Real-time dynamic monitoring system and method for water saving in agricultural irrigation areas

Abstract

The present application discloses a real-time dynamic monitoring system and method for water conservation in agricultural irrigation areas, which uses a water level sensor to monitor the water level data in the observation area in real time to obtain the time series of the water level data, and collects the growth state image of the target crop object through a camera, and uses data analysis and image processing technology based on deep learning to perform time series implicit association analysis on the time series of the water level data, and at the same time performs mask area instance segmentation and growth state feature extraction on the growth state image of the target crop object, so as to achieve accurate recommendation of valve opening value based on the positive and negative attention interaction representation between the implicit association feature of the water level time series and the growth state feature of the target crop object. This not only helps to improve irrigation efficiency and water resource utilization, but also reduces dependence on manual intervention, promotes the transformation of water conservancy facilities from automation to intelligent and standardized operation mode, and provides more scientific and efficient management means for modern agriculture.

Description

Technical Field

[0001] The present application relates to the field of intelligent monitoring, and more specifically, to a real-time dynamic monitoring system and method for water conservation in agricultural irrigation areas. Background Art

[0002] With the modernization of the water conservancy industry, data telemetry, analysis and early warning, remote control and storage sharing have gradually become the focus of the industry. How to transform the existing manual operation mode into automation and further promote it to a standardized operation mode is the key task of current water conservancy management.

[0003] At present, water control node facilities in most areas are still in the basic operation mode of manual operation of equipment and on-site observation of water levels for adjustment. This traditional method is not only inefficient, but also easily affected by human factors, and it is difficult to meet the requirements of modern water conservancy projects for accuracy and response speed.

[0004] However, water conservancy facilities in some areas have completed automation reconstruction and realized functions such as remote control, which has significantly improved work efficiency and emergency response capabilities. Nevertheless, these automation systems still require the assistance and cooperation of on-site personnel in actual applications, especially in complex or abnormal situations. It is crucial to ensure the stability and reliability of the system.

[0005] Therefore, a real-time dynamic monitoring solution for water conservation in agricultural irrigation areas is desired, which can utilize advanced sensor technology and data analysis to reduce human intervention to achieve precise management of water resources. Summary of the invention

[0006] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a real-time dynamic monitoring system and method for water saving in agricultural irrigation areas, which uses a water level sensor to monitor the water level data in the polder area in real time to obtain the time series of the water level data, and collects the growth status image of the target crop object through a camera, and uses data analysis and image processing technology based on deep learning to perform time series implicit association analysis on the time series of the water level data, and at the same time, the growth status image of the target crop object is subjected to mask region instance segmentation and growth status feature extraction, so as to achieve accurate recommendation of valve opening value based on the positive and negative attention interaction representation between the implicit association features of the water level time series and the growth status features of the target crop object. This not only helps to improve irrigation efficiency and water resource utilization, but also reduces dependence on manual intervention, promotes the transformation of water conservancy facilities from automation to intelligent and standardized operation mode, and provides modern agriculture with more scientific and efficient management means.

[0007] According to one aspect of the present application, a real-time dynamic monitoring system for water conservation in agricultural irrigation areas is provided, which includes: a water level data acquisition module, which is used to monitor the water level data in the observation area in real time through a water level sensor to obtain a time series of the water level data; a growth status image acquisition module, which is used to collect the growth status image of the target crop object through a camera; a data image transmission module, which is used to transmit the time series of the water level data and the growth status image of the target crop object to a cloud control center; a water level time series encoding module, which is used to time series encode the time series of the water level data in the cloud control center to obtain the implicit correlation characteristics of the water level time series; the growth status image acquisition module of the target crop object A long state extraction module is used to perform crop instance mask segmentation and crop growth state feature extraction on the growth state image of the target crop object in the cloud-based central control center to obtain the growth state feature of the target crop object; a growth state-water level interaction module is used to perform feature positive and negative attention interaction response on the water level time series implicit association feature and the growth state feature of the target crop object in the cloud-based central control center to obtain the crop growth state-water level time series semantic interaction response feature; a valve opening recommendation module is used to perform irrigation valve control in the cloud-based central control center based on the crop growth state-water level time series semantic interaction response feature to obtain a valve opening recommendation value.

[0008] According to another aspect of the present application, a real-time dynamic monitoring method for water conservation in agricultural irrigation areas is provided, which includes: monitoring water level data in an observation area in real time through a water level sensor to obtain a time series of water level data; collecting a growth status image of a target crop object through a camera; transmitting the time series of water level data and the growth status image of the target crop object to a cloud-based central control center; in the cloud-based central control center, performing time series encoding on the time series of water level data to obtain water level time series implicit correlation features; in the cloud-based central control center, performing crop instance mask segmentation and crop growth status feature extraction on the growth status image of the target crop object to obtain target crop object growth status features; in the cloud-based central control center, performing feature positive and negative attention interactive responses on the water level time series implicit correlation features and the target crop object growth status features to obtain crop growth status-water level time series semantic interactive response features; in the cloud-based central control center, performing irrigation valve control based on the crop growth status-water level time series semantic interactive response features to obtain valve opening recommended values.

