Digital agricultural automatic control management system

Abstract

The invention relates to the technical field of agricultural management, in particular to a digital agricultural automatic control management system, which comprises the following steps of: acquiring a plant production condition picture in a region by using a satellite, segmenting the picture, dividing the picture into different regions, numbering the regions, and storing the numbered regions into a database; utilizing a GPS positioning technology to match farmland data sensors into the divided areas; obtaining a plant growth image training set, establishing a plant growth prediction model of an ST-LSTM architecture, and inputting the plant growth images into the prediction model to obtain a trained model; the method comprises the steps that a satellite picture is divided for three times, the divided picture is input into a trained model, a plant growth state prediction result is obtained, a plant growth model is compared according to the plant growth state prediction result, the plant growth state is judged, a corresponding area is matched according to the result, and nutrients are supplemented in time according to data. Pictures are finely divided, so that the soil condition of each planting area is conveniently obtained, and a corresponding cultivation scheme is provided.

Description

Technical Field

[0001] This invention relates to the field of agricultural management technology, and more specifically to a digital agricultural automatic control and management system. Background Technology

[0002] The Internet of Things (IoT) is an emerging network communication technology in recent years. It utilizes communication technologies such as local area networks (LANs) or the internet to connect sensors, controllers, machines, people, and objects in new ways, forming a network that connects people and things, and things with each other, enabling information-based, remote management and control, and intelligent operation. The Internet of Things (IoT) in agriculture is driving automation, informatization, and intelligentization in agriculture, attracting significant attention both domestically and internationally. Some Western countries have begun using dedicated resource satellites for real-time monitoring of farmland resources, employing sensor fusion technology for real-time monitoring of agricultural ecology, and utilizing precision fertilization and irrigation systems to improve fertilizer utilization. In my country, GPS positioning technology and other wireless sensor network technologies are used to collect, transmit, and manage parameters of large-scale, contiguous agricultural production areas, and ground monitoring stations are used to build agricultural environmental monitoring systems. However, existing agricultural planting management systems struggle to provide corresponding cultivation plans based on the soil conditions of each planting area within a large-scale, contiguous agricultural production area over a given period. Furthermore, they struggle to assess the environment and growth status of crops during their growth process, making timely nutrient or water supplementation difficult, and failing to make timely and precise adjustments to crop growth in small areas. Therefore, we propose a digital agricultural automatic control and management system. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a digital agricultural automatic control and management system, including an image segmentation module, a model training module, and a growth prediction module.

[0004] The image segmentation module includes using satellite-collected images of plant production status in the area to segment the images, dividing them into different regions and numbering the regions, and using GPS positioning technology to match farmland data sensors to the divided regions.

[0005] The model training module includes acquiring a training set of plant growth images, establishing a plant growth prediction model with an ST-LSTM architecture, inputting plant growth images into the prediction model, and obtaining the trained model.

[0006] The growth prediction module includes inputting the segmented graphics into the trained model to obtain plant growth status prediction results, comparing the plant growth status prediction results with the plant growth model to determine the plant growth status, matching the corresponding area according to the results, and replenishing nutrients or water in a timely manner based on the data.

[0007] Furthermore, the acquisition of the plant growth image training set includes,

[0008] Obtain crop growth images and spatial characteristics for the Rth crop during time period T. The spatial characteristics include temperature, humidity, light intensity, rainfall, and soil conditions.

[0009] Plant images are processed using CNN to obtain plant feature maps, and then the plant feature maps are flattened to have x positions, each position having y-dimensional visual features;

[0010] The spatial state features are edited to have x1 locations, each with a y1-dimensional environmental feature.

[0011] The first and second dimensions of the plant features are swapped. The swapped plant features are then multiplied by the environmental features in a batch matrix. Softmax is then applied to the second dimension to convert the similarity into a normalized attention weight matrix. Each row represents a set of weights, indicating the degree of attention a plant location receives from all environmental locations.

[0012] Perform a batch matrix multiplication between the attention weight matrix and the original plant features to obtain the fused features, and use the final fused features as the training set of plant growth images.

