Intelligent irrigation control method and system for vegetation in front of seawall

Abstract

The invention provides an intelligent irrigation control method and system for vegetation in front of a seawall, and relates to the technical field of vegetation watering. Obtaining a vegetation image; according to the vegetation image, through a convolutional neural network, vegetation growth stage identification is carried out; dividing vegetation areas according to vegetation types and the identified vegetation growth stages; according to the soil granularity, the average temperature, the precipitation amount, the air humidity, the normalized water body index, the vegetation type and the vegetation growth stage, predicting the water supply amount per unit area required by each vegetation region through an artificial neural network; and according to the water supply amount per unit area required by each vegetation area, irrigating vegetation in each vegetation area through a seawall front irrigation system. The method can provide the best growth condition for the vegetation in front of the seawall, promote the health and growth of the vegetation, help to maintain and improve the stability of the ecological system of the area in front of the seawall, and support biodiversity.

Description

Technical Field

[0001] The present invention relates to the technical field of vegetation watering, and in particular to an intelligent irrigation control method and system for vegetation in front of a seawall. Background Art

[0002] Planting measures in front of the dike are an important way to transform the seawall into an ecological one. They are an effective measure to solve the current situation of "three-sided light" of my country's seawalls and increase the porosity of the protective structure, thereby improving the habitat of the coastline and enriching biodiversity. The stems and leaves of plants can also increase the roughness of the protective structure, improve the moisture-proof standard by reducing the height of wave climbing, and thus achieve the ecological disaster reduction function.

[0003] In order to provide suitable water conditions for the vegetation in front of the seawall, the vegetation needs to be irrigated. Currently, flooding and artificial watering are mainly used to irrigate the vegetation in front of the seawall.

[0004] However, different types of vegetation have different demands for water, and vegetation at different growth stages also has different demands for water. Flood irrigation cannot distinguish the vegetation types in different areas, which can easily lead to excessive or insufficient water supply, affecting the healthy growth of vegetation. Manual watering is time-consuming and labor-intensive, and it is difficult to accurately grasp the amount of water for vegetation of different types and growth stages, which affects the healthy growth of vegetation. Summary of the invention

[0005] In order to solve the technical problem that the current technology is difficult to accurately grasp the amount of watering for vegetation of different types and at different growth stages, which affects the healthy growth of vegetation, the present invention provides an intelligent irrigation control method and system for vegetation in front of a seawall.

[0006] The technical solution provided by the embodiment of the present invention is as follows:

[0007] First aspect:

[0008] An embodiment of the present invention provides a method for controlling intelligent irrigation of vegetation in front of a seawall, comprising:

[0009] S1: Construct an irrigation system in front of the seawall;

[0010] S2: obtain vegetation images;

[0011] S3: According to the vegetation image, identifying the vegetation growth stage through a convolutional neural network;

[0012] S4: Divide vegetation areas according to vegetation types and identified vegetation growth stages;

[0013] S5: Based on soil particle size, average temperature, precipitation, air humidity, normalized water index, vegetation type and vegetation growth stage, the required water supply per unit area of ​​each vegetation area is predicted through artificial neural network;

[0014] S6: Irrigating the vegetation in each vegetation area through the seawall irrigation system according to the water supply per unit area required by each vegetation area.

[0015] Second aspect:

[0016] An embodiment of the present invention provides an intelligent irrigation control system for vegetation in front of a seawall, comprising:

[0017] processor;

[0018] A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the intelligent irrigation control method for vegetation in front of a seawall as described in the first aspect is implemented.

[0019] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0020] In the present invention, the required water supply per unit area for each vegetation area is predicted through an artificial neural network according to soil particle size, average temperature, precipitation, air humidity, normalized water index, vegetation type and vegetation growth stage. Through the irrigation system in front of the seawall, the vegetation in each vegetation area is intelligently irrigated, which reduces the cost of manual operation and management, can provide the best growth conditions for the vegetation in front of the seawall, promote the health and growth of the vegetation, help maintain and improve the stability of the ecosystem in the area in front of the seawall, and support biodiversity. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 A schematic diagram of a flow chart of a method for controlling intelligent irrigation of vegetation in front of a seawall provided by an embodiment of the present invention;

[0023] Figure 2 A schematic diagram of the structure of a method for intelligent irrigation control of vegetation in front of a seawall provided by an embodiment of the present invention;

[0024] Figure 3 A schematic structural diagram of an intelligent irrigation control system for vegetation in front of a seawall provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0026] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0027] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same. "of", "corresponding, relevant" and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same.

