Cold region saline-alkali soil rice growth assessment method and system based on image data fusion

Abstract

The invention discloses a cold region saline-alkali soil rice growth assessment method and system based on image data fusion, and relates to the technical field of rice growth monitoring. According to the main technical scheme, the method comprises the steps of obtaining multi-source fusion image data, environment monitoring data and actually measured growth indexes corresponding to different growth stages of rice in a cold region saline-alkali soil region in a historical period; performing deep feature extraction on the multi-source fusion image data to obtain key image features; performing deep feature extraction on the environment monitoring data to obtain key environment features; and respectively processing the key image features and the key environment features corresponding to different growth stages in a historical period by using an attention mechanism, and constructing a growth evaluation model corresponding to the rice in the cold region saline-alkali soil region based on the processed key image features, the processed key environment features and actually measured growth indexes. The main purpose is to improve the accuracy of rice growth evaluation of the cold region saline-alkali soil.

Description

Technical Field

[0001] The present application relates to the technical field of rice growth monitoring, and in particular to a method and system for evaluating rice growth in cold saline-alkali land based on image data fusion. Background Art

[0002] With global climate change and increasingly tight land resources, agricultural production faces many challenges, especially in special environments such as cold saline-alkali land. Cold saline-alkali land is mainly soda saline-alkali land, with heavy soil texture and poor water retention capacity, which puts higher requirements on crop growth. As one of the most important food crops in the world, rice is particularly important for achieving stable and efficient production in these special environments. The development of precision agriculture technology provides a new idea for solving this problem. By combining advanced imaging technology and data analysis methods, it can more accurately evaluate the growth status of crops, optimize resource allocation, and improve yield and quality.

[0003] At present, the technology for evaluating the growth of rice in saline-alkali land in cold regions mainly relies on a single data source (such as ground sensors or satellite remote sensing images), or is limited to monitoring environmental factors (such as temperature, humidity, soil Ph value, and soil salinity). However, the information provided by a single data source is limited, and it is difficult to fully reflect the actual growth status of rice in such a complex environment as saline-alkali land in cold regions. In addition, the identification of deep characteristics is not in-depth and comprehensive enough, resulting in large errors in the growth evaluation of rice in saline-alkali land in cold regions, making it difficult to provide reliable decision support. Summary of the invention

[0004] In view of the above problems, the present application provides a method and system for evaluating the growth of rice in saline-alkali land in cold regions based on image data fusion. The main purpose is to improve the accuracy of the evaluation of the growth of rice in saline-alkali land in cold regions to provide reliable decision support.

[0005] In order to solve the above technical problems, this application proposes the following solutions: In a first aspect, the present application provides a method for accurately evaluating rice growth in saline-alkali land in cold regions based on image data fusion, the method comprising: Obtain multi-source fusion image data, environmental monitoring data, and measured growth indicators corresponding to different growth stages of rice in cold saline-alkali land areas in historical periods; Performing deep feature extraction on the multi-source fusion image data according to the measured growth index to obtain key image features corresponding to different growth stages of rice in the cold saline-alkali land area in historical periods; Performing deep feature extraction on the environmental monitoring data based on the multi-source image data to obtain key environmental features corresponding to different growth stages of rice in cold saline-alkali land areas in historical periods; The attention mechanism is used to process the key image features and the key environmental features corresponding to different growth stages in the historical period respectively, and a growth assessment model corresponding to rice in the cold saline-alkali land area is constructed based on the processed key image features, the processed key environmental features and the measured growth indicators, wherein the growth assessment model is used to output the predicted growth indicators corresponding to different growth stages of rice in the cold saline-alkali land area in the future period after inputting the processed current key image features and the processed current key environmental features corresponding to different growth stages in the current period.

[0006] In a second aspect, the present application provides a cold-region saline-alkali land rice growth assessment system based on image data fusion, the system comprising: An acquisition unit is used to obtain multi-source fusion image data, environmental monitoring data and measured growth indicators corresponding to different growth stages of rice in cold saline-alkali land areas in historical periods; A first extraction unit is used to perform deep feature extraction on the multi-source fusion image data according to the measured growth index to obtain key image features corresponding to different growth stages of rice in the cold saline-alkali land area in historical periods; A second extraction unit is used to perform deep feature extraction on the environmental monitoring data according to the multi-source image data to obtain key environmental features corresponding to different growth stages of rice in the cold saline-alkali land area in historical periods; A construction unit is used to use an attention mechanism to process the key image features and the key environmental features corresponding to different growth stages in the historical period respectively, and to construct a growth assessment model corresponding to rice in the cold saline-alkali land area based on the processed key image features, the processed key environmental features and the measured growth indicators, wherein the growth assessment model is used to output predicted growth indicators corresponding to different growth stages of rice in the cold saline-alkali land area in the future period after inputting the processed current key image features and the processed current key environmental features corresponding to different growth stages in the current period.

[0007] In order to achieve the above-mentioned purpose, according to the third aspect of the present application, a storage medium is provided, which includes a stored program, wherein when the program is running, the device where the storage medium is located is controlled to execute the cold-region saline-alkali land rice growth assessment method based on image data fusion according to the above-mentioned first aspect.

[0008] In order to achieve the above-mentioned purpose, according to the fourth aspect of the present application, a processor is provided, and the processor is used to run a program, wherein when the program is running, the method for evaluating rice growth in cold saline-alkali land based on image data fusion of the above-mentioned first aspect is executed.