[0013] Furthermore, the ST-LSTM architecture includes an encoder and a predictor. The encoder includes a shallow coding layer, a downsampling coding layer, and a bottleneck coding layer. The predictor includes two upsampling layers, one refining layer, and one multi-task output layer.

[0014] Furthermore, the step of inputting plant growth images into the prediction model to obtain a trained model includes,

[0015] Obtain a training set of plant growth images, and divide the training set into a test set and a validation set;

[0016] Initialize the ST-LSTM architecture, input the test set into the encoder to first downsample and encode the spatiotemporal features, then upsample and decode, and finally output the growth feature vector of the future crop. Train with composite loss and use the validation set to validate the output results.

[0017] The formula for suture loss is:

[0018]

[0019] in, C represents temporal memory, M represents spatial memory, and L represents... O Let λ be the composite loss and λ be the task weight. Let y be the predicted value output by the model, γ be the true label, and γ be the mean squared error function.

[0020] Furthermore, the method of matching farmland data sensors to the pre-defined areas using GPS positioning technology includes,

[0021] Acquire crop growth maps taken by satellite, including but not limited to the seedling stage, growth stage, and maturity stage of crops; acquire the location of ground sensors and the ID of the devices, wherein the sensor location is real-time latitude and longitude.

[0022] Preprocess the satellite growth map to generate a raster NDVI map with a resolution of 1 to 3 meters;

[0023] Obtain the farmland boundary vectors from the NDVI map, divide the farmland into multiple closed polygon vector feature sets, including the vertex coordinates of the polygons, field number, field area, soil type and crop type;

[0024] Obtain the field area from the above closed polygon vector feature set. If the field area is less than 1 hectare, divide the field into four 25m x 25m grids. If the field area is between 1 and 5 hectares, generate a fishing net according to 50m x 50m. If the field area is greater than 5 hectares, set the grid side length to 100m to obtain a secondary grid and number each sub-region.

[0025] The secondary grid is divided into microgrids of 10m x 10m, and the microgrids are numbered to obtain the tertiary grid;

[0026] Obtain the latitude and longitude of the sensor, convert the latitude and longitude to planar coordinates, map the planar coordinates into the inner side of the NDVI map, and for each sensor point, find the grid with the shortest center point distance in the three-level grid, and bind the sensor to the grid with the shortest center point distance.

[0027] Furthermore, the timely replenishment of nutrients or water based on data includes:

[0028] Acquire plant images and sensor data in a three-level grid, extract plant image features and spatial features collected by the sensors, fuse the two sets of features to obtain a fused feature set, input the fused feature set into a trained plant growth model, predict the future growth results of the plant, and establish a local crop variety database, wherein the database includes, but is not limited to, data on the three growth cycles of the plant.

[0029] The predicted future plant growth data is compared with local crop growth data. If the predicted data is better than the local data, the plant growth conditions are not changed. If the predicted data is worse than the local data, the plant growth conditions are optimized.

[0030] The beneficial effects of this invention are reflected in:

[0031] 1. Preprocess satellite-captured farmland photos, then divide the farmland into multiple closed polygon vector feature sets based on the farmland's boundary vectors. Further subdivide these closed polygon vector feature sets to obtain a secondary grid. Number each sub-region, and then divide the secondary grid according to a 10m x 10m grid. The image is segmented a third time at 10m to obtain a three-level grid. Then, the latitude and longitude of the sensor are obtained, and the sensor is bound to the grid with the shortest distance to the center point in the three-level grid. The satellite image is divided into three parts for fine segmentation, which facilitates the acquisition of soil conditions in each planting area to provide corresponding cultivation plans. The plant features and spatial features of the three-level grid are fused, and the fused data is input into the trained plant growth model to predict the future growth results of the plant. The prediction results are compared with local data, and optimization is performed based on the comparison values. The environmental conditions and growth status of the crop during the growth process are judged to facilitate timely replenishment of nutrients or water and timely and accurate adjustments during the crop growth process. An ST-LSTM architecture including encoder and predictor is established, and the architecture is trained by collecting a training set of plant growth images to obtain a trained model for easy prediction of future plant growth. Attached Figure Description

[0032] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. The elements or parts in the drawings are not necessarily drawn to scale.

