Urban garden greening management data processing system and method based on cloud platform

Abstract

The invention relates to the field of greening management, and particularly discloses an urban landscaping management data processing system and method based on a cloud platform, and the method comprises the steps: firstly obtaining the current plant growth environment data obtained by the cloud platform, different plant moisture demand management data obtained by the cloud platform, and a plant growth image collected by a camera; the deep learning technology is used for carrying out feature extraction and correlation analysis on the three plants, and a plant irrigation suggestion is generated through a generator, so that a customized irrigation scheme is obtained according to the growth environment and water demand characteristics of each plant, excessive irrigation or insufficient irrigation is effectively avoided, water resources are saved, healthy growth of the plants is improved, and the method is suitable for large-scale popularization and application. And the efficiency and precision of landscaping management are improved.

Description

Technical Field

[0001] This application relates to the field of greening management, and more specifically, to a data processing system and method for urban landscaping management based on a cloud platform. Background Art

[0002] A garden is a beautiful natural environment and recreation area created by applying engineering techniques and artistic means in a certain area, through means such as transforming the terrain or further planting trees, flowers, and plants, constructing buildings, and arranging garden paths. As the functions of urban gardens gradually transition from traditional beautification, viewing, and recreation to comprehensive functions such as ecological protection and environmental improvement, the greening area in urban gardens is continuously increasing, and the types of seedlings are becoming more and more diverse. In view of this situation, the greening management of urban gardens has become increasingly important.

[0003] However, at present, most urban landscaping uses garden sprinkler trucks or automatic irrigation systems for watering. Although this method improves work efficiency to a certain extent, it can only be watered on schedule. Since the water shortage degree of plants cannot be known, plants can only be watered with approximately the same amount of water equally, resulting in the embarrassing situation of insufficient irrigation in some areas and excessive irrigation in some areas. Moreover, over-irrigation or under-irrigation of some vegetation not only wastes manpower and material resources but also consumes a large amount of water resources.

[0004] Therefore, a data processing system and method for urban landscaping management based on a cloud platform are desired. Summary of the Invention

[0005] In order to solve the above technical problems, this application is proposed. Embodiments of this application provide a data processing system and method for urban landscaping management based on a cloud platform. First, it obtains the current plant growth environment data obtained from the cloud platform, the water demand management data of different plants obtained from the cloud platform, and the plant growth images collected by a camera. Using deep learning technology, it performs feature extraction and correlation analysis on the three, and through a generator, generates plant irrigation suggestions, so as to obtain a customized irrigation plan according to the growth environment and water demand characteristics of each plant, effectively avoiding over-irrigation or under-irrigation, saving water resources and improving the healthy growth of plants, and improving the efficiency and accuracy of landscaping management.

[0006] According to one aspect of this application, a data processing system for urban landscaping management based on a cloud platform is provided, which includes:

[0007] An urban landscaping management data acquisition module, configured to obtain the current plant growth environment data obtained from the cloud platform, the water demand management data of different plants obtained from the cloud platform, and the plant growth images collected by a camera;

[0008] An urban landscaping management data extraction module, configured to extract plant growth multi-modal correlation feature vectors and plant growth attention fully-connected feature vectors from the plant current growth environment data obtained from the cloud platform, the different plant water requirement management data obtained from the cloud platform, and the plant growth images collected by the camera;

[0009] A plant irrigation recommendation generation module, configured to generate plant irrigation recommendations based on the plant growth multi-modal correlation feature vectors and the plant growth attention fully-connected feature vectors.

[0010] According to another aspect of the present application, there is provided a method for processing urban landscaping management data based on a cloud platform, which includes:

[0011] Obtain the plant current growth environment data obtained from the cloud platform, the different plant water requirement management data obtained from the cloud platform, and the plant growth images collected by the camera;

[0012] Extract plant growth multi-modal correlation feature vectors and plant growth attention fully-connected feature vectors from the plant current growth environment data obtained from the cloud platform, the different plant water requirement management data obtained from the cloud platform, and the plant growth images collected by the camera;

[0013] Generate plant irrigation recommendations based on the plant growth multi-modal correlation feature vectors and the plant growth attention fully-connected feature vectors.

[0014] Compared with the prior art, the urban landscaping management data processing system and method based on a cloud platform provided by the present application first obtain the plant current growth environment data obtained from the cloud platform, the different plant water requirement management data obtained from the cloud platform, and the plant growth images collected by the camera, use deep learning technology to perform feature extraction and correlation analysis on the three, and through a generator, generate plant irrigation recommendations, so as to obtain a customized irrigation plan according to the growth environment and water requirement characteristics of each plant, effectively avoid over-irrigation or under-irrigation, save water resources and improve the healthy growth of plants, and improve the efficiency and accuracy of landscaping management. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] By describing the embodiments of the present application in more detail in conjunction with the drawings, the above and other objects, features and advantages of the present application will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the present application together with the embodiments of the present application, and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1It is a block diagram of an urban landscaping management data processing system based on a cloud platform according to an embodiment of the present application.

[0017] Figure 2 It is a block diagram of an urban landscaping management data extraction module in an urban landscaping management data processing system based on a cloud platform according to an embodiment of the present application.

[0018] Figure 3 It is a block diagram of a plant irrigation recommendation generation module in an urban landscaping management data processing system based on a cloud platform according to an embodiment of the present application.

[0019] Figure 4 It is a flowchart of an urban landscaping management data processing method based on a cloud platform according to an embodiment of the present application. Detailed implementation manners

[0020] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0021] Figure 1 It is a block diagram of an urban landscaping management data processing system based on a cloud platform according to an embodiment of the present application. As Figure 1 shown, an urban landscaping management data processing system 100 based on a cloud platform according to an embodiment of the present application includes: an urban landscaping management data acquisition module 110, configured to acquire plant current growth environment data obtained from the cloud platform, different plant water requirement management data obtained from the cloud platform, and plant growth images collected by a camera; an urban landscaping management data extraction module 120, configured to extract a plant growth multi-modal correlation feature vector and a plant growth attention fully connected feature vector from the plant current growth environment data obtained from the cloud platform, the different plant water requirement management data obtained from the cloud platform, and the plant growth images collected by the camera; a plant irrigation recommendation generation module 130, configured to generate a plant irrigation recommendation based on the plant growth multi-modal correlation feature vector and the plant growth attention fully connected feature vector.