[0028] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0029] Reference Manual Attached Figure 1 , showing a flow chart of a method for intelligent irrigation control of vegetation in front of a seawall provided by an embodiment of the present invention.

[0030] Reference Manual Attached Figure 2 , showing a structural schematic diagram of a method for intelligent irrigation control of vegetation in front of a seawall provided by an embodiment of the present invention.

[0031] The embodiment of the present invention provides a method for controlling the intelligent irrigation of vegetation in front of a seawall, which can be implemented by an intelligent irrigation control device for vegetation in front of a seawall, and the intelligent irrigation control device for vegetation in front of a seawall can be a terminal or a server. The processing flow of the method for controlling the intelligent irrigation of vegetation in front of a seawall can include the following steps:

[0032] S1: Construct an irrigation system in front of the seawall.

[0033] In a possible implementation, the irrigation system in front of the seawall specifically includes: a water diversion channel, a sedimentation tank, a water storage tank, a hose and a valve. The water diversion channel outside the seawall is used as a water source, and the seawater enters the water storage tank after being precipitated in the sedimentation tank. The water storage tank is connected to the vegetation planting area through a hose, and an irrigation hole is opened on the hose. A valve is set at the irrigation hole, and the vegetation is drip irrigated by opening the valve.

[0034] It should be noted that the water diversion channel outside the seawall is used as a water source to directly obtain seawater from the ocean for irrigation. This design utilizes the abundant seawater resources in coastal areas and reduces dependence on freshwater resources. Especially in areas with tight freshwater resources, this method can significantly alleviate the problem of water shortage.

[0035] Furthermore, the seawater is first precipitated in the sedimentation tank before entering the reservoir, which can effectively remove the silt, impurities and suspended matter in the seawater and prevent these impurities from entering the irrigation system, thereby preventing the blockage of irrigation pipes and irrigation holes and ensuring the long-term stable operation of the irrigation system.

[0036] Furthermore, the reservoir can store settled seawater and regulate the amount of water, ensuring that the irrigation system can provide a stable water source at different times. In particular, when the tide changes or the water diversion is unstable, the reservoir can act as a buffer to ensure a continuous water supply.

[0037] Furthermore, the design of the hose and irrigation holes enables the drip irrigation system to deliver water directly to the roots of the vegetation, reducing evaporation and runoff losses. Drip irrigation can better control the amount of water to meet the needs of different vegetation areas and is particularly suitable for refined management.

[0038] Furthermore, the valve can flexibly adjust the irrigation volume of each area according to actual needs, ensuring that different plants or vegetation at different growth stages can get the right amount of water. Through the automated control system, the valve can also be adjusted in real time to further improve the accuracy of irrigation.

[0039] S2: Acquire vegetation images.

[0040] Optionally, the vegetation image may be acquired through a satellite or through a wide-angle camera. The present invention does not limit the specific method of acquiring the vegetation image.

[0041] S3: Based on the vegetation image, the vegetation growth stage is identified through the convolutional neural network.

[0042] Among them, Convolutional Neural Network (CNN) is a powerful and widely used deep learning model, which is particularly good at processing image data. In the task of identifying the growth stage of vegetation, CNN can effectively extract image features and automatically identify the growth stage of vegetation, providing technical support for precision agriculture and ecological management.

[0043] In a possible implementation, the convolutional neural network includes: a backbone feature extraction network, a feature fusion network, and a vegetation growth stage detection head. S3 specifically includes sub-steps S301 to S303:

[0044] S301: extracting a multi-scale feature map in a vegetation image through a backbone feature extraction network.