[0009] By means of the above technical scheme, the present application provides a method and system for evaluating the growth of rice in cold saline-alkali land based on image data fusion. First, multi-source fused image data, environmental monitoring data and measured growth indicators corresponding to different growth stages of rice in cold saline-alkali land area in historical period are obtained. Then, key image features corresponding to different growth stages of rice in cold saline-alkali land area in historical period are extracted from multi-source fused image data according to measured growth indicators. Then, key environmental features corresponding to different growth stages of rice in cold saline-alkali land area in historical period are extracted from environmental monitoring data according to multi-source image data. Finally, the attention mechanism is used to process the key image features and key environmental features corresponding to different growth stages in historical period respectively. Based on the processed key image features, the processed key environmental features and the measured growth indicators, a growth evaluation model corresponding to rice in cold saline-alkali land area is constructed. After inputting the processed current key image features and the processed current key environmental features corresponding to different growth stages in the current period, the predicted growth indicators corresponding to different growth stages of rice in cold saline-alkali land area in future period are output. The technical solution provided by the present application ensures that the model can obtain comprehensive information input by obtaining multi-source fusion image data, environmental monitoring data and measured growth indicators corresponding to different growth stages of rice in cold saline-alkali land areas in historical periods. The multi-source fusion image data not only improves the richness and comprehensiveness of the data, but also enhances the model's adaptability to complex environmental changes in cold saline-alkali land areas. By extracting deep key image features through measured growth indicators, image features that have a direct impact on rice growth status, yield, etc. can be screened out, ensuring that the model can focus on truly important information and reduce the interference of redundant features. By extracting deep key environmental features from environmental monitoring data, we can indirectly capture the complex impact of environmental factors on rice growth, analyze the mechanism of action of environmental factors in more detail, and provide a more comprehensive environmental adaptability assessment. The environment of cold saline-alkali land is changeable, and image features can reflect these changes in real time, enabling the model to better adapt to different environmental conditions. At the same time, the introduction of an adaptive attention mechanism to process key features enhances the model's attention to more important features among the key features, further improving the accuracy of the growth assessment model for evaluating rice growth in cold saline-alkali land, thereby providing users with reliable decision support.

[0010] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings: Figure 1 A flow chart of a method for evaluating rice growth in saline-alkali land in cold regions based on image data fusion provided in an embodiment of the present application is shown; Figure 2 A block diagram of a cold-region saline-alkali land rice growth assessment system based on image data fusion provided in an embodiment of the present application is shown; Figure 3 A block diagram showing the composition of another cold-region saline-alkali land rice growth assessment system based on image data fusion provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0013] The embodiment of the present application provides a method for evaluating the growth of rice in saline-alkali land in cold regions based on image data fusion. The method can improve the accuracy of evaluating the growth of rice in saline-alkali land in cold regions to provide reliable decision support. The specific execution steps are as follows: Figure 1 As shown, including: 101. Obtain multi-source fusion image data, environmental monitoring data and measured growth indicators corresponding to different growth stages of rice in cold saline-alkali land areas in historical periods.

[0014] In this step, the cold saline-alkali land area is the experimental field for growing rice in this embodiment.

[0015] The historical period can be specifically the rice growing season within the last three or five years. The multi-source fusion image data is obtained by fusion of panoramic images, drone low-altitude images, IoT real-time images, and local close-up images of artificial plants according to timestamp registration. Panoramic images are obtained by regular shooting with a 360-degree camera installed on a fixed observation tower. Low-altitude drone images are obtained by using a commercial drone equipped with a high-resolution camera to fly once a week to cover the entire experimental field. IoT real-time images are obtained by small cameras deployed in multiple IoT sensor nodes in the field to upload pictures regularly every day to capture dynamic changes in the field. Local close-up images of artificial plants are obtained by trained technicians regularly entering the field using handheld cameras or mobile phones to take pictures, focusing on capturing representative local features of the plants.

[0016] Environmental monitoring data include temperature, humidity, light intensity, precipitation, soil Ph value, soil salinity and moisture content, etc., which can be obtained through distributed sensor networks.

[0017] The measured growth indicators can be extracted from the field measurement data of the experimental field, and the measured growth indicators include but are not limited to data in multiple dimensions such as growth status, yield, environmental adaptability, management strategy and long-term trend evaluation. Among them, growth status includes health status and development progress, etc. Health status includes but is not limited to chlorophyll content (SPAD value), incidence of pests and diseases, leaf health index and root vitality, and development progress includes but is not limited to stem height, number and length of leaves, number of tillers, heading period and filling period; yield includes yield level and yield fluctuation, yield level includes but is not limited to effective number of ears, number of grains per ear and thousand-grain weight, and yield fluctuation includes but is not limited to inter-annual variation, field differences and extreme weather impacts; environmental adaptability includes temperature adaptability, water adaptability and soil adaptability, etc. Temperature adaptability includes cold resistance and heat stress tolerance, water adaptability includes drought resistance and waterlogging resistance, and soil adaptability includes salt tolerance and nutrient utilization efficiency; management strategies include but are not limited to fertilization management, irrigation management, pest and disease control and farming activities; long-term trend evaluation includes overall trend of the growing season, impact of climate change and long-term management effect, etc. The overall trend of the growing season includes but is not limited to inter-annual variation and seasonal variation, etc. The impact of climate change includes but is not limited to temperature change, precipitation pattern and extreme weather events, etc., and long-term management effect includes but is not limited to cumulative effect and sustainability.

[0018] The multi-source fusion image data, environmental monitoring data and measured growth indicators can be associated with each other, and the data association includes timestamp matching and geographic coordinate synchronization. Timestamp matching refers to associating the features and growth indicators at the same time point to ensure that the timestamps of the image data, environmental monitoring data and field measurement data are consistent, while geographic coordinate synchronization refers to using GPS information to synchronize geographic coordinates to ensure consistency of spatial positions. The multi-source fusion image data, the environmental monitoring data and the measured growth indicators can be obtained by presetting a period, such as the rice growing season in the last three or five years.