[0022] In the above-mentioned urban landscaping management data processing system 100 based on the cloud platform, the urban landscaping management data acquisition module 110 is used to obtain the current plant growth environment data obtained from the cloud platform, the water demand management data of different plants obtained from the cloud platform, and the plant growth images collected by the camera. It should be understood that the current plant growth environment data (such as temperature, humidity, light, etc.) can reflect the specific conditions of plant growth, which is crucial for determining whether the plant needs irrigation. Secondly, the water demand management data of different plants provides information on the amount of water required for specific plant species, ensuring that each plant can obtain an appropriate amount of water and avoiding the problems of over-irrigation or under-irrigation. Finally, the plant growth images collected by the camera are used to visually monitor the growth status of the plant. Analyzing these images using deep learning technology can identify the plant health status and its response to environmental changes. Among them, the current plant growth environment data and the water demand management data of different plants can be collected through the sensor network integrated in the urban garden and the cloud platform. These sensors are distributed throughout the garden, real-time monitoring environmental parameters and uploading the data to the cloud platform for storage and processing. For the water demand management data of different plants, it can be provided through database query or expert system, which is established based on a large number of experimental studies and historical data and can accurately give the water demand of various plants at different stages of their life cycles. As for the plant growth images collected by the camera, they are completed by the high-definition cameras installed in the garden. These cameras regularly take photos or videos of the plants and transmit them to the central server for analysis. In this way, a customized irrigation plan can be formulated according to the specific growth environment and water demand characteristics of each plant, not only effectively saving water resources, but also improving the healthy growth level of the plants and enhancing the overall efficiency and accuracy of urban landscaping management.

[0023] Specifically, the management of urban landscaping has become increasingly important. However, currently, most urban landscaping relies on sprinkler trucks or automatic irrigation systems for watering. Although these methods have improved work efficiency to a certain extent, their watering methods are often carried out according to a fixed cycle. Since the water requirements of plants cannot be accurately understood, all plants are irrigated with the same amount of water, resulting in insufficient water in some areas and excessive irrigation in some areas. This not only causes waste of water resources, but also may affect the growth of plants, and at the same time wastes manpower and material resources. Therefore, in the technical solution of this application, by obtaining the current plant growth environment data obtained from the cloud platform, the water requirement management data of different plants obtained from the cloud platform, and the plant growth images collected by the camera, and combining deep learning technology, personalized plant irrigation suggestions are generated to provide a tailored irrigation plan according to the growth environment and water requirement characteristics of each plant. This method effectively avoids the situation of over-watering or under-watering, saves water resources, and promotes the healthy growth of plants. At the same time, it can also improve the efficiency and accuracy of urban landscaping management.

[0024] In the above-mentioned cloud platform-based urban landscaping management data processing system 100, the urban landscaping management data extraction module 120 is used to extract plant growth multi-modal association feature vectors and plant growth attention fully connected feature vectors from the current plant growth environment data obtained from the cloud platform, the water requirement management data of different plants obtained from the cloud platform, and the plant growth images collected by the camera. It should be understood that plant growth is a complex process affected by various factors, including but not limited to environmental conditions such as temperature, humidity, and light, the differences in water requirements of different plant species, and the growth status of the plants themselves. By integrating and analyzing these heterogeneous data sources, the growth status of plants and their environmental adaptability can be comprehensively understood, so as to more accurately understand the specific needs of each plant and generate effective irrigation suggestions. This method not only improves the efficiency and accuracy of urban landscaping management, but also maximally saves water resources and promotes the healthy growth of plants.

[0025] Figure 2 It is a block diagram of the urban landscaping management data extraction module in the cloud platform-based urban landscaping management data processing system according to an embodiment of the present application. As Figure 2As shown, in a specific embodiment of the present application, the urban landscaping management data extraction module 120 includes: a plant current growth environment data feature extraction unit 121, which is used to extract features from the plant current growth environment data obtained from the cloud platform to obtain a multi-scale plant current growth environment understanding feature vector; a water demand management data feature extraction unit 122, which is used to extract features from the water demand management data of different plants obtained from the cloud platform to obtain a plant water demand management semantic association feature vector; a plant growth multi-modal feature association unit 123, which is used to associate the multi-scale plant current growth environment understanding feature vector and the plant water demand management semantic association feature vector to obtain the plant growth multi-modal association feature vector; a plant growth attention feature extraction unit 124, which is used to extract features from the plant growth images collected by the camera to obtain the plant growth attention fully connected feature vector.

[0026] In the technical solution of the present application, the plant current growth environment data feature extraction unit 121 is used to extract features from the plant current growth environment data obtained from the cloud platform to obtain a multi-scale plant current growth environment understanding feature vector. It should be understood that the healthy growth of plants depends on various environmental factors such as temperature, humidity, and light. The changes of these factors have complex and multi-level impacts on plants. By extracting the features of the plant current growth environment data, different levels of information in the plant current growth environment data can be captured, better understanding of the growth environment requirements of plants at different scales, and more accurate and effective management strategies can be formulated accordingly.

[0027] In the technical solution of the present application, the water demand management data feature extraction unit 122 is used to extract features from the water demand management data of different plants obtained from the cloud platform to obtain a plant water demand management semantic association feature vector. It should be understood that the healthy growth of plants not only depends on factors such as temperature, humidity, and light in its growth environment, but also highly depends on the rationality of water supply. However, different types of plants have different water demand characteristics, and even the same species may have different water requirements at different growth stages or in different environments. Therefore, by extracting the plant water demand management semantic association feature vector through feature extraction, the water demands of different plants in specific growth environments can be more accurately understood and simulated.