[0009] Compared with the prior art, the present application provides a real-time dynamic monitoring system and method for water conservation in agricultural irrigation areas, which uses a water level sensor to monitor the water level data in the polder area in real time to obtain the time series of the water level data, and collects the growth status image of the target crop object through a camera, and uses data analysis and image processing technology based on deep learning to perform time series implicit association analysis on the time series of the water level data, and at the same time, performs mask area instance segmentation and growth status feature extraction on the growth status image of the target crop object, so as to achieve accurate recommendation of the valve opening value based on the positive and negative attention interaction representation between the implicit association features of the water level time series and the growth status features of the target crop object. This not only helps to improve irrigation efficiency and water resource utilization, but also reduces dependence on manual intervention, promotes the transformation of water conservancy facilities from automation to intelligent and standardized operation mode, and provides more scientific and efficient management means for modern agriculture. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0011] Figure 1 It is a block diagram of a real-time dynamic monitoring system for water saving in agricultural irrigation areas according to an embodiment of the present application.

[0012] Figure 2 This is a data flow diagram of a real-time dynamic monitoring system for water conservation in agricultural irrigation areas according to an embodiment of the present application.

[0013] Figure 3 It is a block diagram of a growth status-water level interaction module in a real-time dynamic monitoring system for water saving in agricultural irrigation areas according to an embodiment of the present application.

[0014] Figure 4 The present invention is a flowchart of a real-time dynamic monitoring method for water saving in agricultural irrigation areas according to an embodiment of the present application. DETAILED DESCRIPTION

[0015] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0016] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0017] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0018] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0019] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0020] It should be understandable that crops at different growth stages have very different water requirements. For example, during the sowing period, crops require less water, generally keeping the soil moist, and the amount of irrigation is relatively small; while during the flowering period, the water demand reaches a peak. For example, cotton has a strong demand for water during the flowering period, and at this time, it is necessary to increase the amount of irrigation to ensure sufficient water supply and promote pollination and fruiting. In addition, different crops have different water requirements and their water requirements vary greatly. For example, rice is a water-loving crop that requires more water throughout its growth period, and the amount of irrigation is relatively large; while dry crops such as corn require relatively less water, and the amount of irrigation is also relatively small.

[0021] In addition, when the water level is high, the moisture content in the soil is relatively high, and the irrigation amount may be reduced or the irrigation interval may be extended; when the groundwater level is low, the soil moisture is easily lost, and the irrigation amount needs to be increased to replenish the soil moisture to meet the needs of crop growth. For example, in some farmlands near rivers or lakes, the groundwater level is high, and the irrigation amount is usually less than that of farmlands far away from water sources and with low groundwater levels.

[0022] In the technical solution of the present application, a real-time dynamic monitoring system for water conservation in agricultural irrigation areas is proposed. Figure 1It is a block diagram of a real-time dynamic monitoring system for water saving in agricultural irrigation areas according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of a real-time dynamic monitoring system for water conservation in agricultural irrigation areas according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the embodiment of the present application, the real-time dynamic monitoring system 300 for water conservation in agricultural irrigation areas includes: a water level data acquisition module 310, which is used to monitor the water level data in the observation area in real time through a water level sensor to obtain a time series of water level data; a growth status image acquisition module 320, which is used to collect the growth status image of the target crop object through a camera; a data image transmission module 330, which is used to transmit the time series of the water level data and the growth status image of the target crop object to the cloud control center; a water level time series encoding module 340, which is used to perform time series encoding on the time series of the water level data in the cloud control center to obtain the implicit correlation characteristics of the water level time series; the target crop The growth status extraction module 350 is used to perform crop instance mask segmentation and crop growth status feature extraction on the growth status image of the target crop object in the cloud-based central control center to obtain the growth status features of the target crop object; the growth status-water level interaction module 360 ​​is used to perform feature positive and negative attention interaction response on the water level time series implicit association features and the growth status features of the target crop object in the cloud-based central control center to obtain the crop growth status-water level time series semantic interaction response features; the valve opening recommendation module 370 is used to perform irrigation valve control in the cloud-based central control center based on the crop growth status-water level time series semantic interaction response features to obtain the valve opening recommendation value.

[0023] In particular, the water level data acquisition module 310 is used to monitor the water level data in the observation area in real time through a water level sensor to obtain a time series of water level data. It should be understood that the water level directly reflects the water content in the soil, which is crucial for evaluating the irrigation needs of crops. Different crops have different water requirements at different growth stages, and accurate water level data can help determine whether irrigation is needed and the amount of irrigation. Therefore, in the technical solution of the present application, the water level data in the observation area is monitored in real time through a water level sensor to obtain a time series of water level data.

[0024] In one example, first, several representative observation points are selected in the agricultural irrigation area. These observation points are usually located at important nodes of the irrigation system or in areas sensitive to water level changes. A specially designed water level sensor is installed at each observation point. This type of sensor can accurately sense and record the water level information at the location. It is worth mentioning that in order to ensure the accuracy and reliability of the data, the selected water level sensor must have the characteristics of high precision, stable performance and strong anti-interference ability. Then, the water level sensor continuously detects the water level at a preset time interval and converts the data measured each time into an electrical signal or other form of signal. These signals are then transmitted to a local data acquisition terminal, which is responsible for collecting data from multiple sensors and performing preliminary processing, such as removing noise, calibrating deviations, etc. The processed water level data will form a series of ordered values, that is, the time series of the water level data, each of which corresponds to the water level state at a specific moment.