[0019] Furthermore, the specific execution process of obtaining multi-source fusion image data and environmental monitoring data corresponding to different growth stages of rice in cold saline-alkali land areas in historical periods is as follows: obtaining environmental monitoring data corresponding to rice in cold saline-alkali land areas within a preset period; obtaining panoramic images, low-altitude images of unmanned aerial vehicles, real-time images of the Internet of Things, and local close-up images of artificial plants corresponding to rice in cold saline-alkali land areas according to the timestamp of the environmental monitoring data; pre-processing the panoramic images, low-altitude images of unmanned aerial vehicles, real-time images of the Internet of Things, and local close-up images of artificial plants, and using a preset matching algorithm to align and fuse the pre-processed panoramic images, pre-processed low-altitude images of unmanned aerial vehicles, pre-processed real-time images of the Internet of Things, and pre-processed local close-up images of artificial plants to obtain multi-source fusion image data; obtaining field measurement data corresponding to rice in cold saline-alkali land areas according to the timestamp of the environmental monitoring data, and extracting growth status, yield, environmental adaptability, management strategy, and long-term trend evaluation from the field measurement data as measured growth indicators.

[0020] In this step, the preset period is usually one or more complete rice growing seasons (such as 3-6 months), which can be customized according to research needs. Environmental monitoring data specifically include temperature, humidity, light intensity, precipitation, soil Ph value, soil salinity, moisture content and nutrient content. Among them, temperature, humidity and light intensity can be regularly collected by automatic weather stations pre-deployed in the experimental fields, soil Ph value, soil salinity, moisture content and nutrient content can be regularly detected by multiple soil sensor nodes distributed in the experimental fields, and precipitation can be deployed in rain gauges in the experimental fields to automatically record each rainfall and its time. In order to ensure the accuracy of environmental monitoring data, outliers (such as extreme values ​​caused by sensor failures) can be identified and removed, and missing values ​​can be filled using interpolation or other statistical methods, while alignment based on timestamps is convenient for subsequent analysis. Based on the timestamp of the environmental monitoring data, the image data from different sources such as panoramic images, low-altitude images of drones, real-time images of the Internet of Things, and local close-up images of artificial plants are first aligned with the environmental monitoring data, and then the measured growth indicators are aligned with the environmental monitoring data to ensure that all data at the same time point are accurately associated. Before fusion, the panoramic images, low-altitude images of drones, real-time images of the Internet of Things, and local close-up images of artificial plants can be preprocessed separately. The preprocessing includes cleaning, standardization, and adding labels. By adding labels, healthy plants and diseased plants or other abnormal conditions can be marked for subsequent analysis. After preprocessing, the SIFT (Scale Invariant Feature Transform) or SURF (Speeded Robust Features) algorithm is used to identify and match the same positions from different perspectives to achieve accurate registration between images. The key points are extracted by Harris corner detection or FAST algorithm to ensure the accuracy of matching. A fusion method based on weighted average is adopted. The specific expression is as follows: ; in, is the fused image at coordinates The pixel value at Indicates The pixel value of the same position in the image, It is The weight of each image can be dynamically adjusted according to the image quality or signal-to-noise ratio.

[0021] In order to ensure the fusion quality, after obtaining the multi-source fusion image data, the structural similarity index of the images before and after fusion can also be calculated. and peak signal-to-noise ratio , evaluate the fusion effect. The specific expression is as follows: ; ; in, and are the average brightness of the image before and after fusion, and is the standard deviation of the two, is the covariance of the two, and is a constant, is the maximum possible pixel value, is the mean square error.

[0022] The value range is [-1,1], where 1 means that the image before and after fusion are exactly the same, 0 means no similarity, and negative values ​​mean no similarity. A high-quality threshold can be pre-set, such as 0.95. When , it indicates that the fusion effect is good, indicating that the fused image retains the main structural information of each image before fusion, and the visual effect is close to the original image. The higher the value, the smaller the difference between the fused image and the images before and after fusion, that is, the higher the signal-to-noise ratio, the less noise. Similarly, a high-quality threshold can be pre-set, such as 30 dB. When , it means the fusion effect is good, indicating that the difference between the fused image and the images before fusion is small, and the visual effect is almost lossless. and , calculate the difference between the two and the high-quality threshold respectively, and set an interval for the difference. Set different weight coefficients for each interval according to the user's preset emphasis on the two. Get the final score through weighted calculation, and compare the final score with the preset score threshold to measure the fusion quality.

[0023] 102. Deep feature extraction is performed on multi-source fusion image data based on measured growth indicators to obtain the key image features corresponding to different growth stages of rice in cold saline-alkali land areas in historical periods.

[0024] In this step, the image features include but are not limited to macro texture and spatial distribution features, micro physiological features and individual fine features, etc. The macro texture and spatial distribution features are used to reflect the large-area field landscape, such as field layout, canopy coverage, vegetation index (NDVI, EVI), etc. The micro physiological features involve microscopic information such as spectral reflectance, cell structure, physiological indicators (SPAD value, Fv / Fm ratio), etc. The individual fine features are used to capture the specific morphology of a single plant, such as morphological characteristics (stem height, leaf length and width, number of grains per ear), local close-ups (lesions, insect pests), 3D reconstruction models, etc. Since the above-mentioned image features include not only features that have a significant impact on the growth indicators corresponding to different growth stages in the historical period, but also other redundant features, in order to improve the efficiency and accuracy of the subsequent model, it is necessary to extract key features that have a strong correlation with the measured growth indicators after determining the image features. Specifically, the Pearson correlation coefficient, Spearman rank correlation coefficient, mutual information and other correlation algorithms can be used to calculate the first correlation between the above-mentioned image features and the measured growth indicators, and compare them with a preset correlation threshold (such as 0.8) to screen out features that have a significant impact on the growth indicators, namely, key image features.

[0025] Furthermore, deep feature extraction is performed on the multi-source fusion image data to obtain the key features corresponding to different growth stages of rice in the cold saline-alkali land area in the historical period. The specific execution process is: deep feature extraction is performed on the multi-source fusion image data to obtain the first image feature, the second image feature and the third image feature; the first correlation between each feature of the first image feature, the second image feature and the third image feature and the measured growth index is calculated respectively; the feature with the first correlation higher than the preset image correlation threshold is used as the key image feature.

