Plant health condition assessment method and system based on AI model, and medium
Through the multimodal data evaluation method based on AI model, the problem of insufficient accuracy and robustness of traditional evaluation methods is solved, and the accurate assessment of plant health status and disease identification is achieved, providing higher accuracy and convenience .
Patent Information
- Application Number
- CN202411578249.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-07
- Publication Date
- 2024-12-03
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional plant health assessment methods rely on single modal data, are time-consuming and labor-intensive, have limited accuracy and robustness, cannot fully reflect plant health status, and are susceptible to noise, making it difficult to accurately identify plant pest and disease types.
Using an AI model-based plant health assessment method, the contrast-enhanced pretreatment and denoising treatment are performed by acquiring multimodal data (image, spectral, meteorological and soil data), and combining multimodal deep learning models to comprehensively evaluate plant health. situation.
Accurate assessment of plant health status is achieved, evaluation errors are reduced, accuracy and convenience are improved, and the health information and disease status of plants can be captured more comprehensively.
Smart Images

Figure CN119066527A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of intelligent crop identification, and more specifically, to a plant health status assessment method, system and medium based on an AI model. Background Art
[0002] The health status of plants directly affects agricultural production and the ecological environment. Traditional plant health assessment methods mainly rely on manual observation and single-modal data (such as images or spectral data). These methods are not only time-consuming and labor-intensive, but also have limited accuracy and robustness, and cannot fully reflect the health status of plants. In addition, plant health is affected by many factors, such as pests and diseases, nutritional status, environmental pressure, etc., and single-modal data is difficult to fully capture this complex information; in addition, existing assessment methods such as ResNet models are affected by noise, and image and other data processing may not accurately obtain the detailed status of plant health, such as the inability to obtain more detailed plant pest and disease types, etc., so that it is impossible to accurately assess the health status of plants. Summary of the invention
[0003] The present application is provided to solve the above-mentioned defects existing in the prior art. A plant health status assessment method, system and medium based on an AI model are needed, which can comprehensively assess the health status of plants based on multimodal data, remove the influence of noise, and accurately obtain the health status of plants.
[0004] In a first aspect of the present application, a plant health assessment method based on an AI model is provided, the plant health assessment method comprising: obtaining multimodal data of a plant to be tested, the multimodal data comprising at least image data, spectral data, meteorological data and soil data; performing contrast enhancement preprocessing on the image data, the contrast enhancement preprocessing comprising: performing segmentation processing on each image in the image data respectively to obtain each image block corresponding to each image; obtaining the grayscale distribution corresponding to each image block of each image according to a preset grayscale and pixel value distribution of the image block; obtaining the grayscale distribution of each image based on the grayscale distribution corresponding to each image block of each image; performing equalization processing on each image based on the grayscale distribution of each image to obtain image data after contrast enhancement preprocessing; inputting the image data after contrast enhancement preprocessing as well as the spectral data, meteorological data and soil data into a multimodal deep learning model to obtain an assessment result of the plant health status.
[0005] In a second aspect of the present application, a plant health assessment system based on an AI model is provided, the plant health assessment system comprising a data preprocessing module and an evaluation module, the preprocessing module being configured to: obtain multimodal data of a plant to be tested, the multimodal data comprising at least image data, spectral data, meteorological data and soil data; perform contrast enhancement preprocessing on the image data, the contrast enhancement preprocessing comprising: segmenting each image in the image data respectively to obtain each image block corresponding to each image; obtaining the grayscale distribution corresponding to each image block of each image according to a preset grayscale and pixel value distribution of the image block; obtaining the grayscale distribution of each image based on the grayscale distribution corresponding to each image block of each image; performing equalization processing on each image based on the grayscale distribution of each image to obtain image data after contrast enhancement preprocessing; the evaluation module being configured to: obtain an assessment result of plant health status based on the image data after contrast enhancement preprocessing as well as the spectral data, meteorological data and soil data.
[0006] The third aspect of the present application provides a non-temporary computer-readable medium having instructions stored thereon, which, when executed by a processor, execute the steps of the plant health assessment method based on the AI model as described in any embodiment of the present application.