[0045] It should be noted that through the backbone feature extraction network, multi-scale feature maps can be extracted from vegetation images. Feature maps of different scales can capture information at different levels in the image. Small-scale feature maps capture details, such as the edges and textures of leaves, while large-scale feature maps capture global features, such as the morphology and structure of vegetation. Multi-scale feature extraction can retain the multi-level information of the image, so that the model can make full use of this information in subsequent processing, thereby improving the understanding and processing capabilities of complex scenes.

[0046] Optionally, the backbone feature extraction network includes convolution units connected in series, namely, a first convolution unit, a second convolution unit, and a third convolution unit. S301 specifically includes:

[0047] S3011: Taking the vegetation image as the input of the first convolution unit, and outputting a small-scale feature map in the vegetation image through the first convolution unit.

[0048] Optionally, S3011 specifically includes: performing convolution processing on the fabric image through a first 3×3 convolution layer to obtain an initial feature map.

[0049] The initial feature map is copied to obtain a first initial feature map and a second initial feature map.

[0050] The first initial feature map is separated by channel to obtain the first branch feature map and the second branch feature map. The second branch feature map is convolved through the first grouped convolution layer, the depthwise separable convolution layer, and the second grouped convolution layer in sequence to obtain the second branch separation feature map. Grouped convolution and depthwise separable convolution can significantly reduce model parameters and computational complexity while retaining sufficient feature extraction capabilities. Especially in deep networks, this design can reduce the computational burden and improve the efficiency of the model.

[0051] The first branch feature map is connected with the second branch separation feature map to obtain a connection feature map. The connection feature map is shuffled to obtain a shuffled feature map. Through separation, connection, shuffling and other operations, the information between channels can be fully exchanged and integrated, so that the network can capture the complex relationship between features in the image and improve the expression ability of the feature map.

[0052] The second initial feature map is symmetrically flipped to obtain a flipped feature map. The symmetrical flipping operation enables the model to process images from different directions or angles. When the image is flipped or rotated in the actual scene, the model can still recognize the features in it. This can improve the performance of the model in various situations, especially when the input image may have multiple perspectives.

[0053] The shuffled feature map and the flipped feature map are element-wise multiplied and convolved through the second 3×3 convolutional layer to obtain the enhanced feature map. Symmetric flipping and element-wise multiplication of the feature map can enhance the model's robustness to image transformations. This allows the model to accurately identify and classify input images at different angles or deformations.

[0054] The enhanced feature map is added to the initial feature map, and then convolved through the third 3×3 convolution layer to obtain a small-scale feature map. By adding the processed feature map to the initial feature map, the original key information can be retained while incorporating the enhanced new features to avoid information loss and enhance the representativeness of the features.

[0055] S3012: Using the small-scale feature map as input to the second convolution unit, and outputting a medium-scale feature map in the vegetation image through the second convolution unit.

[0056] Optionally, S3012 specifically includes: separating the small-scale feature map according to channels to obtain a first small-scale branch feature map and a second small-scale branch feature map. Convolving the second small-scale branch feature map through the first grouped convolution layer, the depth-separable convolution layer and the second grouped convolution layer in sequence to obtain a second small-scale branch separation feature map. Connecting the first small-scale branch feature map and the second small-scale branch separation feature map to obtain a connection feature map. Channel shuffling the connection feature map to obtain a shuffled feature map. Adding the shuffled feature map and the small-scale feature map, and convolving through a 3×3 third convolution layer to obtain a medium-scale feature map.

[0057] S3013: taking the medium-scale feature map as the input of the third convolution unit, and outputting the large-scale feature map in the vegetation image through the third convolution unit.

[0058] Optionally, S3013 specifically includes: separating the mesoscale feature map according to channels to obtain a first mesoscale branch feature map and a second mesoscale branch feature map. Convolving the second mesoscale branch feature map through the first grouped convolution layer, the depth-separable convolution layer and the second grouped convolution layer in sequence to obtain a second mesoscale branch separation feature map. Connecting the second mesoscale branch feature map with the second mesoscale branch separation feature map to obtain a connection feature map. Channel shuffling the connection feature map to obtain a shuffled feature map. Adding the shuffled feature map and the mesoscale feature map, and convolving through a 3×3 third convolution layer to obtain a large-scale feature map.

