value for leaf vein highlighting, used to quantify the intensity of leaf vein contrast in the entire sub-region. H , W are the height and width of the image, is the position weight function, which weights the pixel points according to the importance of the leaf area. Assign different weights (for example, the weight of the center area of ​​the leaf is greater than that of the edge area). is a dynamic threshold function used to remove the background noise of normal leaf vein contrast, defined as: ,in Pixel The distance to the center of the blade, and is a parameter used to control threshold attenuation; This step generates a global reference value for vein highlighting by comprehensively considering the spectral contrast between leaf veins and surrounding tissues and the importance of leaf areas. The introduction of position weight function and dynamic threshold function helps to eliminate interference from normal contrast areas and focus on abnormally prominent vein areas, further improving the accuracy and reliability of pest and disease detection.

[0032] It can be seen from the vein highlight reference value that the larger the performance value of the vein highlight reference value generated after analyzing the contrast between the veins and the surrounding leaf tissues in each sub-region, the stronger the contrast between the veins and the surrounding leaf tissues in the sub-region, which usually means that the leaves may be more seriously affected by pests and diseases or other stresses, thus reflecting the greater possibility of hidden dangers of pests and diseases in the sub-region. On the contrary, when the vein highlight reference value is small, it indicates that the contrast between the veins and the surrounding leaf tissues is low, and the physiological structure and color changes of the leaves are more normal, indicating that the possibility of hidden dangers of pests and diseases in the sub-region is smaller. The vein highlight reference value provides an important basis for early warning of pests and diseases in the sub-region by quantifying the contrast degree, which helps the system to locate problem areas in a timely manner and take intervention measures.

[0033] When the color of leaves in a sub-region shows a gradual color shift, it usually indicates that there may be potential hidden dangers of pests and diseases in the sub-region. Gradual color shift refers to the process in which the color of leaves gradually transitions from normal green to abnormal tones such as yellow-green, brown, gray, and red. This phenomenon is usually an early manifestation of plants being subjected to external stress (such as pathogen infection, pest invasion, or adverse environmental conditions), and may reflect the degradation of chlorophyll, abnormal cell metabolism, or abnormal accumulation of secondary pigments (such as carotenoids and anthocyanins). By analyzing the gradual change trajectory of leaf color, the early abnormal reactions of plants can be captured before pests and diseases have spread significantly, thereby providing a basis for timely diagnosis and intervention to avoid the expansion of hidden dangers.

[0034] The specific steps of analyzing the progressive color shift changes of the leaves in each sub-region and generating the progressive color shift reference value are as follows: By performing multi-dimensional color analysis on the leaf images in each sub-region, a color shift feature matrix is ​​extracted. The color shift feature matrix is ​​based on the leaf's position at time t and space The color change is calculated, focusing on reflecting the gradual change trend of color in different bands. The calculation expression is: ,in, is the color shift feature matrix, indicating the spatial position at time t The degree of color shift, Indicated in wavelength Lower space position The color reflectance value can be collected by multispectral imaging in the visible light (400-700 nm) and near infrared (700-1100 nm) bands. Indicates the last time point The color reflectance value, is the band weight coefficient, which is used to emphasize the sensitivity of specific bands (such as red light and near infrared) to pests and diseases; By calculating the time shift of the color in each band, the color shift feature matrix It reflects the cumulative effect of leaf color changes at different locations and time points. The focus is on extracting the trend characteristics of the gradual change of leaf color from green to yellow, brown, red, etc., thus laying the foundation for subsequent index generation.

[0035] The color shift feature matrix and spatial shift weight are used to calculate the color shift gradual reference value of the sub-region and quantify the degree of gradual shift of the leaf color in time or space. The calculation expression is: ,in, It is the reference value of color shift progression, which quantifies the overall degree of gradual shift of leaf color in the sub-region. The higher the value, the more obvious the color change. The spatial offset weight is set based on the spatial concentration of leaf density and color offset in the sub-region. The higher the weight, the greater the impact of the location on the overall index. It can be calculated by leaf coverage and image texture changes. No specific restrictions are made here. is the smoothing factor to prevent The denominator is zero when This step comprehensively considers the color changes in the time dimension and the weight distribution in space. By accumulating the color shifts in the sub-areas point by point, an index reflecting the overall change trend is generated to determine whether there are abnormal gradual color changes in the area (such as from green to yellow or brown), thereby identifying potential pest and disease risks and helping to quickly locate high-risk areas for in-depth analysis.