[0007] The AI model-based plant health assessment method, system and medium provided in each embodiment of the present application acquire multimodal data of image data, spectral data, meteorological data and soil data, and then preprocess the image block data through a preprocessing module. Through the pixel processing process of each image block in the image, the influence of noise is removed, so that the image can be clearer and the abnormal conditions in the image can be better highlighted, which facilitates the accurate assessment of the results of plant health status, reduces the error of the assessment, and brings accuracy and convenience to crop management. BRIEF DESCRIPTION OF THE DRAWINGS
[0008] In the drawings, which are not necessarily drawn to scale, the same reference numerals may describe similar parts in different views. The same reference numerals with letter suffixes or different letter suffixes may represent different instances of similar parts. The accompanying drawings generally illustrate various embodiments by way of example and not limitation, and together with the specification and claims, are used to illustrate the embodiments applied for. When appropriate, the same reference numerals are used throughout the drawings to refer to the same or similar parts. Such embodiments are illustrative and are not intended to be exhaustive or exclusive embodiments of the present apparatus or method.
[0009] Figure 1 A flow chart showing a method for assessing plant health status based on an AI model according to an embodiment of the present application; Figure 2 A schematic diagram showing the structure of a plant health status assessment system based on an AI model according to an embodiment of the present application; Figure 3 A comparison chart of the accuracy of each prediction method is shown. DETAILED DESCRIPTION
[0010] In order to enable those skilled in the art to better understand the technical solution of the present application, the present application is described in detail below in conjunction with the accompanying drawings and specific implementation examples. The embodiments of the present application are further described in detail below in conjunction with the accompanying drawings and specific implementation examples, but are not intended to limit the present application.
[0011] The words "first", "second" and similar words used in this application do not indicate any order, quantity or importance, but are only used to distinguish. The words "include" or "comprise" and similar words mean that the elements before the word include the elements listed after the word, and do not exclude the possibility of including other elements.
[0012] According to an embodiment of the present application, a plant health status assessment method based on an AI model is provided. Figure 1 As shown, the plant health assessment method includes the following steps. In step 101, multimodal data of the plant to be tested is obtained, and the multimodal data includes at least image data, spectral data, meteorological data and soil data. In step 102, the image data is subjected to contrast enhancement preprocessing, and the contrast enhancement preprocessing includes the following steps. In step 1021, each image in the image data is segmented to obtain each image block corresponding to each image. In step 1022, according to the preset grayscale and pixel value distribution of the image block, the grayscale distribution corresponding to each image block of each image is obtained. In step 1023, based on the grayscale distribution corresponding to each image block of each image, the grayscale distribution of each image is obtained. In step 1024, based on the grayscale distribution of each image, each image is equalized to obtain image data after contrast enhancement preprocessing. In step 103, the image data after contrast enhancement preprocessing, as well as the spectral data, meteorological data and soil data are input into the multimodal deep learning model to obtain the assessment result of the plant health status.
[0013] The image data may include image information of plants obtained by RGB cameras, near infrared cameras and / or multi-spectral cameras. It can be expressed as: . Represents an image.
[0014] The spectral data may be the spectral reflectance characteristics of plants obtained using a hyperspectral instrument. It can be expressed as: in, It is Spectral curve collected.
[0015] Weather data includes local weather information, such as temperature ,humidity , light intensity , rainfall Etc., constitute the meteorological data set .
[0016] Soil data including soil nutrients , pH value, water content Etc., constitute the soil data set .
[0017] The image data in the multimodal data can obtain the image information of the plant, and the color and shape of the plant can be obtained through the image information; the spectral data obtained by the hyperspectral instrument can obtain the spectral reflectance characteristics of the plant, and then obtain information such as the growth status of the plant. The temperature and other environmental factors of the plant can be obtained through meteorological data. The soil data can be used to obtain the soil in which the plant lives. Therefore, a more comprehensive and integrated analysis of the health status of the plant can be performed through multimodal data, so that the health status of the plant can be obtained more accurately.
[0018] Due to the plant growth environment and other reasons, the acquired image data may not clearly reflect the plant's condition. For example, due to the influence of the temperature and water mist in the greenhouse, the image may not be clear, so the analysis of the plant's condition may not be accurate. Contrast enhancement preprocessing can better highlight abnormal conditions in the image.
[0019] In the process of contrast enhancement preprocessing, the image is first divided into image blocks, and then according to the pixel value of each image block and the pre-set grayscale, it is possible to obtain which grayscale each image block belongs to, and then obtain the grayscale distribution of each image. In some embodiments, according to the pixel value of each image block and the pre-set grayscale, the local histogram of the image can be obtained. In this way, the local histogram information corresponding to each image block can be obtained. In some embodiments, the cumulative distribution function is calculated based on the local histogram corresponding to each image.