[0028] In the technical solution of the present application, the plant growth multi-modal feature association unit 123 is used to associate the multi-scale plant current growth environment understanding feature vector and the plant water demand management semantic association feature vector to obtain the plant growth multi-modal association feature vector. It should be understood that plants have different water requirements at different growth stages and in different environments, and these requirements are affected by various environmental factors. By combining multi-scale data reflecting environmental conditions with water demand data of specific plant species, the actual growth requirements of plants can be captured more accurately. Specifically, the multi-scale plant current growth environment understanding feature vector contains multi-dimensional information such as temperature, humidity, and light, and takes into account the dynamic changes in time and the distribution differences in space, providing rich background knowledge; while the plant water demand management semantic association feature vector details the water quantity information required by different plant species and at each growth stage, revealing the specific water demand patterns of plants. The combination of the two can provide customized irrigation suggestions for each plant, avoiding the problems of over-irrigation or under-irrigation.

[0029] In the technical solution of the present application, the plant growth attention feature extraction unit 124 is used to extract features from the plant growth image collected by the camera to obtain the plant growth attention fully connected feature vector. It should be understood that the health of plants is not only affected by environmental factors but also closely related to their appearance features, such as the color, shape, and texture of leaves. Through image analysis, these features can be visually observed, and meaningful information can be extracted from them using deep learning techniques to accurately capture and understand the visual growth state of plants and their changing trends, and highlight the local details that are crucial for plant health, helping the system to more accurately determine whether the plant needs irrigation or other management measures.

[0030] In a specific embodiment of the present application, the plant current growth environment data feature extraction unit 121 includes: passing the plant current growth environment data obtained from the cloud platform through a plant current growth environment data context encoder including an embedding layer to obtain multiple plant current growth environment text feature vectors; arranging the multiple plant current growth environment text feature vectors into a plant current growth environment text input vector; passing the plant current growth environment text input vector through a plant current growth environment multi-scale neighborhood feature extraction module to obtain the multi-scale plant current growth environment understanding feature vector.

[0031] In the technical solution of this application, the plant current growth environment data obtained from the cloud platform is passed through a plant current growth environment data context encoder including an embedding layer to obtain multiple plant current growth environment text feature vectors. It should be understood that plant growth is affected by various environmental factors, such as temperature, humidity, light, etc., and this data usually exists in the form of time series or structured data. However, directly using these raw data for analysis often makes it difficult to capture their inherent complex patterns and interrelationships. Through the embedding layer and the context encoder, these data can be transformed into feature vectors with rich semantic information, so as to better capture and represent complex environmental information. Specifically, in the technical solution of this application, the main function of the plant current growth environment data context encoder including the embedding layer is to map the raw data to a high-dimensional embedding space, in which each data point not only represents its own numerical value, but also contains the context relationship with other data points. The role of the embedding layer is to transform these raw numerical values into higher-dimensional feature representations, so that they can better express these complex associations in the embedding space. The context encoder captures the long-term dependencies and context information in the sequence data. Finally, the multiple plant current growth environment text feature vectors generated after being processed by the context encoder not only contain the basic information of the original environmental data, but also incorporate the context relationships and patterns between the data. In the technical solution of this application, passing the plant current growth environment data obtained from the cloud platform through a plant current growth environment data context encoder including an embedding layer to obtain multiple plant current growth environment text feature vectors includes: performing word segmentation processing on the plant current growth environment data obtained from the cloud platform to obtain a plant current growth environment word sequence; using the embedding layer of the plant current growth environment data context encoder including the embedding layer to map each plant current growth environment word in the plant current growth environment word sequence into a word embedding vector respectively to obtain a sequence of plant current growth environment word embedding vectors; using the Transformer-based Bert model of the plant water demand management semantic encoder to perform global context semantic encoding on the sequence of plant current growth environment word embedding vectors to obtain multiple plant current growth environment text feature vectors

[0032] In the technical solution of this application, the multiple plant current growth environment text feature vectors are arranged into a plant current growth environment text input vector. It should be understood that the growth of plants is affected by various environmental factors, such as temperature, humidity, light, etc., and the data of these factors usually exist in the form of time series or structured form. Each feature vector represents the environmental state information of a certain specific environmental parameter or time period. However, directly using these scattered feature vectors for the training or inference of a deep learning model is often not effective enough because they lack integrity and context connection. By arranging these feature vectors into a unified input vector, the internal relationships and patterns between environmental data can be better captured, thereby improving the prediction accuracy and robustness of the model. When specifically implementing this process, it can be completed through a simple vector concatenation operation, that is, connecting each feature vector in sequence into a larger vector. This concatenation method not only retains the information inside each feature vector but also can reflect the interaction and dependence relationship between them to a certain extent.

[0033] In the technical solution of this application, the plant's current growth environment text input vector is passed through the plant's current growth environment multi-scale neighborhood feature extraction module to obtain the multi-scale plant current growth environment understanding feature vector. It should be understood that the healthy growth of plants not only depends on environmental parameters (such as temperature, humidity, light, etc.) at a single time point or local area, but also needs to consider the change trends and interactions of these parameters at different time and space scales. By processing the plant's current growth environment text input vector through the multi-scale neighborhood feature extraction module, a richer and more detailed feature representation can be generated, helping the system to more accurately simulate and predict the growth requirements of plants. Specifically, considering that the plant's current growth environment text input vector already contains information on multiple time points or environmental parameters, but this information is usually discrete and independent. To better understand and utilize this data, a multi-scale analysis method must be introduced. Among them, the input vector is fed into a multi-scale convolutional layer, which can capture the change patterns of environmental parameters from different scales and angles. For example, a convolutional kernel may focus on short-term time series changes (such as temperature fluctuations within a day), while another convolutional kernel may focus on long-term trends (such as the average temperature change within a month). In this way, the convolutional layer can extract meaningful features at different scales and combine them into a preliminary feature map. In the technical solution of this application, passing the plant's current growth environment text input vector through the plant's current growth environment multi-scale neighborhood feature extraction module to obtain the multi-scale plant current growth environment understanding feature vector includes: a first convolutional layer, a second convolutional layer parallel to the first convolutional layer, and a concatenation layer connected to the first convolutional layer and the second convolutional layer, where the first convolutional layer uses a one-dimensional convolutional kernel with a first scale, and the second convolutional layer uses a one-dimensional convolutional kernel with a second scale. More specifically, the plant's current growth environment multi-scale neighborhood feature extraction module includes: performing one-dimensional convolutional encoding on the plant's current growth environment text input vector using the first convolutional layer of the plant's current growth environment multi-scale neighborhood feature extraction module to obtain a first-scale feature vector; performing one-dimensional convolutional encoding on the plant's current growth environment text input vector using the second convolutional layer of the plant's current growth environment multi-scale neighborhood feature extraction module to obtain a second-scale feature vector; concatenating the first-scale feature vector and the second-scale feature vector to obtain the multi-scale plant current growth environment understanding feature vector.