[0036] It can be seen from the color deviation progressive reference value that the larger the performance value of the color deviation progressive reference value generated after analyzing the progressive color deviation change of the leaf color in each sub-region, the more significant the progressive color deviation change of the leaf color in the sub-region has occurred, such as a gradual shift from green to yellow, brown or other abnormal colors. This progressive color deviation is usually caused by the early physiological response of plants under the stress of pests and diseases, such as chlorophyll degradation, abnormal cell metabolism, or pigment accumulation (such as carotenoids and anthocyanins). Therefore, when the color deviation progressive reference value is large, it means that there may be a high risk of pests and diseases in the area; on the contrary, when the color deviation progressive reference value is small, it indicates that the color change of the leaves is small or close to normal, and the risk of hidden dangers is low. Therefore, the color deviation progressive reference value can be used as an important basis for judging the health status of the region, helping to achieve early detection and precise intervention of pests and diseases.

[0037] The vein highlighting reference value and color deviation gradual reference value generated after analyzing the contrast between the leaf veins and surrounding leaf tissues and the gradual color deviation change of the leaf color are input into the pre-trained convolutional neural network, and the pest and disease index is generated based on the convolutional neural network. The pest and disease index is used to predict the pests and diseases in the sub-area.

[0038] A pre-trained convolutional neural network (CNN) refers to a deep learning model that has been preliminarily learned and optimized based on a large amount of relevant data. This convolutional neural network has been trained with a large number of historical plant pest and disease images and related features (such as reference values ​​for leaf vein highlighting and reference values ​​for color gradient), and can accurately identify and classify the health status of plants and the types of pests and diseases. During the training process, the convolutional neural network extracts local patterns of image or feature data through multi-layer convolution operations, such as changes in leaf vein texture and the trend of leaf color deviation; reduces the dimension of the data through the pooling layer to retain key information; and finally integrates these features through the fully connected layer to predict and quantify the risk of pests and diseases. Pre-training means that the model already has a high generalization ability for specific tasks (such as plant pest and disease identification) and can be directly applied to the prediction of new data without the need to train from scratch.

[0039] In the task of disease and insect pest prediction, such a pre-trained model has two important functions: first, through multi-level feature extraction, the input leaf vein highlighting reference value and color deviation gradual reference value and other information are integrated to capture the hidden patterns that may be related to diseases and insect pests. For example, the convolutional neural network can detect the nonlinear relationship between abnormal leaf vein highlighting and color deviation, which may indicate the risk of viral infection, fungal invasion or pest gnawing; second, with the help of pre-trained weights, the model can be quickly applied to disease and insect pest detection in different cultivation areas without retraining, thus saving computing resources and time. In addition, since the convolutional neural network has learned the characteristic patterns of multiple types of diseases and insect pests during the training process, it can adapt to different plant varieties, disease types and environmental conditions, improving the accuracy and robustness of the prediction of the disease and insect pest index.

[0040] In this scenario, the input of the pre-trained convolutional neural network includes two main features: the reference value of leaf vein prominence and the reference value of color deviation. These indices reflect the contrast between the leaf veins and the surrounding leaf tissues and the gradual change trend of leaf color, respectively, and can be used to capture potential early signs of pests and diseases. The convolutional neural network automatically learns the complex associations between these features and the risk of pests and diseases by performing multi-dimensional analysis on these features. For example, when the reference value of color deviation is high and a specific pattern appears in the reference value of leaf vein prominence, the convolutional neural network may identify this as a high-risk sign of a fungal disease (such as rust). For different combinations of indices, the convolutional neural network can also predict the possible type and severity of pests and diseases based on the pre-trained weights, and generate a pest and disease index to quantify the level of hidden dangers of pests and diseases in the sub-area.