[0026] In this step, the first image feature is the image feature of macro texture and spatial distribution feature, the second image feature is the micro physiological feature, and the third image feature is the individual fine feature. For the extraction of macro texture and spatial distribution features, an image segmentation algorithm (such as GrabCut, U-Net) is used to separate the background and foreground in the image, highlight the rice plants, and use gray level co-occurrence matrix (GLCM), local binary pattern (LBP) and other methods to extract texture features such as contrast, correlation, energy, entropy, etc., and by calculating field layout features such as plant spacing, density, uniformity, etc., the overall distribution is reflected, such as rasterization processing and statistical analysis methods. For the extraction of micro physiological features, a hyperspectral camera is used to obtain the spectral reflectance curve of each rice plant, identify physiological parameters such as chlorophyll content and water status, and shoot the leaf cell structure through a microscope or scanning electron microscope. Combined with the deep learning model, the cell morphology, size, arrangement and other features are automatically identified, and combined with the data collected in the field (such as SPAD value, Fv / Fm ratio), the photosynthesis efficiency and health status of the plant are evaluated. For the individual fine feature extraction, edge detection and shape descriptors (such as Hu moments and Zernike moments) are used to extract the specific morphological features of the stems, leaves, ears and other parts. For the manually collected local close-up images, the small features of the key parts, such as disease spots and traces of insect pests, are extracted. The three-dimensional model of the rice plant is generated through multi-view image reconstruction technology, and its geometric characteristics and spatial structure are further analyzed.

[0027] After obtaining the above-mentioned first image feature, second image feature and third image feature, the first correlation between each feature and the measured growth index can be calculated using correlation algorithms such as Pearson correlation coefficient, Spearman rank correlation coefficient, mutual information, etc. According to the actual situation, an image correlation threshold is pre-set, such as 0.8, and the feature with a first correlation higher than the preset image correlation threshold is used as the key image feature, which is considered to have a significant impact on the growth status. On this basis, the principal component analysis (PCA) or ridge regression methods can be used to further remove redundant features in the key image features, thereby retaining the most representative features, improving the accuracy of subsequent models, and realizing accurate evaluation of rice growth in cold saline-alkali land.

[0028] With respect to the above description of the correlation algorithm, a specific example of using the Pearson correlation coefficient to calculate the first correlation is as follows: ; in, and Image features and growth indicators In the The value at the sample point, and Image features and growth indicators The mean of .

[0029] 103. Deep feature extraction of environmental monitoring data is carried out based on multi-source image data to obtain the key environmental characteristics corresponding to different growth stages of rice in cold saline-alkali areas in historical periods.

[0030] In this step, the environmental change trend characteristics include meteorological condition change trend characteristics (such as temperature, humidity, light intensity, wind speed and wind direction trend changes), soil condition change trend characteristics (such as soil Ph value, soil salt content, water content, nutrient content trend changes) and precipitation change trend characteristics (such as cumulative precipitation, rainfall frequency, and trend changes of heavy rainfall events). Since the above environmental change trend characteristics not only have features that have a significant impact on the image features corresponding to different growth stages in the historical period, but also have other redundant features, therefore, in order to improve the efficiency and accuracy of the subsequent model, it is necessary to extract key features that have a strong correlation with the first image feature, the second image feature, and the third image feature in step 102 after determining the environmental change trend characteristics. Specifically, the second correlation between the above environmental change trend characteristics and the image features can also be calculated using correlation algorithms such as Pearson correlation coefficient, Spearman rank correlation coefficient, mutual information, etc., and compared with a preset correlation threshold (such as 0.6), so as to screen out features that have a significant impact on the image features, that is, key environmental features.

[0031] Furthermore, deep feature extraction is performed on the environmental monitoring data according to the second correlation between the environmental change trend characteristics corresponding to the environmental monitoring data and the image characteristics corresponding to the multi-source image data to obtain the key environmental characteristics corresponding to the different growth stages of rice in the cold saline-alkali land area in the historical period. The specific execution process is: deep feature extraction is performed on the environmental monitoring data to obtain the environmental change trend characteristics corresponding to the different growth stages of rice in the cold saline-alkali land area in the historical period; the second correlation between each feature of the first image feature, the second image feature and the third image feature and the environmental change trend characteristics is calculated respectively; and the features with the second correlation higher than the preset environmental correlation threshold are taken as key environmental features.

[0032] In this step, the environmental change trend characteristics are used to reflect the changing rules and patterns of environmental variables over time, which is crucial for understanding the impact on rice growth. The environmental change trend characteristics specifically include meteorological condition change trends (temperature trend, humidity trend, light intensity trend), soil condition change trends (soil Ph value trend, soil salt content trend, soil moisture trend, soil nutrient trend), and precipitation change trends (cumulative precipitation trend, rainfall frequency trend, and heavy rainfall event trend). For each environmental change trend characteristic , a linear regression model can be used to capture its trend changes. The specific expression is: ; in, is the environmental characteristic value on day t, and is the regression coefficient, is the error term.

[0033] The second correlation between each image feature in step 102 and the above-mentioned environmental change trend feature is calculated respectively. According to the actual situation, an environmental correlation threshold is pre-set, such as 0.6, and the feature with the second correlation higher than the preset environmental correlation threshold is used as the key environmental feature, which is considered to have a significant impact on the growth status of rice at different growth stages. On this basis, the redundant features in the key environmental features can also be further removed by using methods such as principal component analysis (PCA) or ridge regression, so as to retain the most representative features, improve the accuracy of the subsequent model, and achieve accurate evaluation of rice growth in cold saline-alkali land.

[0034] With respect to the above description of the correlation algorithm, a specific example of using the Pearson correlation coefficient to calculate the second correlation is as follows: ; in, and Environmental characteristics and imaging features In the The value at the sample point, and Environmental characteristics and imaging features The mean of .

[0035] In order to further optimize the selection of key environmental features, a method based on information gain or mutual information can also be introduced to evaluate the importance of features. The specific expression is: ; in, It is an environmental feature and imaging features The mutual information between is the joint distribution probability, and is the marginal probability distribution. By calculating the mutual information, the environmental features that have the greatest information contribution to the image features can be selected as key environmental features, thereby improving the prediction ability of the model.