[0025] In particular, the growth status image acquisition module 320 is used to acquire the growth status image of the target crop object through a camera. It should be understood that the appearance characteristics of the crop, such as the color, shape and size of the leaves, are important indicators for measuring the growth status of the crop. Different types of crops and their different growth stages have very different water requirements. By obtaining the specific growth stage information of the crop through image recognition technology, the irrigation strategy can be adjusted more accurately to ensure that the actual water requirements of the crop are met without causing waste. Therefore, in the technical solution of the present application, the growth status image of the target crop object is acquired by a camera.

[0026] In one example, first, suitable locations are selected in agricultural irrigation areas to install cameras. These locations are usually located in places that can cover the main crop planting areas and ensure that the camera's field of view can clearly capture all parts of the crop, including key parts such as leaves, stems and fruits. In order to meet the shooting needs under different environmental conditions, the selected camera not only needs to have high resolution to obtain detailed images, but also needs to be able to work under low light conditions and have functions such as waterproof and dustproof to ensure long-term stable operation. Then, the camera automatically takes pictures according to the preset time interval or trigger conditions. This time interval can be adjusted according to the type of crop and its growth rate. For example, a more frequent image acquisition frequency may be required during a rapid growth period. For some specific cases, such as detecting pests and diseases, the sensor linkage can be set to immediately trigger the camera to take pictures when an abnormal situation is detected. The purpose of this is to capture subtle changes in the growth process of crops in a timely manner and provide detailed data support for subsequent analysis.

[0027] In particular, the data image transmission module 330 is used to transmit the time series of the water level data and the growth status image of the target crop object to the cloud control center. That is, in one example, first, various sensors (such as water level sensors) and cameras deployed in the agricultural irrigation area will collect data according to predetermined time intervals or specific trigger conditions. These devices are usually distributed in different locations in the field to ensure that the entire monitoring area is covered and representative information is obtained. Each collected data will be timestamped and accompanied by corresponding geographic location information so that it can be accurately associated with a specific location and time in subsequent analysis. Next, the collected data needs to undergo preliminary data preprocessing. Specifically, for the water level sensor, this means calibrating the original reading to eliminate possible measurement errors; and the image captured by the camera must be compressed to reduce the file size while maintaining sufficient resolution for subsequent image recognition and feature extraction. In addition, some basic quality control checks will be performed, such as detecting whether there are obvious outliers or blurred images, so as to ensure the quality of the data uploaded to the cloud. Then, choose a suitable communication method to efficiently and stably transmit the pre-processed data to the cloud control center. In modern agricultural environments, wireless network technologies such as Wi-Fi, 4G / 5G cellular networks, and even low-power wide area networks such as LoRaWAN are common choices. According to the specific geographical environment and cost considerations, the most suitable communication method can be selected. For example, in remote areas or large areas of farmland, satellite communication may be the only feasible way; while in places with good network infrastructure, high-speed and stable ground networks can be used to achieve fast data transmission. It is worth mentioning that in the actual transmission process, the data will be packaged into messages that conform to a specific protocol format, which not only helps to ensure the correct routing of data packets in the network, but also facilitates parsing by the cloud receiving end. Commonly used protocols include MQTT (Message Queuing Telemetry Transport), HTTP(S), CoAP (Constrained Application Protocol), etc. Different protocols are suitable for different application scenarios. For example, MQTT is widely used in the field of Internet of Things due to its lightweight characteristics and good QoS (quality of service) guarantee; and for one-time large-scale data upload, HTTPS provides higher security and reliability. Once the data is successfully uploaded to the cloud control center, it enters the data processing process inside the cloud platform.

[0028] In particular, the water level time series encoding module 340 is used to perform time series encoding on the time series of the water level data in the cloud control center to obtain the implicit correlation characteristics of the water level time series. That is, in the technical solution of the present application, the time series of the water level data is passed through a water level time series feature extractor based on a void causal convolutional network to obtain a water level time series implicit correlation feature vector as the water level time series implicit correlation feature. Considering that the water level data exists in the form of a time series, it has a temporal sequence and potential dependencies, that is, the water level data has time series feature information and associations on a time scale. However, considering that ordinary convolution operations may be difficult to fully explore the time series features within a long time span due to reasons such as a fixed receptive field. Based on this, the present application uses a water level time series feature extractor based on a void causal convolutional network to perform time series feature analysis on the time series of the water level data to effectively capture the long-term dependencies in the time series and obtain a water level time series implicit correlation feature vector. It should be understandable that the dilated causal convolutional network has a unique advantage in processing time series data. It can effectively capture the correlation between water level data at different time intervals on the basis of considering the time sequence (causality ensures that only the data of the past and current moments are used, and no future information is introduced, which conforms to the actual time series analysis logic). In other words, the dilated convolution can expand the receptive field by setting different dilation rates, better capture the implicit correlation of long-distance time series, and clearly analyze the changing rules of water levels at different time stages and the deep correlation between water levels at different times.