[0034] In a specific embodiment of this application, the water demand management data feature extraction unit 122 includes: passing the different plant water demand management data obtained from the cloud platform through a plant water demand management semantic encoder to obtain multiple plant water demand management semantic feature vectors; concatenating the multiple plant water demand management semantic feature vectors into the plant water demand management semantic association feature vector.

[0035] In the technical solution of this application, the different plant water requirement management data obtained from the cloud platform are processed through a plant water requirement management semantic encoder to obtain multiple plant water requirement management semantic feature vectors. It should be understood that different plant species have significant differences in water requirements at different stages of their life cycles. By extracting these complex water requirement information through a semantic encoder and converting it into feature vectors with rich semantic information, it can provide a scientific basis for formulating precise irrigation strategies. In the technical solution of this application, the different plant water requirement management data obtained from the cloud platform are processed through a plant water requirement management semantic encoder to obtain multiple plant water requirement management semantic feature vectors, including: performing word segmentation on the different plant water requirement management data obtained from the cloud platform to obtain a water requirement word sequence; using the embedding layer of the plant water requirement management semantic encoder to map each water requirement word in the water requirement word sequence into a word embedding vector respectively to obtain a sequence of water requirement word embedding vectors; using the Transformer-based Bert model of the plant water requirement management semantic encoder to perform global context semantic encoding on the sequence of word embedding vectors to obtain multiple plant water requirement management semantic feature vectors.

[0036] In the technical solution of this application, the multiple plant water requirement management semantic feature vectors are cascaded into the plant water requirement management semantic association feature vector. It should be understood that the multiple plant water requirement management semantic feature vectors already contain rich semantic information, and each vector represents the water requirement characteristics of a plant at a certain specific growth stage. However, using these feature vectors alone may not be able to fully reflect the mutual relationship between different plants and the overall water requirement pattern. Cascading these feature vectors can integrate the scattered information together to form a unified and structured input representation. Specifically, concatenation is the most direct way, which connects each feature vector in sequence into a larger vector. This concatenation method not only retains the information inside each feature vector but also can reflect the interaction and dependency relationship between them to a certain extent.

[0037] In a specific embodiment of the present application, the plant growth multimodal feature association unit includes: creating a Spring Boot project and initializing the project through the Spring Initializr tool, selecting the Spring Web dependency to build a Web application environment; defining data model classes for representing the current growth environment understanding feature vectors of multi-scale plants, the semantic association feature vectors of plant water demand management, and the generated plant growth multimodal association feature vectors; defining the FeatureAssociationService service layer for receiving input parameters and executing the association logic, generating new feature vectors and encapsulating them into a PlantMultimodalFeatureVector object for return; creating a FeatureController controller class for defining REST API interfaces and defining POST request handling methods, receiving a request body in JSON format, parsing it into a FeatureRequest object, calling the service layer method and returning the result; configuring the application startup parameters and setting the server port; packaging the Spring Boot application into a JAR file and deploying it for running.

[0038] Among them, the deployment code for part of the data model definition is as follows.

[0039]

[0040]

[0041] Among them. The deployment code for part of the service layer implementation is as follows.

[0042]

[0043]

[0044] Among them, the deployment code for part of the controller definition is as follows.

[0045]

[0046]

[0047] Among them, the deployment code for part of the request body class is as follows.

[0048]

[0049]

[0050] Among them, the deployment code for part of the configuration and startup is as follows.

[0051] server.port=8080

[0052] mvn spring-boot:run

[0053] It should be understood that in the technical solution of this application, the core logic is encapsulated in the FeatureAssociationService service class, and two input feature vectors are combined into a comprehensive feature vector through a simple splicing algorithm. This design not only simplifies the feature fusion process, but also ensures the efficiency and scalability of the system. Specifically, first of all, the design of the data model is the basis of the entire system. By defining three classes, namely PlantEnvironmentFeatureVector, PlantWaterDemandFeatureVector, and PlantMultimodalFeatureVector, corresponding to different types of feature vectors respectively, the feature data can be stored and transmitted in a structured manner. These model classes provide a clear data interface for the subsequent service layer logic, ensuring the readability and maintainability of the code. The FeatureAssociationService in the service layer implements the core feature association logic, and splices the two feature vectors through the Java Streams API to finally generate a new feature vector. Although this method is simple, it can effectively capture the basic relationship between the two features in practical applications, providing high-quality input data for subsequent deep learning or machine learning models. The implementation of the controller layer further enhances the interaction ability of the system. By defining the FeatureController class and mapping it to the / api / plant-features path, the system provides a RESTful API interface that allows external clients to call the feature association function through an HTTP POST request. The client only needs to send a JSON object containing two feature vectors to obtain the processed multimodal feature vector as a response. This design conforms to the standards of modern web development, facilitates integration with other systems, and also supports cross-platform calls. In this way, the complex feature fusion process can be abstracted into a set of simple API interfaces, greatly reducing the usage threshold. At the same time, since the system is built based on the Spring Boot framework, it has high portability and scalability, and more complex feature fusion algorithms (such as weighted summation or attention mechanism) can be introduced in the future to further improve performance. Generally speaking, this deployment solution not only achieves the expected functional goals, but also lays a solid foundation for subsequent technology upgrades and function expansions.

[0054] In a specific embodiment of the present application, the plant growth attention feature extraction unit 124 includes: passing the plant growth image collected by the camera through a plant growth attention feature extractor based on a spatial attention network to obtain a plant growth attention feature map; pooling the plant growth attention feature map into a plant growth attention pooled feature vector; passing the plant growth attention pooled feature vector through a plant growth attention feature encoder based on a fully connected layer to obtain the plant growth attention fully connected feature vector.