[0033] Figure 1 This is a structural diagram of the digital agricultural automatic control and management system in Example 1. Detailed Implementation

[0034] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0035] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application should have the ordinary meaning as understood by those skilled in the art to which this invention pertains.

[0036] like Figure 1 As shown, a digital agricultural automatic control and management system includes an image segmentation module S1, a model training module S2, and a growth prediction module S3.

[0037] The image segmentation module S1 includes using satellite-collected images of plant production status in the area to segment the images, dividing them into different regions and numbering the regions, and using GPS positioning technology to match farmland data sensors to the divided regions.

[0038] The model training module S2 includes acquiring a training set of plant growth images, establishing a plant growth prediction model with an ST-LSTM architecture, inputting plant growth images into the prediction model, and obtaining a trained model.

[0039] The growth prediction module S3 includes inputting the divided graphics into the trained model to obtain the plant growth status prediction results, comparing the plant growth status prediction results with the plant growth model to determine the plant growth status, matching the corresponding area according to the results, and timely supplementing nutrients or water according to the data.

[0040] In this embodiment, a training set of plant growth images is obtained, including:

[0041] Obtain crop growth images and spatial characteristics for the Rth crop during time period T. Spatial characteristics include temperature, humidity, light intensity, rainfall, and soil conditions.

[0042] Plant images are processed using CNN to obtain plant feature maps, and then the plant feature maps are flattened to have x positions, each position having y-dimensional visual features;

[0043] The spatial state features are edited to have x1 locations, each with a y1-dimensional environmental feature.

[0044] The first and second dimensions of the plant features are swapped. The swapped plant features are then multiplied by the environmental features in a batch matrix. Softmax is then applied to the second dimension to convert the similarity into a normalized attention weight matrix. Each row represents a set of weights, indicating the degree of attention a plant location receives from all environmental locations.

[0045] Perform a batch matrix multiplication between the attention weight matrix and the original plant features to obtain the fused features, and use the final fused features as the training set of plant growth images.

[0046] For example, assuming the number of images input into the CNN at one time is B, the resulting image shape is (B, 512, 7, 7). Flattening this image yields a shape of (B, 512, 49), which means unfolding a 7x7 grid into 49 locations, each with 512 visual features. Inputting 5 sensor variables and a 7x7 grid into the encoder outputs environmental features of (B, 5, 49). Similarly, with 49 locations and 5-dimensional environmental information for each location, swapping the first and second dimensions of the plant features yields (B, 49, 512). This is then compared with the environmental features (B, 5...). Perform a batch matrix multiplication (49) to obtain a 49×49 similarity matrix. Then, perform Softmax on the second dimension (i.e., the dimension with 49 positions) to convert the similarity into a normalized attention weight matrix, maintaining the shape (B, 49, 49). Each row is a set of weights, representing the degree of attention a certain plant position receives from all environmental positions. For each sample, multiply the 512×49 plant features by the 49×49 weight matrix. This is equivalent to using attention weights to perform a weighted summation of the 49 plant features to finally obtain a fused feature with the shape (B, 512, 49).

[0047] In this embodiment, the ST-LSTM architecture includes an encoder and a predictor. The encoder includes a shallow coding layer, a downsampling coding layer, and a bottleneck coding layer. The predictor includes two upsampling layers, a refinement layer, and a multi-task output layer.

[0048] In this embodiment, plant growth images are input into a prediction model to obtain a trained model, including...

[0049] Obtain a training set of plant growth images, and divide the training set into a test set and a validation set;

[0050] Initialize the ST-LSTM architecture, input the test set into the encoder to first downsample and encode the spatiotemporal features, then upsample and decode, and finally output the growth feature vector of the future crop. Train with composite loss and use the validation set to validate the output results.

[0051] The formula for suture loss is:

[0052]

[0053] in, C represents temporal memory, M represents spatial memory, and L represents... O For recovery

[0054] The combined loss λ is the task weight. Let y be the predicted value output by the model, γ be the true label, and γ be the mean squared error function.