[0029] In particular, the target crop growth state extraction module 350 is used to perform crop instance mask segmentation and crop growth state feature extraction on the growth state image of the target crop object in the cloud control center to obtain the growth state feature of the target crop object. That is, in the technical solution of the present application, first, the growth state image of the target crop object is passed through a crop instance segmenter based on a masked region convolutional neural network to obtain a target crop object growth state mask image. Considering that in the actual scene of the agricultural irrigation area, the growth state image of the target crop object collected by the camera often contains the crop itself and complex background information, such as soil, field paths, surrounding weeds, etc. Therefore, in order to accurately segment the target crop from the complex background, provide pure image data that only focuses on the individual crop for the subsequent analysis of the growth state of the crop itself, and avoid interference caused by background factors, in the technical solution of the present application, the growth state image of the target crop object is passed through a crop instance segmenter based on a masked region convolutional neural network to obtain a target crop object growth state mask image. It can be understood that the masked region convolutional neural network (MaskR-CNN) is an instance segmentation algorithm based on a convolutional neural network. It can also perform pixel-level segmentation on each detected target and generate a mask (i.e., a binary image) for each instance with the same resolution as the input image or scaled proportionally. Since it can provide pixel-level masks, it can accurately locate the specific position and shape of crops, thus enabling more refined agricultural management, such as precise irrigation for specific crops or regions.

[0030] In one example, first, the growth state image of the target crop object is preprocessed, including adjusting the image size to adapt to the model input requirements, standardizing the pixel value distribution and other operations to ensure that the image data meets the requirements of the algorithm, thereby improving the detection accuracy of the model; next, MaskR-CNN uses its unique architecture to process the growth state image of the target crop object. The network consists of a basic network (usually ResNet or a similar deep convolutional network) and two branches: a region proposal network for generating bounding box proposals, and the other is a task-specific head for performing classification and mask prediction. The basic network is responsible for extracting the growth state features of the target crop object in the growth state image of the target crop object, and the RPN proposes candidate regions that may contain crop instances based on these features. For each proposed region, the task-specific head further refines the position of the bounding box and predicts a binary mask for each instance, that is, a mask with the same resolution or scaled as the growth state image of the target crop object, which is used to represent the specific position and shape of the crop. It is worth mentioning that in MaskR-CNN, efficient feature reuse is achieved by sharing convolutional layers, so that the model can complete target detection and pixel-level segmentation at the same time. This design not only improves computational efficiency, but also enhances the accuracy of the model because it allows the network to learn richer spatial information. For agricultural applications, this means being able to more accurately locate crops and distinguish individual plants even in densely planted areas.

[0031] Then, the target crop object growth state mask image is passed through the crop growth state feature extractor based on the CNN-LSTM model to obtain the target crop object growth state feature vector as the target crop object growth state feature. Considering that the target crop object growth state mask image contains the spatial morphological information of the crop at a specific moment, such as the shape of the crop, leaf distribution, geometric relationship of different parts, etc., and there is a mutual implicit correlation and influence between different local spatial position information. Considering that CNN is good at extracting spatial features of images, it can capture the shape, texture and other detailed features of the crop in the mask image. Specifically, CNN can effectively extract these spatial features by sliding the convolution kernel in the convolution layer on the image for convolution operation. For example, the lower layer convolution kernel can capture basic features such as edges and textures, and as the number of network layers increases, the high-level convolution kernel can obtain more abstract semantic features, such as the overall outline and structure of the crop. These spatial features are very critical for understanding the growth state of the crop at a certain moment. LSTM performs well in processing data dependencies. Based on this, in the technical solution of the present application, the growth status mask image of the target crop object is passed through a crop growth status feature extractor based on a CNN-LSTM model to extract visual features closely related to the health status of the crop while capturing the correlation between the local growth information of each crop, thereby obtaining a growth status feature vector of the target crop object.

[0032] In particular, the growth status-water level interaction module 360 ​​is used to perform feature positive and negative attention interaction responses on the implicit association features of the water level time series and the growth status features of the target crop object in the cloud control center to obtain crop growth status-water level time series semantic interaction response features. Considering that the implicit association features of the water level time series reflect the inherent laws and hidden associations of the water level changes over time in agricultural irrigation areas, these associations involve the dynamic situation of the water level under the influence of multiple factors such as irrigation water, precipitation, and soil infiltration. The growth status features of the target crop object cover the status information of crops in different aspects (such as morphology, health, growth rate, etc.), which are affected by the irrigation conditions brought about by water level changes, and are also fed back to the irrigation demand. There is a complex and mutually influential relationship between the two, but this relationship is implicit and difficult to analyze directly. Based on this, this application introduces a feature positive and negative attention interactive response mechanism to interactively respond to the implicit correlation features of the water level time series and the growth status features of the target crop object to use positive and negative attention to dig out deep and complex semantic associations, such as how the current growth status features of the crop indicate the need for subsequent water level regulation, etc., which helps to more comprehensively grasp the internal connection between the two and provide a more realistic basis for subsequent accurate irrigation decisions. Specifically, in a specific example of this application, such as Figure 3As shown, the growth state-water level interaction module 360 ​​includes: a homography projection transformation unit 361, which is used to perform homography projection transformation on the water level time series implicit association feature vector and the target crop object growth state feature vector to obtain a water level time series implicit association homography projection coding vector and a target crop object growth state homography projection coding vector; a bidirectional attention balance field construction unit 362, which is used to construct a positive and negative growth state-water level bidirectional attention balance field between the water level time series implicit association homography projection coding vector and the target crop object growth state homography projection coding vector; a growth state-water level time series attention modulation response unit 363, which is used to perform a growth state-water level time series modulation response on the water level time series implicit association homography projection coding vector and the target crop object growth state homography projection coding vector based on the positive and negative growth state-water level bidirectional attention balance field to obtain a crop growth state-water level time series semantic interaction response coding vector as the crop growth state-water level time series semantic interaction response feature.