[0055] In the technical solution of the present application, the plant growth image collected by the camera is passed through a plant growth attention feature extractor based on a spatial attention network to obtain a plant growth attention feature map. It should be understood that the health status of a plant depends not only on environmental factors but is also closely related to its appearance features, such as the color, shape, texture of the leaves, etc. However, the plant growth image collected by the camera usually contains rich visual information, but not all information is helpful for judging the plant health status. Directly using the original image for analysis often makes it difficult to effectively identify these key features, especially when the image contains a large amount of background information or noise. Therefore, through a Spatial Attention Network, it is possible to automatically focus on the part of the image that best reflects the plant health status and generate a highly targeted feature map. In the technical solution of the present application, passing the plant growth image collected by the camera through a plant growth attention feature extractor based on a spatial attention network to obtain a plant growth attention feature map includes: using the convolutional encoding part of the plant growth attention feature extractor based on the spatial attention network to perform deep convolutional encoding on the plant growth image collected by the camera to obtain an initial convolutional feature map; inputting the initial convolutional feature map into the spatial attention part of the plant growth attention feature extractor based on the spatial attention network to obtain a spatial attention map; passing the spatial attention map through the Softmax activation function to obtain a spatial attention feature map; calculating the element-wise multiplication of the spatial attention feature map and the initial convolutional feature map to obtain the plant growth attention feature map. More specifically, inputting the initial convolutional feature map into the spatial attention part of the plant growth attention feature extractor based on the spatial attention network to obtain a spatial attention map includes: performing average pooling and max pooling on the initial convolutional feature map along the channel dimension respectively to obtain an average feature matrix and a max feature matrix; concatenating and channel-adjusting the average feature matrix and the max feature matrix to obtain a channel feature matrix; using the convolutional layer of the spatial attention feature map to perform convolutional encoding on the channel feature matrix to obtain a spatial attention map.

[0056] In the technical solution of this application, the plant growth attention feature map is pooled into a plant growth attention pooled feature vector. It should be understood that when the plant growth image obtained from the camera undergoes feature extraction based on the spatial attention network, a plant growth attention feature map highlighting the key regions of the plant health state is obtained. However, directly using such a high-resolution feature map for analysis may lead to a huge consumption of computing resources and is prone to introducing unnecessary noise. Therefore, it is necessary to convert it into a more compact and easy-to-process form. Commonly used pooling methods include Max Pooling, Average Pooling, etc. Max Pooling represents the information of each local region by selecting the maximum value in the region, which helps to retain the most significant features; while Average Pooling calculates the average value of each local region, which helps to smooth the features and reduce the influence of noise. Among them, the pooling window determines the size of the image region processed each time, and the stride determines the distance between adjacent processed regions. Generally, a larger pooling window and stride can significantly reduce the size of the output feature map, but may also cause some detailed information to be lost; on the contrary, a smaller pooling window and stride can retain more details, but will increase the computational burden. After the pooling operation is completed, the result is a feature map with a lower resolution but more concentrated information. To convert it into a vector form for subsequent processing by deep learning models, a global pooling operation is also required. Global pooling refers to applying the pooling operation over the entire feature map rather than being limited to a certain local region. For example, global average pooling can be adopted, that is, taking the average of all pixel values of the entire feature map to generate a single value. This not only greatly reduces the amount of data but also effectively summarizes the information of the entire feature map to form a feature vector with a fixed length.

[0057] In the technical solution of this application, the plant growth attention pooling feature vector is passed through a plant growth attention feature encoder based on a fully connected layer to obtain the plant growth attention fully connected feature vector. It should be understood that although the pooled feature vectors have greatly simplified the complexity of the original data, they may not be directly applicable to specific task requirements. A fully connected layer (FC) can map the input data to a new feature space, achieving a non-linear transformation in this process and providing a richer representation for subsequent tasks. Specifically, first, the input pooled feature vectors are flattened into a one-dimensional vector form to ensure that they can be directly used as the input of the fully connected layer. Then, these vectors pass through a fully connected layer composed of multiple neurons. Each neuron has its own weight and bias term, which are used to calculate the weighted sum of the input vectors, and a non-linear element is introduced through an activation function (such as ReLU or Sigmoid) to enhance the model's expression ability. During the whole process, the model adjusts its internal parameters according to a specific task (such as predicting the water requirement of plants) to minimize the prediction error. This usually involves the application of optimization techniques such as the backpropagation algorithm and the gradient descent method to continuously adjust the weight values in the fully connected layer until the optimal solution is found. Finally, the plant growth attention fully connected feature vector obtained after a series of fully connected layer processes not only contains the important information in the original image but also incorporates the influence of other relevant factors (such as plant type, growth environment, etc.).

[0058] Figure 3 Block diagram of the plant irrigation recommendation generation module in the urban landscaping management data processing system based on the cloud platform according to an embodiment of the present application. As Figure 3 shown, in the above-mentioned urban landscaping management data processing system 100 based on the cloud platform, the plant irrigation recommendation generation module 130 is used to generate plant irrigation recommendations based on the plant growth multi-modal association feature vector and the plant growth attention fully connected feature vector, including: an urban landscaping feature fusion unit 131, which is used to fuse the plant growth multi-modal association feature vector and the plant growth attention fully connected feature vector to obtain a plant growth irrigation management feature vector; an urban landscaping feature optimization unit 132, which is used to perform feature core factor embedding optimization based on sparse fusion on the plant growth irrigation management feature vector to obtain an optimized plant growth irrigation management feature vector; and an urban landscaping recommendation reporting unit 133, which is used to pass the optimized plant growth irrigation management feature vector through a generator to generate plant irrigation recommendations.

[0059] In the technical solution of this application, the urban landscaping feature fusion unit 131 is used to combine the plant growth multi-modal correlation feature vector and the plant growth attention fully-connected feature vector to obtain a plant growth irrigation management feature vector. It should be understood that the plant growth multi-modal correlation feature vector contains rich information extracted from environmental data (such as temperature, humidity, light, etc.) and water demand data, reflecting environmental changes at different time and space scales and their impacts on plants. The plant growth attention fully-connected feature vector mainly comes from image analysis, highlighting key areas in the plant appearance (such as the color and shape of leaves), and revealing the current health status of the plant. To fuse these two types of feature vectors into a unified representation, feature fusion techniques are usually adopted. In the technical solution of this application, feature fusion is performed through concatenation and weighted sum.