[0041] Specifically, the convolutional neural network will extract key patterns from the reference values ​​for leaf vein highlighting and color gradient based on the weights learned during training, and the pooling layer will further reduce the dimension to remove irrelevant information while retaining the significant parts of the features. In the fully connected layer, the model will match the extracted features with the pest risk pattern and output the final pest index. This pre-training-based approach can maximize the use of existing pest images and feature data, thereby achieving efficient and accurate pest prediction and providing strong support for the intelligent management of soilless cultivation environments.

[0042] The above convolutional neural network is not specifically limited here, and can realize the reference value of highlighting leaf veins and color cast gradient reference value Conduct comprehensive analysis to generate pest and disease index In order to realize the technical solution of the present invention, the present invention provides a specific implementation method; the pest index The resulting calculation formula is: , where , Reference values ​​for leaf vein prominence and color cast gradient reference value The preset scaling factor of , Both are greater than 0.

[0043] It can be seen from the pest and disease index that the greater the expression value of the leaf vein highlighting reference value generated after analyzing the contrast between the leaf veins and the surrounding leaf tissues in each sub-region, the greater the expression value of the color deviation progressive reference value generated after analyzing the progressive color deviation change of the leaf color in each sub-region, that is, the greater the expression value of the pest and disease index generated when predicting the pests and diseases in the sub-region through the pre-trained convolutional neural network, the greater the hidden danger of pest and disease invasion in the sub-region, and vice versa.

[0044] Based on the prediction results of the convolutional neural network, the health status of each sub-area is classified into "potential pest and disease area" and "normal growth area"; The pest index generated by predicting pests and diseases in sub-regions through the pre-trained convolutional neural network is compared and analyzed with the pre-set pest index reference threshold to classify the health status of each sub-region. The specific classification is as follows: If the pest index generated under a sub-region is greater than the pest index reference threshold, the health status of the sub-region is classified as a potential pest area; Potential pest and disease areas refer to sub-areas where the pest and disease index generated by convolutional neural network analysis exceeds the preset pest and disease index reference threshold. This classification indicates that the health of plants in this sub-area may have been threatened by early pests and diseases, such as tiny lesions on leaves, gradual changes in color abnormalities, prominent vein structure, or other possible signs of pests and diseases. Although the impact of pests and diseases at this stage may not be obvious as large-scale damage to crops, it has the potential to spread and requires focus and timely further intervention measures.

[0045] If the pest index generated under the sub-region is less than or equal to the pest index reference threshold, the health status of the sub-region is classified as a normal growth area.

[0046] The normal growth area refers to the sub-area where the pest index generated by the convolutional neural network analysis is less than or equal to the reference threshold of the pest index. This classification indicates that the plants in this area are in good health, no obvious early signs of pests or potential threats are found, the color and texture of the leaves are consistent with the normal state, and there are no abnormal progressive color deviations or increased contrast of leaf veins.

[0047] This link automatically distinguishes healthy and potential pest and disease areas through the prediction results of the convolutional neural network, thereby reducing indiscriminate monitoring of all areas. Through intelligent classification, the system can focus resources on areas with potential pest and disease risks for priority monitoring and intervention, while regular monitoring is performed on normal areas, avoiding the waste of resources caused by excessive monitoring and improving monitoring efficiency.

[0048] For normal growth areas, continue to maintain the initial status for pest and disease monitoring; For sub-areas classified as normal growth areas, since the plants are in good health and no early signs or potential risks of pests and diseases are detected, there is no need to adjust the monitoring method or frequency, and the initial monitoring settings can be continued. By maintaining the initial status monitoring, basic monitoring of normal areas can be ensured while avoiding unnecessary waste of resources, so that more system computing resources and management focus can be concentrated on high-risk potential pest and disease areas, achieving optimization of monitoring efficiency and precise management of the cultivation environment.