[0033] Specifically, the homography projection transformation unit 361 is used to perform homography projection transformation on the water level time series implicit association feature vector and the target crop object growth state feature vector to obtain a water level time series implicit association homography projection coding vector and a target crop object growth state homography projection coding vector. Since the water level time series implicit association feature vector and the target crop object growth state feature vector may have different scales or distributions, direct comparison or combination may lead to inaccurate results. By mapping the water level time series implicit association feature vector and the target crop object growth state feature vector to a new feature space, the two feature vectors can be compared or interacted in the same space to capture the potential relationship between the two. This helps to more accurately identify the correlation pattern between water level changes and crop growth status in the future. More specifically, the water level time series implicit association feature vector and the target crop object growth state feature vector are homographically projected to obtain a water level time series implicit association homographic projection coding vector and a target crop object growth state homographic projection coding vector using the following homographic projection transformation formula; wherein the homographic projection transformation formula is: ;in, is the implicit correlation feature vector of the water level time series, is the growth state feature vector of the target crop object, is the homography matrix of water level time series mapping, is the target crop mapping homography matrix, is the homography projection encoding vector of the implicit association of water level time series, is the homography projection encoding vector of the growth state of the target crop object.

[0034] Specifically, the bidirectional attention balance field construction unit 362 is used to construct a positive and negative growth state-water level bidirectional attention balance field between the water level time series implicitly associated homography projection coding vector and the target crop object growth state homography projection coding vector. In a specific example of the present application, first, the positive growth state-water level attention score field of the water level time series implicitly associated homography projection coding vector relative to the target crop object growth state homography projection coding vector is calculated to evaluate the importance distribution of the water level time series implicitly associated homography projection coding vector relative to the target crop object growth state homography projection coding vector. The core of the positive attention mechanism is to quantify the correlation between the two coding vectors, so that the model can focus on the part of the water level time series implicitly associated homography projection coding vector that best reflects the correlation with the target crop object growth state homography projection coding vector, and at the same time act as a filter to filter out irrelevant noise information. Next, the reverse growth state-water level attention score field of the target crop object growth state homography projection coding vector relative to the water level time series implicitly associated homography projection coding vector is calculated. This process provides an opportunity to understand the relationship between the growth state-water level features from the opposite direction, ensuring that important interaction details are not missed. In fact, the positive and negative attention score fields work together to form a more complete and detailed relationship map, which helps to capture asymmetric relationships. Furthermore, in order to integrate the information of the positive and negative attention score fields and comprehensively consider the mutual influence between the two feature vectors, in the technical solution of the present application, the positive and negative growth state-water level bidirectional attention balance field between the positive growth state-water level attention score field and the negative growth state-water level attention score field is constructed. In this way, it can be ensured that in the final interactive response encoding, neither one side is overly biased nor the contribution of any party is ignored. More specifically, the positive and negative growth state-water level bidirectional attention balance field between the water level time series implicit association homography projection coding vector and the target crop object growth state homography projection coding vector is constructed using the following attention balance field construction formula; wherein, the attention balance field construction formula is: ;in, yes The transposed vector of yes The transposed vector of is matrix multiplication, yes Length, yes Length, It is a positive growth state - water level attention score field, It is the reverse growth state - water level attention score field, is the feature concatenation operation, is a convolutional code with a convolution kernel of 3×3. It is a positive and negative growth state-water level two-way attention balance field.