[0060] In the technical solution of this application, the urban landscaping feature optimization unit 132 is used to perform feature core factor embedding optimization based on sparse fusion on the plant growth irrigation management feature vector to obtain an optimized plant growth irrigation management feature vector. It should be understood that in the technical solution of this application, since both the current plant growth environment data and the plant water demand management data are related to the growth status of plants, if the feature information in both is similar or highly correlated, it may lead to redundant features, thereby affecting the efficiency and effectiveness of the model. Moreover, different data sources (such as growth environment data, plant water demand data, plant growth images) have different feature dimensions. After passing through their respective encoders and feature extraction modules, these data may be fused in different dimensions or scales. If the feature dimension differences are too large, directly fusing multi-modal features may lead to dimensional imbalance, resulting in insufficient attention of the model to certain features and affecting the final prediction accuracy. In this way, when the data volume is insufficient or the data distribution is biased, the model may over-rely on the noise or local patterns in the training data, resulting in limited generalization ability. In this case, the model's prediction ability for unseen data is poor, and errors are likely to occur in actual applications. Therefore, in the technical solution of this application, feature core factor embedding optimization based on sparse fusion is performed on the plant growth irrigation management feature vector to obtain an optimized plant growth irrigation management feature vector.

[0061] Among them, performing feature core factor embedding optimization based on sparse fusion on the plant growth irrigation management feature vector to obtain an optimized plant growth irrigation management feature vector includes:

[0062] First, extract the prior parameter base matrix for plant growth irrigation. It should be understood that extracting the prior parameter base matrix for plant growth irrigation is not merely an information acquisition process, but rather a crucial operation for structuring and computabilizing prior knowledge. The principle lies in condensing domain expertise, model design concepts, or macroscopic laws inverted from data into a mathematical form of the prior parameter base matrix for plant growth irrigation, so as to facilitate the effective utilization of subsequent algorithms.

[0063] Then, perform core prior information feature selection on the prior parameter base matrix for plant growth irrigation to obtain a set of latent factor decomposition coding vectors of the core prior information of the plant growth irrigation model, which is expressed by the core prior information feature selection formula as:

[0064]

[0065] where U represents the set of latent factor decomposition coding vectors of the core prior information of the plant growth irrigation model, v1, v2, v m represent the first, second, and m-th latent factor decomposition coding vectors of the core prior information of the plant growth irrigation model respectively, T represents the transpose operation, CoreExtraction represents core prior information feature selection, M p represents the prior parameter base matrix for plant growth irrigation, Λ represents a diagonal matrix, and λ1, λ m represent the values at the first and m-th positions on the diagonal of the diagonal matrix respectively. It should be understood that the prior parameter base matrix for plant growth irrigation may contain redundant or high-dimensional information, and direct application may lead to an excessive computational burden and interference from non-critical information in the optimization process. Therefore, the principle of core prior information feature selection is to denoise and refine prior knowledge, similar to the filtering process in signal processing, retaining the main components and filtering out redundancy.

[0066] Next, construct a core prior information response potential mapping matrix of the plant growth irrigation model between the plant growth irrigation management feature vector and each latent factor decomposition coding vector in the set of latent factor decomposition coding vectors of the core prior information of the plant growth irrigation model to obtain a set of core prior information response potential mapping matrices of the plant growth irrigation model, which is expressed by the potential mapping formula as:

[0067]

[0068] where x o represents the plant growth irrigation management feature vector, l i (x o ) represents x oPerform a linear transformation. The eigenvector after the linear transformation has the same eigen-scale as the corresponding latent factor decomposition coding vector of the core prior information of the plant growth irrigation model, v i represents the i-th latent factor decomposition coding vector of the core prior information of the plant growth irrigation model, represents matrix multiplication, L represents the length of the latent factor decomposition coding vector of the core prior information of the plant growth irrigation model, MR i represents the i-th latent mapping matrix of the core prior information response of the plant growth irrigation model. It should be understood that the core of this step lies in constructing the interaction relationship between the plant growth irrigation management eigenvector and the refined prior knowledge. Specifically, through latent space mapping, the non-linear response pattern of the plant growth irrigation management eigenvector to prior knowledge in different aspects is learned. Essentially, the latent mapping matrix of the core prior information response of the plant growth irrigation model is a re-coding of the plant growth irrigation management eigenvector from the perspective of prior knowledge, incorporating the interpretation and processing of knowledge. Its function exceeds information association, achieving directional enhancement of feature representation and extraction of multi-perspective feature information.

[0069] Immediately afterwards, calculate the kernel correlation metric value of the prior information response of the plant growth irrigation for each latent mapping matrix of the core prior information response of the plant growth irrigation model in the set of latent mapping matrices of the core prior information response of the plant growth irrigation model to obtain a set of kernel correlation metric values of the prior information response of the plant growth irrigation, which is expressed by the kernel correlation metric formula as:

[0070] S i = ||MR i || F

[0071] where ||·|| F represents the F-norm of the matrix, and S i represents the i-th kernel correlation metric value of the prior information response of the plant growth irrigation. It should be understood that the principle of this step is to refine key information and streamline feature representation for each latent mapping matrix of the core prior information response of the plant growth irrigation model. Just like generating an information summary, the most representative summary or eigenvector of the core information is extracted. The kernel correlation metric value of the prior information response of the plant growth irrigation needs to be representative and discriminative. It contains the ideas of feature selection and feature aggregation, retaining the most informative parts and aggregating them into a concise kernel correlation metric value to achieve deeper compression and refinement. The function of the kernel correlation metric value of the prior information response of the plant growth irrigation is not only information compression, but more importantly, to improve the efficiency and robustness of the subsequent fusion process, reduce the data dimension, and reduce the computational burden, especially in the case of high-dimensional matrices.