[0049] For the sub-area identified as a potential pest and disease area, the camera control module automatically adjusts the focus of the camera so that the camera focuses on the sub-area. After the camera is focused, it will capture the details of the plants in the sub-area more accurately. For sub-areas identified as potential pest and disease areas, automatic focusing is achieved through the camera's control module. Specifically, the pest and disease index generated by the convolutional neural network first determines the position coordinates of the target sub-area, and the system uses the coordinates as the target point of the camera. Then, using the camera's autofocus function, combined with a depth calculation algorithm (such as measuring the contrast peak through an image clarity assessment algorithm), the camera's focal plane is dynamically adjusted to accurately focus on the target sub-area. In this way, the camera is able to capture high-resolution images of the target area, further refine the observation and analysis of early signs of pests and diseases, and provide more detailed visual data support for precise intervention.

[0050] After the camera focuses on the potential pest area, the current focal length of the camera is shortened according to the prediction results of the convolutional neural network to capture the details of the target area more clearly and identify pests and diseases; After the camera focuses on the potential pest area, the current focal length of the camera is shortened according to the prediction results of the convolutional neural network. The specific steps are as follows: The pest index generated by the convolutional neural network in the sub-area is collected, and the pest index is compared with the preset pest index reference threshold. According to the relationship between the two, the focal length difference factor is calculated. The calculation expression is: ,in, is the focal length difference factor, which is used to calculate the adjusted focal length later. It indicates the pest index of the sub-region predicted by the convolutional neural network, reflecting the strength of the potential pest risks in the sub-region. is the reference threshold of the pest index, indicating the boundary between the normal growth area and the potential pest area. is an adjustment factor used to quantify the effect of the degree to which the pest index exceeds the reference threshold of the pest index on the shortening of the focal length.

[0051] By calculating the extent to which the pest index exceeds the pest index reference threshold, the magnitude of the focal length adjustment is obtained. When the pest index is greater than the pest index reference threshold, it is determined that the current focal length needs to be shortened to capture finer image features.

[0052] According to the current focal length and the calculated focal length difference factor, the adjusted target focal length is calculated and ensured to be within the focal length adjustment range of the camera. The expression for target focal length adjustment is: ,in, The adjusted target focal length is used to achieve clearer image capture. is the current focal length of the camera, and is the physical adjustment range of the camera focal length, ensuring that the calculated target focal length does not exceed the camera hardware capability, where: and They are the maximum and minimum focal lengths of the camera respectively; This step ensures that when the focus is adjusted based on the pest index, the target focus is always within the range allowed by the camera hardware. By dynamically calculating the target focus, the system can achieve more precise focus on the target area, improving image resolution and detail expression.

[0053] The camera’s focal length is automatically adjusted based on the calculated target focal length. The focus result is verified in real time through the image clarity index. If the focus effect does not reach the expected setting value, it is fine-tuned again based on the adjustment range. The clarity index formula is as follows: ,in, is the image clarity index, which indicates the clarity of the image in the target area. Current focal length F (This focal length is the target focal length after adjustment) The image intensity value at is the image gradient, which is used to quantify the change in pixel intensity, N is the total number of pixels in the target area; By adjusting the focus to the target focal length , the image details captured by the camera will be clearer. At the same time, based on the image clarity index The focus effect is verified in real time to ensure the image quality is at the best level. If the verification result does not meet expectations, the system can further optimize the focus adjustment process to achieve efficient and accurate monitoring of potential pest and disease areas.

[0054] Shortening the focal length can provide higher image resolution, making subtle signs of pests and diseases (such as tiny spots, slight leaf discoloration, etc.) clearer, helping the system to accurately identify pests and diseases. Shortening the focal length is an effective means to improve the accuracy of pest and disease identification. After the focal length is shortened, the camera can capture the details of the target area more clearly, especially in the early stages when pests and diseases are not yet obvious. This operation enables the system to detect signs of pests and diseases from small changes. Through this refined monitoring, timely intervention can be carried out at the early stage of pest and disease occurrence, effectively preventing the spread of pests and diseases and ensuring the health of crops in the entire cultivation area.