[0035] Specifically, the growth state-water level time series attention modulation response unit 363 is used to perform growth state-water level time series modulation response on the water level time series implicitly associated homography projection coding vector and the target crop object growth state homography projection coding vector based on the positive and negative growth state-water level bidirectional attention balance field to obtain the crop growth state-water level time series semantic interaction response coding vector as the crop growth state-water level time series semantic interaction response feature. In a specific example of the present application, first, the water level time series implicitly associated homography projection coding vector and the target crop object growth state homography projection coding vector are respectively mapped to the positive and negative growth state-water level bidirectional attention balance field to obtain the water level time series implicitly associated homography projection attention modulation coding vector and the target crop object growth state homography projection attention modulation coding vector. It should be understood that when the coding vector is mapped to the positive and negative bidirectional attention balance field constructed above, a reweighting operation is actually performed. The modulated coding vector not only contains the information of the original feature, but also integrates the influence of the feature from the other party. In this way, the model can understand and represent the complex interaction between the growth state and water level features at a higher level of abstraction, ensuring that each encoding vector adjusts itself according to the guidance provided by the balance field. Further, the response between the water level time series implicitly associated homography projection attention modulation encoding vector and the target crop object growth state homography projection attention modulation encoding vector is calculated to obtain the crop growth state-water level time series semantic interaction response encoding vector. Specifically, the position point division between the water level time series implicitly associated homography projection attention modulation encoding vector and the target crop object growth state homography projection attention modulation encoding vector is calculated to obtain the crop growth state-water level time series semantic interaction response encoding vector. Among them, by dividing by position point, the water level time series implicitly associated homography projection attention modulation encoding vector and the target crop object growth state homography projection attention modulation encoding vector are compared to reveal the new relationship between the two feature vectors after their respective modulations. This deep feature comparison and integration reveals the subtle proportional relationship between the crop growth state-water level features, providing a solid foundation for the accurate recommendation of valve opening value. More specifically, based on the positive and negative growth state-water level bidirectional attention balance field, the water level temporal implicit association homography projection coding vector and the target crop object growth state homography projection coding vector are subjected to growth state-water level temporal modulation response using the following modulation response formula to obtain a crop growth state-water level temporal semantic interaction response coding vector; wherein the modulation response formula is: ;in, is the water level temporal implicit correlation homography projection attention modulation encoding vector, is the target crop object growth state homography projection attention modulation encoding vector, It is the crop growth status-water level temporal semantic interaction response encoding vector.

[0036] In particular, the valve opening recommendation module 370 is used to control the irrigation valve in the cloud control center based on the crop growth state-water level time series semantic interaction response feature to obtain the valve opening recommendation value. In the technical solution of the present application, the crop growth state-water level time series semantic interaction response encoding vector is passed through the irrigation valve controller based on a decoder to obtain the valve opening recommendation value. That is, the crop growth state-water level time series semantic interaction response feature obtained by the positive and negative attention interaction of the water level time series implicit association feature vector and the target crop object growth state feature vector is decoded and processed to achieve accurate recommendation of the valve opening value.

[0037] Here, when the water level time series implicit correlation feature vector and the target crop object growth state feature vector represent the time series correlation features of the water level data and the short-range-long-range bidirectional correlation mask image semantic features of the growth state image of the target crop object respectively, when performing interactive responses based on feature positive and negative attention fields, the complexity of the positive and negative attention field mechanisms of different modal features will lead to insufficient representation of the long-range interactive response of the crop growth state-water level time series semantic interactive response coding vector, thereby reducing the expression effect of the crop growth state-water level time series semantic interactive response coding vector and affecting the accuracy of the valve opening recommendation value obtained by the decoder-based irrigation valve controller.

[0038] Preferably, in one example, the crop growth state-water level time series semantic interaction response coding vector is passed through a decoder-based irrigation valve controller to obtain a valve opening recommendation value, including: determining the number of super-distributed eigenvalues ​​in the crop growth state-water level time series semantic interaction response coding vector whose difference from the eigenvalue mean is greater than a threshold, and respectively calculating the reciprocal of the logarithm of the number of super-distributed eigenvalues ​​with base two and the exponential value of the reciprocal of the number of super-distributed eigenvalues ​​with base natural constants to obtain a first crop growth state-water level time series semantic interaction response field cell representation value and the second crop growth state-water level time series semantic interaction response field cell representation value : ;in, represents the crop growth state-water level temporal semantic interaction response encoding vector, The first one represents the crop growth state-water level temporal semantic interaction response encoding vector eigenvalues, represents the mean of all eigenvalues ​​of the crop growth state-water level temporal semantic interaction response encoding vector, Representation calculation The number of is the number of super-distributed eigenvalues ​​in the crop growth state-water level time series semantic interaction response encoding vector, and is the threshold hyperparameter.

[0039] Calculate the hyperbolic sine function value of the sum of squares of all eigenvalues ​​of the crop growth state-water level time series semantic interaction response encoding vector: ;in, represents the length of the crop growth state-water level temporal semantic interaction response encoding vector, represents the hyperbolic sine function, Represents the value of the hyperbolic sine function.

[0040] And calculate the exponential value of the hyperbolic sine function value with the natural constant as the base and then divide it by the square of the length of the crop growth state-water level time series semantic interaction response encoding vector to obtain the crop growth state-water level time series semantic interaction response multidimensional field manifold value ,in, represents a natural constant, Represents the multidimensional field manifold value of crop growth state-water level temporal semantic interaction response.

[0041] Calculate the multidimensional field manifold value of the crop growth state-water level time series semantic interaction response Relative to the first crop growth state-water level time series semantic interaction response field cell representation value and the second crop growth state-water level time series semantic interaction response field cell representation value The integrated value of the covariance of the semantic interaction response of the crop growth state-water level time series is: ;in, Represents the covariance integration value of the semantic interaction response between crop growth status and water level time series.

[0042] The power function feature vector of the crop growth state-water level time series semantic interaction response encoding vector with the crop growth state-water level time series semantic interaction response covariance integration value as the inverse is calculated, and the autocorrelation matrix of the power function feature vector is calculated to obtain the crop growth state-water level time series semantic interaction response adaptive representation matrix, that is: ;in, represents a power function feature vector with the inverse of the crop growth state-water level time series semantic interaction response covariance integration value for calculating the crop growth state-water level time series semantic interaction response encoding vector, represents matrix multiplication, represents the transpose of a vector, Represents the adaptive representation matrix of crop growth status-water level temporal semantic interaction response.