[0036] Considering the complex and changeable environment of cold saline-alkali land, by analyzing the impact of environmental changes on image features, we can indirectly capture the complex impact of the environment on rice growth and analyze the mechanism of environmental factors in more detail. For example, changes in soil Ph value and soil salt content may first affect the color and morphology of leaves, which are then reflected in image features and ultimately affect the growth status. Image features, as an intermediate layer between the environment and growth indicators, provide more detailed information. By monitoring changes in image features, potential problems such as pests and diseases, insufficient water, etc. can be discovered earlier, so that measures can be taken in advance. Because image features are relatively stable, they can more reliably reflect the actual effects of environmental changes. Directly linking environmental changes with image features can reduce the noise that may be introduced when environmental factors directly act on growth indicators. In addition, by establishing the relationship between environmental changes and image features, new research perspectives can be provided to better understand the specific impact paths of environmental factors on crop growth.

[0037] 104. The attention mechanism is used to process the key image features and key environmental features corresponding to different growth stages in the historical period, and a growth assessment model for rice in cold saline-alkali land areas is constructed based on the processed key image features, processed key environmental features and measured growth indicators.

[0038] Among them, the growth assessment model is used to output the predicted growth indicators corresponding to different growth stages of rice in cold saline-alkali land areas in the future period after inputting the processed current key image features and the processed current key environmental features corresponding to different growth stages in the current period.

[0039] In this step, an adaptive attention mechanism is introduced to perform weighted fusion of the key image features and key environmental features corresponding to different growth stages in the historical period. By learning the importance scores of different features, the weights of each feature can be dynamically adjusted so that the model pays more attention to the most representative key features among the image features and key environmental features, thereby enhancing the model's attention to the important features among the key features, thereby improving the accuracy of the model.

[0040] Furthermore, the attention mechanism is used to process the key image features and key environmental features corresponding to different growth stages in the historical period, and the specific execution process of constructing the growth assessment model of rice in the cold saline-alkali land area is as follows: the attention mechanism is used to perform weighted fusion on the key image features and key environmental features corresponding to different growth stages in the historical period, and the unified image feature vectors and unified environmental feature vectors corresponding to the different growth stages of rice in the cold saline-alkali land in the historical period are obtained; the unified image feature vectors and unified environmental feature vectors corresponding to the different growth stages in the historical period are used as input, the measured growth indicators corresponding to the different growth stages in the historical period are used as output, and a sample set is constructed according to a preset ratio; the sample set is used to train a hybrid model of a convolutional neural network combined with a long short-term memory network, and the growth assessment model of rice in the cold saline-alkali land area is obtained.

[0041] Among them, the attention weights of key image features and key environmental features correspond to the growth stages.

[0042] In this step, the attention weights can be dynamically adjusted according to different growth stages to ensure that the model can capture the most relevant information at each growth stage. Specifically, a set of attention weights is initialized for each key image feature and key environmental feature. The attention weights reflect the importance of different key image features and key environmental features at a specific growth stage. The attention weights are dynamically adjusted according to the characteristics of each growth stage. For example, in the seedling stage, more attention may be paid to microscopic physiological characteristics, while in the heading stage, more attention is paid to macroscopic texture features. Multiply the key image features and key environmental features corresponding to each growth stage by their corresponding attention weights, and then sum them to obtain the unified image feature vector and unified environmental feature vector for that growth stage. The specific expressions are: ; ; in, and They are Key environmental characteristics and Key image features, and They are Key environmental characteristics and The attention weights of each key image feature, and are the sample numbers of key environmental characteristics, respectively.

[0043] The unified image feature vector and unified environmental feature vector corresponding to different growth stages in the historical period are used as input, and the measured growth indicators corresponding to different growth stages in the historical period are used as output to construct a sample set. The data is divided into a training set and a test set according to a preset ratio (such as 80% training set and 20% test set) to ensure the generalization ability of the model. The k-fold cross-validation method is used to further improve the stability and reliability of the model. Based on the sample set, a hybrid model of a convolutional neural network combined with a long short-term memory network is used for training to obtain a growth prediction model. Among them, the convolutional neural network is used to extract the spatial structural information of image features. It can automatically learn local patterns in images, such as textures and edges, and is suitable for processing multi-source image data, while the long short-term memory network is used to capture long-term dependencies in time series data. It is particularly suitable for processing time series features of environmental monitoring data, such as temperature and humidity change trends. It should be noted that the growth prediction model needs to predict multiple growth indicators (such as health status, yield, etc.) at the same time. A multi-task learning framework can be designed to predict multiple future growth indicators at the same time and share the underlying feature representation. The future growth index at least includes the prediction of the future growth state, future yield, future environmental adaptability, future management strategy and future long-term trend evaluation corresponding to different growth stages of rice in the cold saline-alkali land area in the historical period. After inputting the current unified image feature vector and the current unified environmental feature vector corresponding to different growth stages in the current period, the predicted growth index corresponding to different growth stages of rice in the cold saline-alkali land area in the future period is output, including the future growth state, future yield, future environmental adaptability, future management strategy and future long-term trend evaluation.

[0044] Furthermore, in order to ensure the accuracy and timeliness of the model during continuous use, it also includes: monitoring the increment of multi-source fusion image data corresponding to different growth stages of rice in cold saline-alkali land areas in historical periods; if the increment of multi-source fusion image data is higher than the preset increment threshold, the growth assessment model is retrained based on the multi-source fusion image data containing the increment.