[0055] In this embodiment, GPS positioning technology is used to match farmland data sensors to pre-defined areas, including:

[0056] Acquire crop growth maps taken by satellite, including but not limited to the seedling stage, growth stage, and maturity stage of crops; acquire the location of ground sensors and the ID of the devices, where the sensor location is real-time latitude and longitude.

[0057] Preprocess the satellite growth map to generate a raster NDVI map with a resolution of 1 to 3 meters;

[0058] Obtain the farmland boundary vectors from the NDVI map, divide the farmland into multiple closed polygon vector feature sets, including the vertex coordinates of the polygons, field number, field area, soil type and crop type;

[0059] Obtain the field area from the above closed polygon vector feature set. If the field area is less than 1 hectare, divide the field into four 25m x 25m grids. If the field area is between 1 and 5 hectares, generate a fishing net according to 50m x 50m. If the field area is greater than 5 hectares, set the grid side length to 100m to obtain a secondary grid and number each sub-region.

[0060] The secondary grid is divided into microgrids of 10m x 10m, and the microgrids are numbered to obtain the tertiary grid;

[0061] Obtain the latitude and longitude of the sensor, convert the latitude and longitude to planar coordinates, map the planar coordinates into the inner side of the NDVI map, and for each sensor point, find the grid with the shortest center point distance in the three-level grid, and bind the sensor to the grid with the shortest center point distance.

[0062] In this embodiment, replenishing nutrients or water in a timely manner based on data includes:

[0063] Plant images and sensor data from a three-level grid are acquired, and features from the plant images and spatial features collected by the sensors are extracted. The two sets of features are then fused to obtain a fused feature set, which is then input into a trained plant growth model to predict future plant growth outcomes.

[0064] Establish a local crop variety database, which includes, but is not limited to, data on the three growth cycles of plants;

[0065] The predicted future plant growth data is compared with local crop growth data. If the predicted data is better than the local data, the plant growth conditions are not changed. If the predicted data is worse than the local data, the plant growth conditions are optimized.

[0066] It should be noted that, in practice, satellite-captured farmland photos are preprocessed. Then, based on the farmland's boundary vectors, the farmland is divided into multiple closed polygon vector feature sets. These closed polygon vector feature sets are further subdivided to obtain a secondary grid. Each sub-region is numbered. The secondary grid is then subdivided a third time into 10m x 10m grids to obtain a tertiary grid. The sensor's latitude and longitude are obtained, and the grid with the shortest distance to the center point of the tertiary grid is bound to the sensor. Plant features and spatial features from the tertiary grid are fused. The fused data is input into a pre-trained plant growth model to predict future plant growth. The predicted results are compared with local data, and optimization is performed based on the comparison values. An ST-LSTM architecture, including an encoder and a predictor, is established. A training set of plant growth images is collected to train the architecture, resulting in a trained model.

[0067] This embodiment discloses a computer-readable storage medium, including a memory, a processor, and a computer program stored on the memory and running on the processor. When the processor executes the computer program, it implements any of the above-described methods to provide a digital agricultural automatic control and management system.

[0068] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters, weights, and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0069] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the flow or function according to the embodiments of the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired or wireless network. A computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives (SSDs).

[0070] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0071] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0072] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only one method, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0073] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0074] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A digital agricultural automatic control and management system, comprising an image segmentation module, a model training module, and a growth prediction module, characterized in that: The image segmentation module includes using satellite-collected images of plant production status in the area to segment the images, dividing them into different regions and numbering the regions, and using GPS positioning technology to match farmland data sensors to the divided regions. The model training module includes acquiring a training set of plant growth images, establishing a plant growth prediction model with an ST-LSTM architecture, inputting plant growth images into the prediction model, and obtaining the trained model. The growth prediction module includes inputting the segmented graphics into the trained model to obtain plant growth status prediction results, comparing the plant growth status prediction results with the plant growth model to determine the plant growth status, matching the corresponding area according to the results, and replenishing nutrients or water in a timely manner based on the data.