[0020] Equalization processing is performed based on the grayscale distribution of the image. In some embodiments, the grayscale value of each pixel in each image block of the image is mapped to a new grayscale value using the formula shown in formula (1).
[0021] (1) Where L represents the total number of preset gray levels, M×M represents the size of each pixel block, and CDF local (0) represents the cumulative distribution function value corresponding to gray level 0, CDF local ( ) represents the cumulative distribution function value corresponding to the original gray value of the image block, Represents the grayscale value obtained after mapping. The grayscale value of each pixel in each image block is mapped to a new grayscale value after being equalized by the local histogram of the image block.
[0022] In this way, the grayscale value of each image block is changed according to the grayscale distribution of the entire image, so that the contrast of the entire image is enhanced.
[0023] In some embodiments, the plant health status assessment method further includes performing denoising processing on the image data, and the denoising processing includes: performing denoising processing on the image using a non-local mean method.
[0024] The non-local mean is calculated according to equations (2) and (3).
[0025] (2) in, So Centered search window, is the weight, defined as: (3) in, and They are respectively and The neighborhood window centered at It is a parameter that controls the degree of attenuation. De-noising can avoid errors in plant health analysis due to noise.
[0026] In some embodiments, the plant health status assessment method further includes preprocessing the spectral data, and the preprocessing of the spectral data includes: denoising the spectral data to obtain denoised spectral data; normalizing the denoised spectral data to obtain normalized spectral data.
[0027] In some embodiments, a Savitzky-Golay filter is used to smooth the spectral curve to obtain denoised spectral data.
[0028] In some embodiments, the spectral data is normalized based on equation (4).
[0029] (4) in, is the wavelength The reflectivity under are the minimum and maximum values of the spectral curve.
[0030] In some embodiments, the image data after contrast enhancement preprocessing as well as spectral data, meteorological data and soil data are input into a multimodal deep learning model to obtain an assessment result of plant health status, including: extracting image features of image data using an image feature extraction network; extracting spectral features of spectral data using a spectral feature extraction network; performing feature fusion on the image features, the spectral features, meteorological data and soil data to obtain fused features; and inputting the fused features into an output module to obtain an assessment result of plant health.
[0031] In some embodiments, the image feature extraction network may adopt a convolutional neural network (CNN); the spectral feature extraction network may adopt a one-dimensional convolutional neural network (1D-CNN).
[0032] In some embodiments, the meteorological data and soil data are preprocessed using a fully connected layer to obtain preprocessed meteorological data and soil data, and then feature fusion is performed.
[0033] Multimodal data based on plant image data, spectral data, meteorological data and soil data can more comprehensively analyze the health of plants and more accurately obtain the health, sub-health and various disease states of plants. The fusion of multimodal data can comprehensively capture plant health information and reduce the limitations of a single data source.
[0034] The health status assessment results include: health level: such as "healthy", "sub-healthy", "disease", etc.; possible types of pests and diseases: identify possible pests and diseases based on model output; nutritional status assessment: determine whether the plant has nutritional deficiencies or excesses; environmental pressure: assess the impact of meteorological and soil conditions on plants.
[0035] In some embodiments, the image features, the spectral features, the meteorological data and the soil data are feature fused by using a vector splicing method. For example, the feature dimension of the image data is 2048, the feature dimension of the spectral data is 256, the feature dimension of the meteorological data is 128, and the feature dimension of the soil data is 128. The feature dimension of the fused features obtained after vector splicing is 2048.
[0036] In some embodiments, the output module includes at least two fully connected layers, and the fusion features are input into each fully connected layer to obtain the output data of each fully connected layer; the output data of each fully connected layer is input into the output layer, and the output layer uses the output probability distribution function based on the output data of each fully connected layer to obtain the probability of each health status category, and selects the health status category with the largest probability as the evaluation result. The health status category may include health, sub-health, pests and diseases, lack of water and other states.
[0037] The fused feature vector After further processing by several fully connected layers, the final output is the assessment result of plant health status The processing process is as follows: in, and are the weights and biases of the first fully connected layer; and are the weights and biases of the output layer, is the number of health status categories; is the activation function, is the output probability distribution function.
[0038] In some embodiments, the multimodal deep learning model is trained using a cross entropy loss function. The cross entropy loss function is shown in formula (5): (5) in, is the number of samples, It is The true labels of samples, is the probability of each health status category assessed by the model. The base of is 10.