[0045] In this step, a fixed time window (such as 30 days) and an incremental threshold (such as 10% or 50MB) can be pre-set. The time window is used to accumulate and calculate the increment, that is, the multi-source fusion image data currently collected is compared with the data volume of the previous cycle according to the time window, and the newly added data volume is calculated. When the newly added data volume exceeds the incremental threshold, steps 101-104 are triggered to retrain the model. In order to reduce the consumption of computing resources, retraining can adopt a small batch update strategy, and only use newly acquired data for incremental training each time, instead of reusing all historical data. In addition, a unique version number is assigned to each updated model for easy tracking and management. A rollback mechanism can also be set up. If the new model performs poorly, it can be quickly restored to the previous version, that is, the previous version.

[0046] Furthermore, in order to facilitate users to visualize the rice growth assessment results predicted by the model, so as to quickly and accurately know the rice growth status and manage it, it also includes: generating the corresponding growth status information of rice in the cold saline-alkali land area based on the predicted growth index; marking the growth status information on the visualization map corresponding to the rice in the cold saline-alkali land area, and displaying it in a preset display mode, so as to guide users to accurately manage the rice in the cold saline-alkali land area.

[0047] Since the predicted growth indicators include future growth status, future yield, future environmental adaptability, future management strategy and future long-term trend evaluation, in this step, the multi-dimensional data such as future growth status, future yield, future environmental adaptability, future management strategy and future long-term trend evaluation predicted by the growth assessment model are integrated to obtain growth status information. In order to facilitate the user's intuitive understanding, a scoring system (such as 0-10 points) can also be designed for each dimension, and the corresponding score can be assigned according to the prediction results.

[0048] High-resolution satellite images or drone aerial images were used as the base map, and according to the actual layout of the experimental field, the boundaries of the plots were drawn, and the information of different areas was marked to obtain a visual map of the cold saline-alkali land area. The growth status information was marked in the visual map, including the growth status layer: different colors or icons were used to represent the growth status of each plot (such as green for health and red for disease), the yield prediction layer: a heat map was used to show the expected yield of each area, and the color depth reflected the yield, the environmental adaptability layer: a chart or label was used to show the temperature adaptability and water requirements of each area, and the management strategy layer: arrows or symbols were used to indicate the best time and location for management activities such as fertilization, irrigation, and pest control.

[0049] Through API interface or database connection, ensure that the visual map is updated synchronously with the latest data, and set up a regular push function to send the latest growth status report and management suggestions to users. It also allows users to click on any location on the map to view the specific growth status information of that location, and provides filtering and sorting functions to help users quickly find key areas of concern.

[0050] Based on the above Figure 1 It can be seen from the implementation method that the present application provides a method for evaluating the growth of rice in cold saline-alkali land based on image data fusion, which ensures that the model can obtain comprehensive information input by obtaining multi-source fused image data, environmental monitoring data and measured growth indicators corresponding to different growth stages of rice in cold saline-alkali land areas in historical periods. The multi-source fused image data not only improves the richness and comprehensiveness of the data, but also enhances the model's adaptability to complex environmental changes in cold saline-alkali land areas. By extracting deep key image features through measured growth indicators, those image features that have a direct impact on rice growth status, yield, etc. can be screened out, ensuring that the model can focus on what is really important. The information obtained by the image processing can reduce the interference of redundant features. By extracting deep key environmental features from environmental monitoring data, the complex impact of environmental factors on rice growth can be indirectly captured, and the mechanism of action of environmental factors can be analyzed more carefully, providing a more comprehensive environmental adaptability assessment. In addition, the environment of cold saline-alkali land is changeable, and image features can reflect these changes in real time, so that the model can better adapt to different environmental conditions. At the same time, the introduction of an adaptive attention mechanism to process key features enhances the model's attention to more important features among the key features, further improving the accuracy of the growth assessment model in assessing the growth of rice in cold saline-alkali land, thereby providing users with reliable decision support.

[0051] Furthermore, as a response to the above Figure 1 The implementation of the method embodiment shown in the figure, the embodiment of the present application provides a cold-region saline-alkali land rice growth assessment system based on image data fusion, the system is used to improve the accuracy of cold-region saline-alkali land rice growth assessment to provide reliable decision support. The embodiment of the system corresponds to the aforementioned method embodiment. For ease of reading, this embodiment will no longer repeat the details of the aforementioned method embodiment one by one, but it should be clear that the system in this embodiment can correspond to all the contents of the aforementioned method embodiment. Specifically, Figure 2 As shown, the system includes: An acquisition unit 21 is used to acquire multi-source fusion image data, environmental monitoring data and measured growth indicators corresponding to different growth stages of rice in the cold saline-alkali land area in historical periods; A first extraction unit 22 is used to perform deep feature extraction on the multi-source fusion image data according to the measured growth index to obtain key image features corresponding to different growth stages of rice in the cold saline-alkali land area in historical periods; The second extraction unit 23 is used to perform deep feature extraction on the environmental monitoring data according to the multi-source image data to obtain key environmental features corresponding to different growth stages of rice in the cold saline-alkali land area in historical periods; The construction unit 24 is used to use the attention mechanism to process the key image features and the key environmental features corresponding to different growth stages in the historical period respectively, and to construct a growth assessment model corresponding to rice in the cold saline-alkali land area based on the processed key image features, the processed key environmental features and the measured growth indicators, wherein the growth assessment model is used to output the predicted growth indicators corresponding to different growth stages of rice in the cold saline-alkali land area in the future period after inputting the processed current key image features and the processed current key environmental features corresponding to different growth stages in the current period.

[0052] Further, such as Figure 3 As shown, the acquisition unit 21 includes: The first acquisition module 211 is used to acquire the environmental monitoring data corresponding to rice in the cold saline-alkali land area within a preset period; The second acquisition module 212 is used to obtain the panoramic image, the low-altitude image of the drone, the real-time image of the Internet of Things, and the local close-up image of the artificial plant corresponding to the rice in the cold saline-alkali land area according to the timestamp of the environmental monitoring data; The fusion processing module 213 is used to pre-process the panoramic image, the drone low-altitude image, the IoT real-time image and the artificial plant local close-up image, and to register and fuse the pre-processed panoramic image, the pre-processed drone low-altitude image, the pre-processed IoT real-time image and the pre-processed artificial plant local close-up image using a preset matching algorithm to obtain the multi-source fused image data; The third acquisition module 214 is used to obtain the field measurement data corresponding to the rice in the cold saline-alkali land area according to the timestamp of the environmental monitoring data, and extract the growth status, yield, environmental adaptability, management strategy and long-term trend evaluation from the field measurement data as the measured growth indicators.