2. The digital agricultural automatic control and management system according to claim 1, characterized in that: The training set of plant growth images was obtained. include, Obtain crop growth images and spatial characteristics for the Rth crop during time period T. The spatial characteristics include temperature, humidity, light intensity, rainfall, and soil conditions. Plant images are processed using CNN to obtain plant feature maps, and then the plant feature maps are flattened to have x positions, each position having y-dimensional visual features; The spatial state features are edited to have x1 locations, each with a y1-dimensional environmental feature. The first and second dimensions of the plant features are swapped. The swapped plant features are then multiplied by the environmental features in a batch matrix. Softmax is then applied to the second dimension to convert the similarity into a normalized attention weight matrix. Each row represents a set of weights, indicating the degree of attention a plant location receives from all environmental locations. Perform a batch matrix multiplication between the attention weight matrix and the original plant features to obtain the fused features, and use the final fused features as the training set of plant growth images.

3. The digital agricultural automatic control and management system according to claim 2, characterized in that: The ST-LSTM architecture includes an encoder and a predictor. The encoder includes a shallow coding layer, a downsampling coding layer, and a bottleneck coding layer. The predictor includes two upsampling layers, one refinement layer, and one multi-task output layer.

4. The digital agricultural automatic control and management system according to claim 3, characterized in that: The process of inputting plant growth images into a prediction model to obtain a trained model includes... Obtain a training set of plant growth images, and divide the training set into a test set and a validation set; Initialize the ST-LSTM architecture, input the test set into the encoder to first downsample and encode the spatiotemporal features, then upsample and decode, and finally output the growth feature vector of the future crop. Train with composite loss and use the validation set to validate the output results. The formula for suture loss is: in, C represents temporal memory, M represents spatial memory, and L represents... O Let λ be the composite loss and λ be the task weight. Let y be the predicted value output by the model, γ be the true label, and γ be the mean squared error function.

5. The digital agricultural automatic control and management system according to claim 1, characterized in that: The process of matching farmland data sensors to predefined areas using GPS positioning technology includes, Acquire crop growth maps taken by satellite, including but not limited to the seedling stage, growth stage, and maturity stage of crops; acquire the location of ground sensors and the ID of the devices, wherein the sensor location is real-time latitude and longitude. Preprocess the satellite growth map to generate a raster NDVI map with a resolution of 1 to 3 meters; Obtain the farmland boundary vectors from the NDVI map, divide the farmland into multiple closed polygon vector feature sets, including the vertex coordinates of the polygons, field number, field area, soil type and crop type; Obtain the field area from the above closed polygon vector feature set. If the field area is less than 1 hectare, divide the field into four 25m x 25m grids. If the field area is between 1 and 5 hectares, generate a fishing net according to 50m x 50m. If the field area is greater than 5 hectares, set the grid side length to 100m to obtain a secondary grid and number each sub-region. The secondary grid is divided into microgrids of 10m x 10m, and the microgrids are numbered to obtain the tertiary grid; Obtain the latitude and longitude of the sensor, convert the latitude and longitude to planar coordinates, map the planar coordinates into the inner side of the NDVI map, and for each sensor point, find the grid with the shortest center point distance in the three-level grid, and bind the sensor to the grid with the shortest center point distance.

6. The digital agricultural automatic control and management system according to claim 5, characterized in that: The timely replenishment of nutrients or water based on data includes: Plant images and sensor data from a three-level grid are acquired, and features from the plant images and spatial features collected by the sensors are extracted. The two sets of features are then fused to obtain a fused feature set, which is then input into a trained plant growth model to predict future plant growth outcomes. Establish a local crop variety database, which includes, but is not limited to, data on the three growth cycles of plants; The predicted future plant growth data is compared with local crop growth data. If the predicted data is better than the local data, the plant growth conditions are not changed. If the predicted data is worse than the local data, the plant growth conditions are optimized.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the digital agricultural automatic control and management system according to any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed, implements the digital agricultural automatic control and management system according to any one of claims 1 to 6.

Images