[0039] The samples include the collected plant image data, spectral data, meteorological data and soil data. The plant health status corresponding to each sample can be annotated in advance to obtain the label of the real health status of each sample. , Indicates the probability of each category of "health", "sub-health" and "disease".
[0040] In some embodiments, the Adam optimization algorithm is used to adjust the model parameters of the multimodal deep learning model.
[0041] In some embodiments, the plant health status assessment method also includes: obtaining the plant health status at the current moment and multiple time points before the current moment; the plant health trend prediction model selects an action according to the latest strategy, and predicts the plant health status at the next time point based on the plant health status at the current moment and multiple time points before the current moment.
[0042] The interval between the current time point and the next time point or the previous time point can be preset, such as three hours, three days, etc. Based on the latest health trend forecast, it is helpful to adjust agricultural management measures in a timely manner, such as fertilization, irrigation, pest control, etc.
[0043] The multiple time points before the current moment correspond to the plant health status at each time point within at least one month before the current moment, and the time interval between two adjacent time points may be 4 hours, one day, or the like.
[0044] In some embodiments, the plant health trend prediction model is trained online, and the online training method includes: The real-time plant health status output by the multimodal deep learning model is stored in the experience playback pool; The parameters of the multimodal deep learning model are used as the initial values of the parameters of the plant health trend prediction model; Based on the sample data of the experience revisit pool, the plant health trend prediction model is trained online using a deep reinforcement learning algorithm.
[0045] The online training process includes: at the current time point, predicting the plant health status at the predicted time point according to the current strategy selection action to obtain the predicted plant health status; the deep reinforcement learning algorithm obtains the latest experience samples based on the experience replay pool, and the latest experience samples include the real-time plant health status output by the newly generated multimodal deep learning model between the current time point and the predicted time point, and uses the latest experience samples to perform small batch training on the plant health trend prediction model; using the degree of match between the actual plant health status and the predicted plant health status, the plant health trend prediction model is fed back and rewarded, and the feedback results are used to update the model parameters.
[0046] The degree of matching is measured according to the reward value of the calculated reward output. The greater the reward value, the higher the degree of matching.
[0047] Use deep reinforcement learning algorithms to build online prediction models so that the prediction models can be optimized in real time based on newly generated data to improve the accuracy of predictions.
[0048] The choices of actions for the plant health trend prediction model are as follows: in is the output of the deep reinforcement learning algorithm, are the network parameters of the deep reinforcement learning algorithm, is the health status of the plant at the current moment, and A is the set of actions. The output of the deep reinforcement learning algorithm is the result of action selection. This way of selecting actions is to explore the pros and cons of different actions (implicit reward size) during model training and improve model performance.
[0049] By introducing the online learning mechanism of reinforcement learning, the model can process new data in real time, respond quickly to changes in the environment and plant conditions, reduce prediction errors caused by outdated data, provide more accurate and timely health trend forecasts, and help agricultural managers make more effective decisions.
[0050] In some embodiments, rewarding the plant health trend prediction model includes: inputting the prediction result and the actual result of the plant health trend prediction model into a cross entropy function, and then using the result of the cross entropy function as a reward to optimize the parameters of the plant health trend prediction model, and the reward is shown in formula (6): (6) in, is the cross entropy loss function is the actual plant health status; Indicates the performance improvement of the model under new data, is a weight coefficient that measures the effect of online learning. Rewards are given based on the accuracy and real-time performance of the plant health trend prediction model. Refers to the difference in loss of the model before and after parameter update, The calculation process is as follows: 1. Loss before model update : Use the current model parameters (parameters before updating) to predict the new data and get the prediction results With the true value The loss between: 2. Loss after model update : Use the new data to learn or update the model online to obtain updated model parameters. Use the updated model parameters to predict the same new data to obtain new prediction results. Calculate the updated prediction results With the true value The loss between: 3. Computing performance improvement Δ: The difference in loss before and after the update is used as a measure of performance improvement: The new data in step 2 are the newly stored samples in the experience replay pool after the model is updated.