[0072] Subsequently, based on the set of kernel correlation metric values of the prior information responses of plant growth irrigation, a sparse dynamic fusion is performed on the set of potential mapping matrices of the prior information responses of the core of the plant growth irrigation model to obtain the prior information response projection coding matrix of the plant growth irrigation model, which is expressed by the sparse dynamic fusion formula as:

[0073]

[0074] Among them, softmax represents the normalized exponential function, and P represents the prior information response projection coding matrix of the plant growth irrigation model. It should be understood that the core principle of the sparse dynamic fusion lies in emphasizing the adaptive and selective prior information integration strategy. Sparsity reflects that not all prior information responses are equally important, and the fusion should be selective, focusing on more important responses, weakening or ignoring unimportant responses, improving the feature selection ability and generalization ability, and avoiding overfitting. Dynamics means that the weights or methods of sparse dynamic fusion are not fixed, but are adaptively adjusted according to the input data or model state, improving the flexibility and adaptability of the model. The essence of performing sparse dynamic fusion on the set of potential mapping matrices of the prior information responses of the core of the plant growth irrigation model is to optimally combine the response information from different prior knowledge perspectives to form the prior information response projection coding matrix of the plant growth irrigation model that comprehensively reflects the prior response of the model.

[0075] Finally, the plant growth irrigation management feature vector is mapped to the feature space of the prior information response projection coding matrix of the plant growth irrigation model to obtain the optimized plant growth irrigation management feature vector, which is expressed by the mapping formula as:

[0076]

[0077] where x optRepresents the optimized plant growth irrigation management feature vector. It should be understood that the entire adaptation process is finally completed by mapping the plant growth irrigation management feature vector to the feature space of the prior information response projection coding matrix of the plant growth irrigation model to obtain the optimized plant growth irrigation management feature vector. The principle of this step is to utilize the feature space defined by the prior information response projection coding matrix of the plant growth irrigation model and project the plant growth irrigation management feature vector into this space. Substantially, the prior information response projection coding matrix of the plant growth irrigation model acts as a transformation matrix to perform a linear or non-linear mapping transformation on the plant growth irrigation management feature vector, enabling it to be embedded in the feature space integrating the model prior information. The optimized plant growth irrigation management feature vector better conforms to the constraints and guidance of the model prior, achieving boundary adaptation on the feature manifold. Compared with the plant growth irrigation management feature vector, the optimized plant growth irrigation management feature vector is usually significantly improved in terms of expression ability, discriminability, and generalization, providing a better feature representation for subsequent machine learning tasks.

[0078] In the technical solution of this application, the urban landscaping recommendation report unit 133 is used to generate plant irrigation recommendations by passing the optimized plant growth irrigation management feature vector through a generator. It should be understood that the generator is usually a deep neural network model, such as a generative adversarial network (GAN), a variational autoencoder (VAE), or a multi-layer perceptron (MLP). The core function of the generator is to learn and extract information related to irrigation strategies from the input feature vector and convert it into specific recommendations that are easy to understand and execute. For example, the generator can determine whether irrigation is needed and the specific amount and time of irrigation based on the current environmental conditions and the health status of the plants. To achieve this, the generator first performs a series of non-linear transformations on the input optimized feature vector. Each neuron has its own weight and bias term to calculate the weighted sum of the input features and introduce non-linearity through an activation function (such as ReLU or Sigmoid) to enhance the expression ability of the model. This non-linear transformation enables the generator to capture the complex relationships between the input features and map these relationships to the output space. During the training process, the generator continuously adjusts its internal parameters to minimize the prediction error and improve the accuracy of the generated recommendations. In addition, to prevent overfitting and improve the generalization ability of the model, regularization techniques such as Dropout or L2 regularization are often applied in the generator. In this way, the generator can also exhibit good stability and accuracy when facing new data. Finally, the generator converts the optimized plant growth irrigation management feature vector into specific irrigation recommendations. These recommendations may include detailed information such as the time, frequency, and amount of irrigation.

[0079] In summary, in the embodiments of the present application, the current plant growth environment data obtained from the cloud platform, the water demand management data of different plants obtained from the cloud platform, and the plant growth images collected by the camera are first acquired. Using deep learning technology, feature extraction and correlation analysis are performed on the three, and through a generator, a plant irrigation suggestion is generated, so as to obtain a customized irrigation plan according to the characteristics of the growth environment and water demand of each plant, effectively avoiding over-irrigation or under-irrigation, saving water resources, improving the healthy growth of plants, and improving the efficiency and accuracy of urban landscaping management.

[0080] Figure 4 FIG. is a flowchart of a method for processing urban landscaping management data based on a cloud platform according to an embodiment of the present application. As Figure 4 shown, the method for processing urban landscaping management data based on a cloud platform according to an embodiment of the present application includes: S110, acquiring the current plant growth environment data obtained from the cloud platform, the water demand management data of different plants obtained from the cloud platform, and the plant growth images collected by the camera; S120, extracting a plant growth multi-modal correlation feature vector and a plant growth attention fully connected feature vector from the current plant growth environment data obtained from the cloud platform, the water demand management data of different plants obtained from the cloud platform, and the plant growth images collected by the camera; S130, generating a plant irrigation suggestion based on the plant growth multi-modal correlation feature vector and the plant growth attention fully connected feature vector.

[0081] Here, those skilled in the art can understand that the specific operations of each step in the above method for processing urban landscaping management data based on a cloud platform have been described in detail in the description of the cloud platform-based urban landscaping management data processing system above, and therefore, the repeated description thereof will be omitted. Figures 1 to 3 of the cloud platform-based urban landscaping management data processing system, and thus, the repeated description thereof will be omitted.

Claims

1. An urban landscaping management data processing system based on a cloud platform, characterized in that, Including: An urban landscaping management data acquisition module, configured to acquire the current plant growth environment data obtained from a cloud platform, the water requirement management data of different plants obtained from the cloud platform, and the plant growth images collected by a camera; An urban landscaping management data extraction module, configured to extract a plant growth multi-modal association feature vector and a plant growth attention fully-connected feature vector from the current plant growth environment data obtained from the cloud platform, the water requirement management data of different plants obtained from the cloud platform, and the plant growth images collected by the camera; A plant irrigation recommendation generation module, configured to generate a plant irrigation recommendation based on the plant growth multi-modal association feature vector and the plant growth attention fully-connected feature vector.