[0043] The crop growth state-water level time series semantic interaction response coding vector is matrix-multiplied with the crop growth state-water level time series semantic interaction response adaptive representation matrix to obtain an optimized crop growth state-water level time series semantic interaction response coding vector, that is: Here, the crop growth state-water level time series semantic interaction response encoding vector is a row vector, Represents the optimized crop growth state-water level temporal semantic interaction response encoding vector.

[0044] The optimized crop growth state-water level time series semantic interaction response encoding vector is passed through a decoder-based irrigation valve controller to obtain a valve opening recommendation value.

[0045] That is, in view of the lack of global adaptive mapping mechanism caused by cross-distributed field associations in the attribute clusters of the crop growth state-water level time series semantic interaction response coding vector under the preset parameter discrete arrangement, a multi-dimensional tensor field cell representation framework is constructed by the mutual covariance coupling driving scheme of the crop growth state-water level time series semantic interaction response coding vector, and the deep interaction law of its parameter space is revealed by analyzing the fully connected topological association composite pattern. Therefore, by adopting the gradient-based multi-dimensional field implicit factor simulation method, a parameter space nonlinear association modeling system is established, thereby realizing the topological reconstruction of the probabilistic trajectory evolution pattern of the crop growth state-water level time series semantic interaction response coding vector, and improving the accuracy of the valve opening recommendation value obtained by the decoder-based irrigation valve controller.

[0046] As described above, the real-time dynamic monitoring system 300 for water conservation in agricultural irrigation areas according to the embodiment of the present application can be implemented in various wireless terminals, such as a server having a real-time dynamic monitoring algorithm for water conservation in agricultural irrigation areas. In a possible implementation, the real-time dynamic monitoring system 300 for water conservation in agricultural irrigation areas according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the real-time dynamic monitoring system 300 for water conservation in agricultural irrigation areas can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the real-time dynamic monitoring system 300 for water conservation in agricultural irrigation areas can also be one of the many hardware modules of the wireless terminal.

[0047] Alternatively, in another example, the real-time dynamic monitoring system 300 for water conservation in agricultural irrigation areas and the wireless terminal may be separate devices, and the real-time dynamic monitoring system 300 for water conservation in agricultural irrigation areas may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0048] Furthermore, a real-time dynamic monitoring method for water conservation in agricultural irrigation areas is also provided.

[0049] Figure 4 Flow chart of a real-time dynamic monitoring method for water conservation in agricultural irrigation areas according to an embodiment of the present application. Figure 4 As shown, according to the embodiment of the present application, the real-time dynamic monitoring method for water saving in agricultural irrigation areas includes the following steps: S1, real-time monitoring of water level data in the observation area by a water level sensor to obtain a time series of water level data; S2, collecting a growth status image of a target crop object by a camera; S3, transmitting the time series of the water level data and the growth status image of the target crop object to a cloud control center; S4, in the cloud control center, time-series encoding the time series of the water level data to obtain a water level time series implicit correlation feature; S5, in the cloud control center, crop instance mask segmentation and crop growth status feature extraction are performed on the growth status image of the target crop object to obtain a target crop object growth status feature; S6, in the cloud control center, feature positive and negative attention interactive response is performed on the water level time series implicit correlation feature and the target crop object growth status feature to obtain a crop growth status-water level time series semantic interactive response feature; S7, in the cloud control center, irrigation valve control is performed based on the crop growth status-water level time series semantic interactive response feature to obtain a valve opening recommendation value.

[0050] In summary, according to the embodiment of the present application, a real-time dynamic monitoring method for water conservation in agricultural irrigation areas is explained, which uses a water level sensor to monitor the water level data in the polder area in real time to obtain a time series of water level data, and collects the growth status image of the target crop object through a camera, and uses data analysis and image processing technology based on deep learning to perform time series implicit association analysis on the time series of the water level data, and at the same time, performs mask area instance segmentation and growth status feature extraction on the growth status image of the target crop object, so as to achieve accurate recommendation of valve opening value based on the positive and negative attention interaction representation between the implicit association features of the water level time series and the growth status features of the target crop object. This not only helps to improve irrigation efficiency and water resource utilization, but also reduces dependence on manual intervention, promotes the transformation of water conservancy facilities from automation to intelligent and standardized operation modes, and provides more scientific and efficient management means for modern agriculture.

[0051] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A real-time dynamic monitoring system for water saving in agricultural irrigation areas, characterized in that: include: The water level data acquisition module is used to monitor the water level data in the observation area in real time through the water level sensor to obtain the time series of the water level data; the growth status image acquisition module is used to collect the growth status image of the target crop object through the camera; the data image transmission module is used to transmit the time series of the water level data and the growth status image of the target crop object to the cloud control center; the water level time series encoding module is used to perform time series encoding on the time series of the water level data in the cloud control center to obtain the implicit correlation characteristics of the water level time series; the target crop growth status extraction module is used to perform time series encoding on the time series of the water level data in the cloud control center , performing crop instance mask segmentation and crop growth state feature extraction on the growth state image of the target crop object to obtain the growth state feature of the target crop object; a growth state-water level interaction module, used in the cloud control center, to perform feature positive and negative attention interaction response on the water level time series implicit association feature and the growth state feature of the target crop object to obtain the crop growth state-water level time series semantic interaction response feature; a valve opening recommendation module, used in the cloud control center, to control the irrigation valve based on the crop growth state-water level time series semantic interaction response feature to obtain the valve opening recommendation value.