[0051] According to an embodiment of the present application, a plant health status assessment system based on an AI model is also provided. Figure 2 As shown, the plant health assessment system 200 includes a data preprocessing module 201 and an assessment module 202, wherein the preprocessing module 201 is configured to: obtain multimodal data of the plant to be tested, wherein the multimodal data includes at least image data and spectral data; perform contrast enhancement preprocessing on the image data, wherein the contrast enhancement preprocessing includes: segmenting each image in the image data to obtain each image block corresponding to each image; obtaining the grayscale distribution corresponding to each image block of each image according to the preset grayscale and pixel value distribution of the image block; obtaining the grayscale distribution of each image based on the grayscale distribution corresponding to each image block of each image; performing equalization processing on each image based on the grayscale distribution of each image to obtain image data after contrast enhancement preprocessing; and the assessment module 202 is configured to: obtain the assessment result of the plant health status based on the image data after contrast enhancement preprocessing as well as the spectral data, meteorological data and soil data.
[0052] In some embodiments, the pre-processing module 201 is further configured to: perform denoising processing on the image data, and the denoising processing includes: performing denoising processing on the image using a non-local mean method.
[0053] In some embodiments, the plant health status assessment system also includes a prediction module 202, which obtains the plant health status at the current moment and at least two time points before the current moment; the plant health trend prediction model selects actions according to the latest strategy, and predicts the plant health status at the next time point based on the plant health status at the current moment and at least two time points before the current moment.
[0054] According to an embodiment of the present application, a non-temporary computer-readable medium is also provided, on which instructions are stored. When executed by a processor, the instructions execute the steps of the plant health status assessment method based on the AI model as described in any embodiment of the present application.
[0055] Using data sets including different crops (including image, spectral, meteorological, and soil data), the plant health status was evaluated using the plant health status assessment method of the present application and the ResNet model (with image data as input), the support vector machine (with spectral data as input), and the multimodal MLP fusion model (with image, spectral, meteorological, and soil data as input). The results are shown in Table 1.
[0056] Table 1 Evaluation results of different methods method Accuracy Accuracy Recall F1-Score ResNet Model (Image) 0.852 0.835 0.801 0.817 Support Vector Machine (Spectral) 0.784 0.76 0.735 0.747 Existing multimodal MLP fusion model 0.885 0.862 0.84 0.85 This application multimodal AI model 0.947 0.938 0.965 0.930 It can be seen from Table 1 that for different crops, the plant health status assessment method of the present application can improve the accuracy, precision, recall rate and F1-Score compared with the traditional single-modal method and multi-modal method.
[0057] The plant health trend prediction model, ResNet model, support vector machine, and multimodal MLP fusion model of the present application predict the plant health status of multimodal data at 10 consecutive time points. The time interval between adjacent time points is 6 hours, and the accuracy of each time point is obtained by comparing it with the actual plant health status. Figure 3 As shown, the accuracy of each algorithm in predicting plant health status at different time points is shown. The experimental results show that the plant health trend prediction model proposed in this application not only has the highest accuracy, but also shows a continuous upward trend in prediction accuracy at different time points. In contrast, the prediction accuracy of other algorithms such as ResNet model (image), support vector machine (spectrum) and existing multimodal MLP fusion model shows a fluctuating trend, which further verifies the effectiveness of the online learning method proposed in this application.
[0058] In addition, although exemplary embodiments have been described herein, the scope includes any and all embodiments based on the present application with equivalent elements, modifications, omissions, combinations (e.g., various embodiments intersecting schemes), adaptations or changes. The elements in the claims are to be interpreted broadly based on the language adopted in the claims, and are not limited to the examples described in this specification or during the practice of this application, and the examples are to be interpreted as non-exclusive. Therefore, this specification and examples are intended to be considered as examples only, and the true scope and spirit are indicated by the following claims and the full scope of their equivalents.
[0059] The above description is intended to be illustrative rather than restrictive. For example, the above examples (or one or more schemes thereof) may be used in combination with each other. For example, a person of ordinary skill in the art may use other embodiments when reading the above description. In addition, in the above-mentioned specific embodiments, various features may be grouped together to simplify the present application. This should not be interpreted as an intention that a feature of an application that does not require protection is necessary for any claim. On the contrary, the subject matter of the present application may be less than all the features of an embodiment of a particular application. Thus, the following claims are incorporated herein in the specific embodiments as examples or embodiments, wherein each claim is independently a separate embodiment, and it is considered that these embodiments may be combined with each other in various combinations or arrangements. The scope of the present invention should be determined with reference to the appended claims and the full scope of equivalent forms granted by these claims.
[0060] The above embodiments are only exemplary embodiments of the present application and are not intended to limit the present invention. The protection scope of the present invention is defined by the claims. Those skilled in the art may make various modifications or equivalent substitutions to the present invention within the essence and protection scope of the present application, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present invention.