2. The urban landscaping management data processing system based on a cloud platform according to claim 1, wherein The urban landscaping management data extraction module includes: A current plant growth environment data feature extraction unit, configured to perform feature extraction on the current plant growth environment data obtained from the cloud platform to obtain a multi-scale current plant growth environment understanding feature vector; A water requirement management data feature extraction unit, configured to perform feature extraction on the water requirement management data of different plants obtained from the cloud platform to obtain a plant water requirement management semantic association feature vector; A plant growth multi-modal feature association unit, configured to associate the multi-scale current plant growth environment understanding feature vector and the plant water requirement management semantic association feature vector to obtain the plant growth multi-modal association feature vector; A plant growth attention feature extraction unit, configured to perform feature extraction on the plant growth images collected by the camera to obtain the plant growth attention fully-connected feature vector.

3. The urban landscaping management data processing system based on a cloud platform according to claim 2, wherein The current plant growth environment data feature extraction unit includes: Passing the current plant growth environment data obtained from the cloud platform through a current plant growth environment data context encoder including an embedding layer to obtain a plurality of current plant growth environment text feature vectors; Arranging the plurality of current plant growth environment text feature vectors into a current plant growth environment text input vector; Passing the current plant growth environment text input vector through a current plant growth environment multi-scale neighborhood feature extraction module to obtain the multi-scale current plant growth environment understanding feature vector.

4. The urban landscaping management data processing system based on a cloud platform according to claim 3, wherein, The water requirement management data feature extraction unit includes: Passing the water requirement management data of different plants obtained from the cloud platform through a plant water requirement management semantic encoder to obtain a plurality of plant water requirement management semantic feature vectors; Cascading the plurality of plant water requirement management semantic feature vectors into the plant water requirement management semantic association feature vector.

5. The urban landscaping management data processing system based on a cloud platform according to claim 4, wherein The plant growth multi-modal feature association unit includes: Creating a Spring Boot project and initializing the project through the Spring Initializr tool, and selecting Spring Web dependencies to build a Web application environment; Defining a data model class for representing the multi-scale current plant growth environment understanding feature vector, the plant water requirement management semantic association feature vector, and the generated plant growth multi-modal association feature vector. Define the FeatureAssociationService service layer, which is used to receive input parameters, execute association logic, generate new feature vectors, encapsulate them into a PlantMultimodalFeatureVector object and return it; Create a FeatureController controller class, which is used to define REST API interfaces, define a POST request handling method, receive a request body in JSON format, parse it into a FeatureRequest object, call the service layer method and return the result; Configure the application startup parameters and set the server port; Package the Spring Boot application into a JAR file and deploy it for running.

6. The urban landscaping management data processing system based on a cloud platform according to claim 5, wherein, The plant growth attention feature extraction unit includes: Pass the plant growth image collected by the camera through a plant growth attention feature extractor based on a spatial attention network to obtain a plant growth attention feature map; Pool the plant growth attention feature map into a plant growth attention pooled feature vector; Pass the plant growth attention pooled feature vector through a plant growth attention feature encoder based on a fully connected layer to obtain the plant growth attention fully connected feature vector.

7. The urban landscaping management data processing system based on a cloud platform according to claim 6, wherein The plant irrigation recommendation generation module includes: An urban landscaping feature fusion unit, which is used to fuse the plant growth multimodal association feature vector and the plant growth attention fully connected feature vector to obtain a plant growth irrigation management feature vector; An urban landscaping feature optimization unit, which is used to perform feature core factor embedding optimization based on sparse fusion on the plant growth irrigation management feature vector to obtain an optimized plant growth irrigation management feature vector; An urban landscaping recommendation report unit, which is used to pass the optimized plant growth irrigation management feature vector through a generator to generate plant irrigation recommendations.

8. The urban landscaping management data processing system based on a cloud platform according to claim 7, characterized in that The urban landscaping feature optimization unit includes: Extract the plant growth irrigation prior parameter base matrix; Perform core prior information feature selection on the plant growth irrigation prior parameter base matrix to obtain a set of plant growth irrigation model core prior information latent factor decomposition coding vectors; Construct a plant growth irrigation model core prior information response potential mapping matrix between the plant growth irrigation management feature vector and each plant growth irrigation model core prior information latent factor decomposition coding vector in the set of plant growth irrigation model core prior information latent factor decomposition coding vectors to obtain a set of plant growth irrigation model core prior information response potential mapping matrices; Calculate the plant growth irrigation prior information response kernel correlation metric value of each plant growth irrigation model core prior information response potential mapping matrix in the set of plant growth irrigation model core prior information response potential mapping matrices to obtain a set of plant growth irrigation prior information response kernel correlation metric values; Based on the set of prior information response kernel correlation metric values for plant growth irrigation, sparsely and dynamically fuse the set of potential mapping matrices of the core prior information response of the plant growth irrigation model to obtain the prior information response projection coding matrix of the plant growth irrigation model; Map the plant growth irrigation management feature vector to the feature space of the prior information response projection coding matrix of the plant growth irrigation model to obtain the optimized plant growth irrigation management feature vector.

9. A method for processing urban landscaping management data based on a cloud platform, characterized in that, Comprising: Obtain the current plant growth environment data obtained from the cloud platform, the different plant water requirement management data obtained from the cloud platform, and the plant growth images collected by the camera; Extract the plant growth multi-modal association feature vector and the plant growth attention fully connected feature vector from the current plant growth environment data obtained from the cloud platform, the different plant water requirement management data obtained from the cloud platform, and the plant growth images collected by the camera; Generate plant irrigation suggestions based on the plant growth multi-modal association feature vector and the plant growth attention fully connected feature vector.

10. The method for processing urban landscaping management data based on a cloud platform according to claim 9, characterized in that, Extracting the plant growth multi-modal association feature vector and the plant growth attention fully connected feature vector from the current plant growth environment data obtained from the cloud platform, the different plant water requirement management data obtained from the cloud platform, and the plant growth images collected by the camera, includes: Perform feature extraction on the current plant growth environment data obtained from the cloud platform to obtain a multi-scale current plant growth environment understanding feature vector; Perform feature extraction on the different plant water requirement management data obtained from the cloud platform to obtain a plant water requirement management semantic association feature vector; Associate the multi-scale current plant growth environment understanding feature vector and the plant water requirement management semantic association feature vector to obtain the plant growth multi-modal association feature vector; Perform feature extraction on the plant growth images collected by the camera to obtain the plant growth attention fully connected feature vector.

Images