2. The real-time dynamic monitoring system for water saving in agricultural irrigation areas according to claim 1 is characterized in that: The water level time series encoding module is used to: pass the time series of the water level data through a water level time series feature extractor based on a void causal convolutional network to obtain a water level time series implicit correlation feature vector as the water level time series implicit correlation feature.

3. The real-time dynamic monitoring system for water saving in agricultural irrigation areas according to claim 2 is characterized in that: The target crop growth state extraction module includes: a target crop object growth state mask unit, which is used to pass the growth state image of the target crop object through a crop instance segmenter based on a mask region convolutional neural network to obtain a target crop object growth state mask image; and a target crop growth state feature extraction unit, which is used to pass the target crop object growth state mask image through a crop growth state feature extractor based on a CNN-LSTM model to obtain a target crop object growth state feature vector as the target crop object growth state feature.

4. The real-time dynamic monitoring system for water saving in agricultural irrigation areas according to claim 3 is characterized in that: The growth state-water level interaction module includes: a homography projection transformation unit, which is used to perform homography projection transformation on the water level time series implicit association feature vector and the target crop object growth state feature vector to obtain a water level time series implicit association homography projection coding vector and a target crop object growth state homography projection coding vector; a bidirectional attention balance field construction unit, which is used to construct a positive and negative growth state-water level bidirectional attention balance field between the water level time series implicit association homography projection coding vector and the target crop object growth state homography projection coding vector; a growth state-water level time series attention modulation response unit, which is used to perform growth state-water level time series modulation response on the water level time series implicit association homography projection coding vector and the target crop object growth state homography projection coding vector based on the positive and negative growth state-water level bidirectional attention balance field to obtain a crop growth state-water level time series semantic interaction response coding vector as the crop growth state-water level time series semantic interaction response feature.

5. The real-time dynamic monitoring system for water saving in agricultural irrigation areas according to claim 4 is characterized in that: The bidirectional attention balance field construction unit is used to: calculate the forward growth state-water level attention score field of the water level time series implicitly associated homography projection coding vector relative to the target crop object growth state homography projection coding vector; calculate the reverse growth state-water level attention score field of the target crop object growth state homography projection coding vector relative to the water level time series implicitly associated homography projection coding vector; and construct the forward and reverse growth state-water level bidirectional attention balance field between the forward growth state-water level attention score field and the reverse growth state-water level attention score field.

6. The real-time dynamic monitoring system for water saving in agricultural irrigation areas according to claim 5 is characterized in that: The growth state-water level temporal attention modulation response unit includes: a homography projection attention modulation subunit, which is used to map the water level temporal implicitly associated homography projection coding vector and the target crop object growth state homography projection coding vector to the positive and negative growth state-water level bidirectional attention balance field respectively to obtain the water level temporal implicitly associated homography projection attention modulation coding vector and the target crop object growth state homography projection attention modulation coding vector; a growth state-water level response calculation subunit, which is used to calculate the response between the water level temporal implicitly associated homography projection attention modulation coding vector and the target crop object growth state homography projection attention modulation coding vector to obtain the crop growth state-water level temporal semantic interaction response coding vector.

7. The real-time dynamic monitoring system for water saving in agricultural irrigation areas according to claim 6 is characterized in that: The growth status-water level response calculation subunit is used to calculate the water level time series implicit association homography projection attention modulation coding vector and the target crop object growth status homography projection attention modulation coding vector by dividing them by position points to obtain the crop growth status-water level time series semantic interaction response coding vector.

8. The real-time dynamic monitoring system for water saving in agricultural irrigation areas according to claim 7 is characterized in that: The valve opening recommendation module is used to: pass the crop growth state-water level time series semantic interaction response encoding vector through a decoder-based irrigation valve controller to obtain the valve opening recommendation value.

9. A real-time dynamic monitoring method for water conservation in agricultural irrigation areas, characterized in that: include: The water level data in the polder area is monitored in real time by a water level sensor to obtain a time series of the water level data; the growth status image of the target crop object is collected by a camera; the time series of the water level data and the growth status image of the target crop object are transmitted to a cloud control center; in the cloud control center, the time series of the water level data is time-series encoded to obtain water level time series implicit correlation features; in the cloud control center, the growth status image of the target crop object is subjected to crop instance mask segmentation and crop growth status feature extraction to obtain target crop object growth status features; in the cloud control center, the water level time series implicit correlation features and the target crop object growth status features are subjected to feature positive and negative attention interactive responses to obtain crop growth status-water level time series semantic interactive response features; in the cloud control center, irrigation valve control is performed based on the crop growth status-water level time series semantic interactive response features to obtain valve opening recommended values.

Images