Claims
1. A plant health status assessment method based on AI model, It is characterized in that The plant health status assessment method comprises: Acquire multimodal data of the plant to be tested, wherein the multimodal data at least includes image data, spectral data, meteorological data and soil data; The image data is subjected to contrast enhancement preprocessing, wherein the contrast enhancement preprocessing includes: segmenting each image in the image data to obtain each image block corresponding to each image; obtaining the grayscale distribution corresponding to each image block of each image according to the preset grayscale and pixel value distribution of the image block; obtaining the grayscale distribution of each image based on the grayscale distribution corresponding to each image block of each image; and performing equalization processing on each image based on the grayscale distribution of each image to obtain image data after contrast enhancement preprocessing; The image data after contrast enhancement preprocessing as well as spectral data, meteorological data and soil data are input into a multimodal deep learning model to obtain an assessment result of plant health status.
2. The plant health status assessment method according to claim 1, It is characterized in that The plant health status assessment method further includes performing denoising processing on the image data, and the denoising processing includes: performing denoising processing on the image using a non-local mean method.
3. The plant health status assessment method according to claim 1, It is characterized in that The plant health status assessment method also includes preprocessing the spectral data, and the preprocessing of the spectral data includes: denoising the spectral data to obtain denoised spectral data; normalizing the denoised spectral data to obtain normalized spectral data.
4. The plant health status assessment method according to claim 1, It is characterized in that The image data after contrast enhancement preprocessing as well as spectral data, meteorological data and soil data are input into a multimodal deep learning model to obtain an assessment result of plant health status, including: extracting image features of image data using an image feature extraction network; extracting spectral features of spectral data using a spectral feature extraction network; performing feature fusion on the image features, the spectral features, meteorological data and soil data to obtain fused features; and inputting the fused features into an output module to obtain an assessment result of plant health.
5. The plant health status assessment method according to claim 1, It is characterized in that The plant health status assessment method also includes: obtaining the plant health status at the current moment and multiple time points before the current moment; the plant health trend prediction model selects an action according to the latest strategy, and predicts the plant health status at the next time point based on the plant health status at the current moment and the plant health status before the current moment.
6. The method for evaluating plant health status according to claim 1, It is characterized in that The multimodal deep learning model is trained using a cross entropy loss function.
7. A plant health status assessment system based on AI model, It is characterized in that The plant health status assessment system comprises a data preprocessing module and an assessment module, wherein the preprocessing module is configured to: obtain multimodal data of the plant to be tested, wherein the multimodal data comprises at least image data, spectral data, meteorological data and soil data; The image data is subjected to contrast enhancement preprocessing, wherein the contrast enhancement preprocessing includes: segmenting each image in the image data to obtain each image block corresponding to each image; obtaining the grayscale distribution corresponding to each image block of each image according to the preset grayscale and pixel value distribution of the image block; obtaining the grayscale distribution of each image based on the grayscale distribution corresponding to each image block of each image; and performing equalization processing on each image based on the grayscale distribution of each image to obtain image data after contrast enhancement preprocessing; The evaluation module is configured to obtain an evaluation result of plant health status based on the image data after contrast enhancement preprocessing as well as spectral data, meteorological data and soil data.
8. The plant health status assessment system according to claim 7, It is characterized in that The preprocessing module is further configured to: perform denoising processing on the image data, and the denoising processing includes: performing denoising processing on the image using a non-local mean method.
9. The plant health status assessment system according to claim 7, It is characterized in that The plant health status assessment system also includes a prediction module, which is configured to: obtain the plant health status at the current moment and multiple time points before the current moment; the plant health trend prediction model selects actions according to the latest strategy, and predicts the plant health status at the next time point based on the plant health status at the current moment and the plant health status before the current moment.
10. A non-transitory computer-readable medium having instructions stored thereon, which, when executed by a processor, execute the steps of the plant health status assessment method based on an AI model as described in any one of claims 1 to 6.
Citation Information
Patent Citations
Coral distribution and health condition evaluation method based on deep clustering analysis
CN110889844A
Data acquisition and analysis method for AI identification of green plant growth situation
CN117036088A
Crop disease and pest detection method and system based on multi-modal deep learning
CN118258769A
Crop multi-modal data fusion driving method
CN118627008A
Multi-modal perception crop disease and insect pest intelligent identification and precise early warning system
CN118658077A
Cited By
Plant state evaluation method and system based on multi-modal thinking chain technology
CN120747108A