[0053] Further, such as Figure 3 As shown, the first extraction unit 22 includes: A first extraction module 221 performs deep feature extraction on the multi-source fusion image data to obtain a first image feature, a second image feature and a third image feature, wherein the first image feature is a macro texture and spatial distribution feature image feature, the second image feature is a micro physiological feature, and the third image feature is an individual fine feature; A first calculation module 222, used to respectively calculate a first correlation between each of the first image feature, the second image feature and the third image feature and the measured growth index; The first determining module 223 is configured to use the feature whose first correlation is higher than a preset image correlation threshold as the key image feature.

[0054] Further, such as Figure 3 As shown, the second extraction unit 23 includes: The second extraction module 231 is used to perform deep feature extraction on the environmental monitoring data to obtain the environmental change trend characteristics corresponding to different growth stages of rice in the cold saline-alkali land area in the historical period; A second calculation module 232, used to respectively calculate a second correlation between each of the first image feature, the second image feature and the third image feature and the environment change trend feature; The second determining module 233 is configured to use the feature whose second relevance is higher than a preset environment relevance threshold as the key environment feature.

[0055] Further, such as Figure 3 As shown, the construction unit 24 includes: The feature processing module 241 is used to use the attention mechanism to perform weighted fusion on the key image features and the key environmental features corresponding to different growth stages in the historical period, so as to obtain unified image feature vectors and unified environmental feature vectors corresponding to different growth stages of rice in cold saline-alkali land in the historical period, wherein the attention weights of the key image features and the key environmental features correspond to the growth stage; A sample construction module 242, for taking the unified environmental feature vector and the unified environmental feature vector corresponding to different growth stages in the historical period as input, taking the measured growth index corresponding to different growth stages in the historical period as output, and constructing a sample set according to a preset ratio; The model training module 243 is used to use the sample set to train a hybrid model of a convolutional neural network combined with a long short-term memory network to obtain the growth assessment model corresponding to rice in the cold saline-alkali land area.

[0056] Further, such as Figure 3 As shown, the system also includes: The monitoring unit 25 is used to process the key image features and the key environmental features corresponding to different growth stages in the historical period by using the attention mechanism, and build a growth assessment model corresponding to rice in the cold saline-alkali land area based on the processed key image features, the processed key environmental features and the measured growth index, and then monitor the multi-source fusion image data increment corresponding to different growth stages of rice in the cold saline-alkali land area in the historical period; The retraining unit 26 is configured to retrain the growth assessment model based on the multi-source fusion image data including the increment if the increment of the multi-source fusion image data is higher than a preset increment threshold.

[0057] Further, such as Figure 3 As shown, the system also includes: A generating unit 27 is used to process the key image features and the key environmental features corresponding to different growth stages in the historical period by using an attention mechanism, and to construct a growth assessment model corresponding to rice in the cold saline-alkali land area based on the processed key image features, the processed key environmental features and the measured growth index, and then generate growth status information corresponding to rice in the cold saline-alkali land area based on the predicted growth index; The visualization unit 28 is used to mark the growth status information in the visualization map corresponding to the rice in the cold saline-alkali land area, and display it according to a preset display method to guide the user to accurately manage the rice in the cold saline-alkali land area.

[0058] Furthermore, an embodiment of the present application further provides a storage medium, wherein the storage medium is used to store a computer program, wherein when the computer program is running, the device where the storage medium is located is controlled to execute the above Figure 1 The rice growth assessment method in cold saline-alkali land based on image data fusion is described in.

[0059] Furthermore, the embodiment of the present application also provides a processor, the processor is used to run a program, wherein the program executes the above Figure 1 The rice growth assessment method in cold saline-alkali land based on image data fusion is described in.

[0060] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0061] It is understandable that the related features in the above methods and systems can be referenced to each other. In addition, the "first", "second" and the like in the above embodiments are used to distinguish the embodiments, and do not represent the advantages and disadvantages of the embodiments.

[0062] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0063] The algorithm and display provided herein are not inherently related to any particular computer, virtual system or other device. Various general purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing such systems. In addition, the application is not directed to any specific programming language either. It should be understood that various programming languages ​​can be utilized to realize the content of the application described herein, and the description of the specific language above is for the purpose of disclosing the best mode of implementation of the application.

[0064] In addition, the memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0065] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0066] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A system that specifies the functions of a box or multiple boxes.

[0067] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction system, which is implemented in the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0068] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0069] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0070] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0071] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0072] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0073] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0074] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for evaluating rice growth in saline-alkali land in cold regions based on image data fusion, characterized in that: The method comprises: Obtain multi-source fusion image data, environmental monitoring data, and measured growth indicators corresponding to different growth stages of rice in cold saline-alkali land areas in historical periods; Performing deep feature extraction on the multi-source fusion image data according to the measured growth index to obtain key image features corresponding to different growth stages of rice in the cold saline-alkali land area in historical periods; Performing deep feature extraction on the environmental monitoring data according to the multi-source image data to obtain key environmental features corresponding to different growth stages of rice in the cold saline-alkali land area in historical periods; The attention mechanism is used to process the key image features and the key environmental features corresponding to different growth stages in the historical period respectively, and a growth assessment model corresponding to rice in the cold saline-alkali land area is constructed based on the processed key image features, the processed key environmental features and the measured growth indicators, wherein the growth assessment model is used to output the predicted growth indicators corresponding to different growth stages of rice in the cold saline-alkali land area in the future period after inputting the processed key image features and the processed key environmental features corresponding to different growth stages in the current period.