[0055] Through the above-mentioned intelligent soilless cultivation environment monitoring method based on big data, the problem that a single-view camera in the prior art cannot accurately identify the early signs of pests and diseases can be effectively solved. First, by dividing the cultivation area into multiple sub-areas and collecting image data of each sub-area, it can be ensured that the plant characteristics in each sub-area are representative, thereby improving the accuracy of pest and disease monitoring. Secondly, the convolutional neural network is used to extract features and predict pests and diseases from the collected image data, which can not only accurately judge the health status of each sub-area, but also automatically adjust the focus of the camera when a potential pest and disease area is found, focus on details and shorten the focal length, so that the early signs of pests and diseases can be clearly captured. This intelligent and dynamically adjusted monitoring method greatly improves the accuracy of pest and disease identification, can timely discover pest and disease hidden dangers, ensure that pests and diseases are intervened in the early stage, thereby effectively preventing the spread of pests and diseases and ensuring the health of crops in the entire cultivation environment.

[0056] The present invention provides Figure 2 The intelligent soilless cultivation environment monitoring system based on big data shown includes a sub-area division module, an image data acquisition module, a feature extraction and pest prediction module, a health status classification module, a normal area monitoring module, a potential pest focus module, and a focal length optimization and fine recognition module; The sub-area division module divides the soilless cultivation environment area into several sub-areas of equal size. The area of ​​each sub-area is set according to the specific cultivation environment and monitoring requirements to ensure that the plants in each sub-area are representative. The image data acquisition module automatically collects the image data of the plant leaves in each sub-area through the camera, and transmits it to the data processing system in real time for subsequent pest and disease detection and analysis; The feature extraction and pest prediction module summarizes the image data collected from each sub-area, extracts the features reflecting potential pests and diseases from the summary, analyzes the extracted features, inputs the analyzed features into a pre-trained convolutional neural network, and predicts the pests and diseases of the sub-area through the pre-trained convolutional neural network; The health status classification module classifies the health status of each sub-area into "potential pest and disease area" and "normal growth area" based on the prediction results of the convolutional neural network; Normal area monitoring module, for normal growth areas, continues to maintain the initial state for pest and disease monitoring; The potential pest and disease focusing module automatically adjusts the focus of the camera through the camera control module for the sub-area identified as the potential pest and disease area, so that the camera focuses on the sub-area. After the camera is focused, it will capture the details of the plants in the sub-area more accurately. The focal length optimization and fine identification module shortens the current focal length of the camera based on the prediction results of the convolutional neural network after the camera focuses on the potential pest area, so as to capture the details of the target area more clearly and identify pests and diseases; The intelligent soilless cultivation environment monitoring method based on big data provided in the embodiment of the present invention is realized by the above-mentioned intelligent soilless cultivation environment monitoring system based on big data. The specific methods and processes of the intelligent soilless cultivation environment monitoring system based on big data are detailed in the embodiment of the above-mentioned intelligent soilless cultivation environment monitoring method based on big data, which will not be repeated here.

[0057] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0058] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0059] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. An intelligent soilless cultivation environment monitoring method based on big data, characterized in that: The following steps are involved: The soilless cultivation environment area is divided into several sub-areas of equal size. The area of ​​each sub-area is set according to the specific cultivation environment and monitoring requirements to ensure that the plants in each sub-area are representative. In each sub-area, the camera automatically collects image data of plant leaves in the sub-area and transmits it to the data processing system in real time for subsequent pest and disease detection and analysis; Summarize the image data collected from each sub-area, extract the features reflecting potential pests and diseases from the summary, analyze the extracted features, input the analyzed features into a pre-trained convolutional neural network, and predict the pests and diseases in the sub-area through the pre-trained convolutional neural network; Based on the prediction results of the convolutional neural network, the health status of each sub-area is classified into "potential pest and disease area" and "normal growth area"; For normal growth areas, the initial state is maintained for pest and disease monitoring; for sub-areas identified as potential pest and disease areas, the camera control module automatically adjusts the focus of the camera so that the camera focuses on the sub-area. After the camera is focused, it will capture the details of the plants in the sub-area more accurately. After the camera focuses on the potential pest and disease area, the current focal length of the camera is shortened according to the prediction results of the convolutional neural network to capture the details of the target area more clearly and identify pests and diseases.