[0059] In the present invention, each level of feature map represents different levels of information of the image. Small-scale feature maps may contain more details, medium-scale feature maps integrate more local contexts, and large-scale feature maps capture global features. Through step-by-step convolution operations, starting from small-scale feature maps, medium-scale and large-scale feature maps are gradually extracted. This design can capture detailed features (such as texture, edges) to more global features (such as shape, structure) from the image. This feature extraction method from local to global enables the network to gradually understand the overall semantics and details of the image.

[0060] S302: Fusing the multi-scale feature maps through a feature fusion network to obtain a fused feature map.

[0061] It should be noted that by fusing multi-scale feature maps through the feature fusion network, detailed information can be effectively combined with global information to form a more comprehensive and robust feature representation. This fusion helps the model to more accurately understand the overall growth state of vegetation, rather than just local features.

[0062] In a possible implementation, the feature fusion network is specifically a feature pyramid network, and S302 specifically includes: upsampling the large-scale feature map to match the size of the large-scale feature map with the medium-scale feature map. Weighted fusion of the upsampled large-scale feature map and the medium-scale feature map to obtain a preliminary fused feature map. Upsampling the preliminary fused feature map to match the size of the preliminary fused feature map with the small-scale feature map. Weighted fusion of the upsampled preliminary fused feature map and the small-scale feature map to obtain a fused feature map.

[0063] In the present invention, by fusing large-scale, medium-scale and small-scale feature maps, the global information of the image can be combined with the local detail information. The large-scale feature map captures the overall structure of the image, while the small-scale feature map retains more detail information. By fusing these features, the feature pyramid network enables the model to simultaneously utilize information at different scales, thereby better understanding the content of the image. At the same time, in the multi-scale fusion process, due to the hierarchical structure of the feature pyramid network, the gradient can be propagated more smoothly, reducing the risk of gradient disappearance or explosion, making the model training more stable and converging faster.

[0064] S303: Determine the vegetation growth stage according to the fused feature map through the vegetation growth stage detection head.

[0065] The vegetation growth stages include: seedling stage, growth stage, reproductive stage, fruit development stage and decline stage.

[0066] It should be noted that the vegetation growth stage detection head can more accurately identify and classify the growth stage of vegetation based on the fused feature map. Since the fused feature map has integrated information at different scales, the detection head can better distinguish the subtle differences between vegetation at different growth stages.

[0067] In a possible implementation, S303 specifically includes: calculating the probability that the vegetation image belongs to each growth stage according to the fused feature map, and taking the growth stage with the largest probability value as the detection result:

[0068] P = Softmax(WV r +b)

[0069] P=[p1,…,p i ,…,p n ]

[0070] Among them, P represents the probability that the vegetation image belongs to each growth stage, p i represents the probability that the vegetation image belongs to the i-th growth stage, Softmax represents the Softmax activation function, W represents the weight matrix, V r represents the fused feature map, and b represents the bias term.

[0071] It should be noted that by calculating the probability of each growth stage through the Softmax function, the detection head can quantify the possibility that the image belongs to each growth stage.

[0072] In the present invention, by fusing information at different scales, the detection head can better process complex features in the image, especially when the difference in some growth stages is small or the features are not obvious, it can still make accurate judgments.

[0073] S4: Divide vegetation areas according to vegetation types and identified vegetation growth stages.

[0074] Among them, the vegetation commonly planted in front of the seawall include seashore paspalum, Spartina alterniflora, Casuarina equisetifolia, verbena, reed, Suaeda salsa and Robinia pseudoacacia, etc.

[0075] Specifically, vegetation A in the seedling stage can be regarded as one area, vegetation A in the growth stage can be regarded as another area, and vegetation B in the seedling stage can be regarded as another area.

[0076] It should be noted that vegetation of different types and growth stages has different requirements for water, nutrition and light. Dividing vegetation into different areas according to type and growth stage can provide precise irrigation according to the specific needs of each area, provide the most suitable growth conditions for vegetation at each stage, and promote the healthy growth and development of vegetation.