2. The method according to claim 1, characterized in that Obtain multi-source fusion image data, environmental monitoring data, and measured growth indicators corresponding to different growth stages of rice in cold saline-alkali land areas in historical periods, including: Obtain the environmental monitoring data corresponding to rice in the cold saline-alkali land area within a preset period; According to the timestamp of the environmental monitoring data, panoramic images, drone low-altitude images, IoT real-time images, and local close-up images of artificial plants corresponding to rice in the cold saline-alkali land area are obtained respectively; Preprocessing the panoramic image, the drone low-altitude image, the IoT real-time image, and the artificial plant local close-up image, and registering and fusing the preprocessed panoramic image, the preprocessed drone low-altitude image, the preprocessed IoT real-time image, and the preprocessed artificial plant local close-up image using a preset matching algorithm to obtain the multi-source fused image data; According to the timestamp of the environmental monitoring data, the field measurement data corresponding to the rice in the cold saline-alkali land area is obtained, and the growth status, yield, environmental adaptability, management strategy and long-term trend evaluation are extracted from the field measurement data as the measured growth indicators.

3. The method according to claim 1, characterized in that According to the measured growth index, the multi-source fusion image data is subjected to deep feature extraction to obtain key image features corresponding to different growth stages of rice in the cold saline-alkali land area in historical periods, including: Performing deep feature extraction on the multi-source fusion image data to obtain a first image feature, a second image feature and a third image feature, wherein the first image feature is a macro texture and spatial distribution feature image feature, the second image feature is a micro physiological feature, and the third image feature is an individual fine feature; respectively calculating first correlations between the first image feature, the second image feature, and the third image feature and the measured growth index; The feature whose first correlation is higher than a preset image correlation threshold is used as the key image feature.

4. The method according to claim 3, characterized in that: The environmental monitoring data is subjected to deep feature extraction based on the multi-source image data to obtain key environmental features corresponding to different growth stages of rice in the cold saline-alkali land area in historical periods, including: Perform deep feature extraction on the environmental monitoring data to obtain environmental change trend characteristics corresponding to different growth stages of rice in cold saline-alkali land areas in historical periods; respectively calculating second correlations between the first image feature, the second image feature, and the third image feature and the environment change trend feature; The feature whose second relevance is higher than a preset environment relevance threshold is used as the key environment feature.

5. The method according to any one of claims 1 to 4, characterized in that The key image features corresponding to different growth stages in the historical period are processed using the attention mechanism, and a growth assessment model corresponding to rice in the cold saline-alkali land area is constructed based on the processed key image features, the processed key environmental features and the measured growth indicators, including: The attention mechanism is used to perform weighted fusion on the key image features and the key environmental features corresponding to different growth stages in the historical period, so as to obtain unified image feature vectors and unified environmental feature vectors corresponding to different growth stages of rice in cold saline-alkali land in the historical period, wherein the attention weights of the key image features and the key environmental features correspond to the growth stages; Taking the unified environmental characteristic vector and the unified environmental characteristic vector corresponding to different growth stages in the historical period as input, taking the measured growth index corresponding to different growth stages in the historical period as output, and constructing a sample set according to a preset ratio; The sample set is used to train a hybrid model of a convolutional neural network combined with a long short-term memory network to obtain the growth assessment model corresponding to rice in the cold saline-alkali land area.

6. The method according to any one of claims 1 to 4, characterized in that After using the attention mechanism to process the key image features and the key environmental features corresponding to different growth stages in the historical period, and constructing a growth assessment model corresponding to rice in the cold saline-alkali land area based on the processed key image features, the processed key environmental features and the measured growth indicators, the method further includes: Monitor the increment of multi-source fusion image data corresponding to different growth stages of rice in cold saline-alkali land areas in historical periods; If the increment of the multi-source fusion image data is higher than a preset increment threshold, the growth assessment model is retrained based on the multi-source fusion image data containing the increment.

7. The method according to claim 6, characterized in that After using the attention mechanism to process the key image features and the key environmental features corresponding to different growth stages in the historical period, and constructing a growth assessment model corresponding to rice in the cold saline-alkali land area based on the processed key image features, the processed key environmental features and the measured growth indicators, the method further includes: Generate growth status information corresponding to rice in the cold saline-alkali land area based on the predicted growth index; The growth status information is marked correspondingly in a visualization map corresponding to the rice in the cold saline-alkali land area, and displayed in a preset display mode to guide the user to accurately manage the rice in the cold saline-alkali land area.

8. A rice growth assessment system for saline-alkali land in cold regions based on image data fusion, characterized in that: The system comprises: An acquisition unit is used to obtain multi-source fusion image data, environmental monitoring data and measured growth indicators corresponding to different growth stages of rice in cold saline-alkali land areas in historical periods; A first extraction unit is used to perform deep feature extraction on the multi-source fusion image data according to the measured growth index to obtain key image features corresponding to different growth stages of rice in the cold saline-alkali land area in historical periods; A second extraction unit is used to perform deep feature extraction on the environmental monitoring data according to the multi-source image data to obtain key environmental features corresponding to different growth stages of rice in the cold saline-alkali land area in historical periods; A construction unit is used to use an attention mechanism to process the key image features and the key environmental features corresponding to different growth stages in the historical period, and to construct a growth assessment model corresponding to rice in the cold saline-alkali land area based on the processed key image features, the processed key environmental features and the measured growth indicators, wherein the growth assessment model is used to output predicted growth indicators corresponding to different growth stages of rice in the cold saline-alkali land area in the future period after inputting the processed key image features and the processed key environmental features corresponding to different growth stages in the current period.

9. A storage medium, characterized in that: The storage medium includes a stored program, wherein when the program is running, the device where the storage medium is located is controlled to execute the method for evaluating rice growth in cold saline-alkali land based on image data fusion as described in any one of claims 1 to 7.

10. A processor, characterized in that: The processor is used to run a program, wherein the program, when running, executes the method for evaluating rice growth in cold saline-alkali land based on image data fusion as described in any one of claims 1 to 7.

Images