2. The intelligent soilless cultivation environment monitoring method based on big data according to claim 1 is characterized in that: Divide the soilless cultivation environment area into several sub-areas according to the size of the area. The specific steps are as follows: It is necessary to clarify the total area of ​​the monitoring area of ​​the entire soilless cultivation environment and the area ratio required for dividing the sub-areas. The size of each sub-area should be determined according to the shape of the environment, the layout of the cultivation system, the type of plants, and the utilization requirements of the cultivation area. After determining the area of ​​the sub-region, select the region division method to divide the entire cultivation environment area into a number of small squares according to an equal grid, and each small square is a sub-region.

3. The intelligent soilless cultivation environment monitoring method based on big data according to claim 1 is characterized in that: Features reflecting potential diseases and insect pests are extracted from the summary, wherein the extracted features include the contrast between leaf veins and surrounding leaf tissues and the gradual color deviation of leaf color. After analyzing the contrast between leaf veins and surrounding leaf tissues and the gradual color deviation of leaf color in each sub-region, vein highlighting reference values ​​and color deviation gradual reference values ​​are generated respectively. The intensity of the contrast between leaf veins and surrounding leaf tissues is quantified by the vein highlighting reference value, and the degree of gradual deviation of leaf color in time or space is quantified by the color deviation gradual reference value.

4. The intelligent soilless cultivation environment monitoring method based on big data according to claim 3 is characterized in that: The vein highlighting reference value and color deviation gradual reference value generated after analyzing the contrast between the leaf veins and surrounding leaf tissues and the gradual color deviation change of the leaf color are input into the pre-trained convolutional neural network, and the pest and disease index is generated based on the convolutional neural network. The pest and disease index is used to predict the pests and diseases in the sub-area.

5. The intelligent soilless cultivation environment monitoring method based on big data according to claim 4 is characterized in that: The specific steps for generating the reference value of leaf vein prominence after analyzing the contrast between the leaf veins and the surrounding leaf tissues in each sub-region are as follows: The plant images of each sub-region are processed, and the spectral reflectance characteristics of the leaves are extracted using multispectral imaging technology. The spectral contrast matrix between the vein area and the surrounding leaf tissue is constructed through the spectral contrast value. The construction expression of the spectral contrast matrix is: ,in: The pixel points of the leaf veins and surrounding leaf tissues The spectral contrast value, Pixel The reflectance value of the near infrared band in the vein area, Pixels The reflectance value of the green light band in the tissue around the leaf veins, The near-infrared band, It is the green light band; Based on the spectral contrast matrix, the position weight function and dynamic threshold function are introduced to normalize the matrix and calculate the reference value of leaf vein highlighting. The calculation expression is: ,in: It is the reference value for leaf vein highlighting, used to quantify the intensity of leaf vein contrast in the entire sub-region. H , W are the height and width of the image, is the position weight function, which weights the pixel points according to the importance of the leaf area. Assign different weights, is a dynamic threshold function used to remove the background noise of normal leaf vein contrast, defined as: ,in Pixel The distance to the center of the blade, and is a parameter used to control the threshold attenuation.

6. The intelligent soilless cultivation environment monitoring method based on big data according to claim 3 is characterized in that: The specific steps of analyzing the progressive color deviation changes of the leaves in each sub-region and generating the progressive color deviation reference value are as follows: By performing multi-dimensional color analysis on the leaf images in each sub-region, a color shift feature matrix is ​​extracted. The color shift feature matrix is ​​based on the leaf's position at time t and space The color change is calculated, and the calculation expression is: ,in, is the color shift feature matrix, indicating the spatial position at time t The degree of color shift, Indicated in wavelength Lower space position The color reflectance value, Indicates the last time point The color reflectance value, is the band weight coefficient; The color shift feature matrix and spatial shift weight are used to calculate the color shift gradual reference value of the sub-region and quantify the degree of gradual shift of the leaf color in time or space. The calculation expression is: ,in, is the color shift gradual reference value, which quantifies the overall gradual shift degree of leaf color in the sub-region. The spatial shift weight is set based on the leaf density and the spatial concentration of color shift in the sub-region. is the smoothing factor to prevent The denominator is zero.

7. The intelligent soilless cultivation environment monitoring method based on big data according to claim 5 is characterized in that: The pest index generated by predicting pests and diseases in sub-regions through the pre-trained convolutional neural network is compared and analyzed with the pre-set pest index reference threshold to classify the health status of each sub-region. The specific classification is as follows: If the pest index generated under a sub-region is greater than the pest index reference threshold, the health status of the sub-region is classified as a potential pest area; If the pest index generated under the sub-region is less than or equal to the pest index reference threshold, the health status of the sub-region is classified as a normal growth area.

8. The intelligent soilless cultivation environment monitoring method based on big data according to claim 4 is characterized in that: For sub-areas identified as potential pest and disease areas, automatic focusing is achieved through the camera control module, specifically: First, the pest index generated by the convolutional neural network determines the location coordinates of the target sub-area, and the coordinates are used as the target point of the camera; Next, the camera's autofocus function is used in combination with a depth calculation algorithm to dynamically adjust the camera's focal plane so that the camera can accurately focus on the target sub-area.

9. The intelligent soilless cultivation environment monitoring method based on big data according to claim 7 is characterized in that: After the camera focuses on the potential pest area, the current focal length of the camera is shortened according to the prediction results of the convolutional neural network. The specific steps are as follows: The pest index generated by the convolutional neural network in the sub-area is collected, and the pest index is compared with the preset pest index reference threshold. According to the relationship between the two, the focal length difference factor is calculated. The calculation expression is: ,in, is the focal length difference factor, which is used to calculate the adjusted focal length later. represents the pest index of the sub-region predicted by the convolutional neural network. is the reference threshold of pest index, is the adjustment factor used to quantify the effect of the degree to which the pest index exceeds the reference threshold of the pest index on the shortening of the focal length; According to the current focal length and the calculated focal length difference factor, the adjusted target focal length is calculated and ensured to be within the focal length adjustment range of the camera. The expression for target focal length adjustment is: ,in, The adjusted target focal length is used to achieve clearer image capture. is the current focal length of the camera, and is the physical adjustment range of the camera focal length, ensuring that the calculated target focal length does not exceed the camera hardware capability, where: and They are the maximum and minimum focal lengths of the camera respectively; The camera’s focal length is automatically adjusted based on the calculated target focal length. The focus result is verified in real time through the image clarity index. If the focus effect does not reach the expected setting value, it is fine-tuned again based on the adjustment range. The clarity index formula is as follows: ,in, is the image clarity index, which indicates the clarity of the image in the target area. Current focal length F Lower pixel The image intensity value at is the image gradient, which is used to quantify the change in pixel intensity, N is the total number of pixels in the target area.

10. An intelligent soilless cultivation environment monitoring system based on big data, used to implement the intelligent soilless cultivation environment monitoring method based on big data as described in any one of claims 1 to 9, characterized in that: It includes sub-area division module, image data acquisition module, feature extraction and pest prediction module, health status classification module, normal area monitoring module, potential pest focus module, and focal length optimization and fine recognition module; The sub-area division module divides the soilless cultivation environment area into several sub-areas of equal size. The area of ​​each sub-area is set according to the specific cultivation environment and monitoring requirements to ensure that the plants in each sub-area are representative. The image data acquisition module automatically collects the image data of the plant leaves in each sub-area through the camera, and transmits it to the data processing system in real time for subsequent pest and disease detection and analysis; The feature extraction and pest prediction module summarizes the image data collected from each sub-area, extracts the features reflecting potential pests and diseases from the summary, analyzes the extracted features, inputs the analyzed features into a pre-trained convolutional neural network, and predicts the pests and diseases of the sub-area through the pre-trained convolutional neural network; The health status classification module classifies the health status of each sub-area into "potential pest and disease area" and "normal growth area" based on the prediction results of the convolutional neural network; Normal area monitoring module, for normal growth areas, continues to maintain the initial state for pest and disease monitoring; The potential pest and disease focusing module automatically adjusts the focus of the camera through the camera control module for the sub-area identified as the potential pest and disease area, so that the camera focuses on the sub-area. After the camera is focused, it will capture the details of the plants in the sub-area more accurately. The focal length optimization and fine recognition module shortens the current focal length of the camera based on the prediction results of the convolutional neural network after the camera focuses on the potential pest and disease area, so as to capture the details of the target area more clearly and identify pests and diseases.

Images