[0077] S5: Based on soil particle size, average temperature, precipitation, air humidity, normalized water index, vegetation type and vegetation growth stage, the artificial neural network is used to predict the water supply per unit area required for each vegetation area.

[0078] Among them, Artificial Neural Network (ANN) is a computational model inspired by biological neural networks and is widely used in pattern recognition, classification, regression, image processing, speech recognition and other fields. ANN processes complex input data by simulating the connection and signal transmission between brain neurons, and learns to extract features from data and build models to perform prediction or decision-making tasks.

[0079] It should be noted that different vegetation areas have different water requirements due to different soil conditions, climate environments, vegetation types and growth stages. ANN can combine these multi-dimensional information to calculate the precise water supply per unit area for each area, achieving personalized precision irrigation.

[0080] In a possible implementation, the artificial neural network includes an input layer, a hidden layer, an output layer, and a prediction layer. S5 specifically includes sub-steps S501 to S503:

[0081] S501: In the input layer, a state vector is input, the state vector includes multiple state parameters, and the state parameters specifically include: soil particle size, average temperature, precipitation, air humidity, normalized water index, vegetation type and vegetation growth stage.

[0082] S502: In each neuron in the hidden layer, a weighted sum is performed on the input state vector to obtain a hidden state vector:

[0083]

[0084] Among them, y j represents the hidden state of the jth hidden layer neuron output, σ1 represents the hidden layer activation function, W j represents the weight vector of the jth hidden layer neuron, T represents the transpose operation, X represents the input state vector, and b j represents the bias term of the jth hidden layer neuron, ω ij represents the weight of the i-th state parameter in the j-th hidden layer neuron, x i represents the i-th state parameter value, and n represents the total number of state parameters.

[0085] S503: Each neuron in the output layer outputs the predicted value of water supply per unit area:

[0086]

[0087] Among them, a represents the output unit area water supply prediction value, σ2 represents the output layer activation function, ω j represents the connection weight between the jth hidden layer neuron and the output layer, b f represents the bias term of the output layer, and m represents the total number of neurons in the hidden layer.

[0088] In the present invention, artificial neural networks are used to predict the water supply per unit area of ​​vegetation areas. By combining multi-dimensional environmental data and vegetation information, accurate, dynamic and automated irrigation management can be achieved. This can not only improve the efficiency of water resource utilization and promote the healthy growth of vegetation, but also improve the overall management efficiency and ecological benefits, thus providing strong support for sustainable development.

[0089] For the training method of artificial neural network, gradient descent method can be used.

[0090] Specifically, a mean square error loss function for the water supply prediction value can be constructed, and then the network parameters of the artificial neural network can be updated according to the following formula:

[0091]

[0092] Among them, θ t+1 represents the model parameters at the t+1th iteration, θ t represents the model parameters at the tth iteration, θ t-1 represents the model parameters at the t-1th iteration, η t represents the adaptive learning rate at the tth iteration, L represents the loss function, Represents the loss function L on the network parameter θ t is the gradient of , and β represents the momentum coefficient.

[0093] It should be noted that the introduction of the momentum term helps to reduce the oscillation of parameter updates during training, especially when faced with noisy gradients or complex loss surfaces, and can provide a smoother optimization path.

[0094] Optionally, the adaptive learning rate is calculated as:

[0095]

[0096] Among them, η min Represents the minimum learning rate, η max Represents the maximum learning rate, T max represents the maximum number of iterations, and cos represents the cosine function.

[0097] In the present invention, in the early stage of training, a higher learning rate helps to quickly find the convergence direction; in the later stage of training, a lower learning rate helps to carefully optimize the model parameters. By dynamically adjusting the learning rate, the model can be exposed to different learning rate ranges during the training process, which helps to improve the generalization ability of the model and reduce the risk of overfitting.

[0098] S6: According to the water supply per unit area required by each vegetation area, the vegetation in each vegetation area is irrigated through the seawall front irrigation system.

[0099] In a possible implementation, S6 specifically includes sub-steps S601 to S607:

[0100] S601: Calculate the total water supply according to the water supply per unit area required by each vegetation area:

[0101]

[0102] Among them, A represents the total water supply, a k represents the water supply per unit area required by the kth vegetation area, S k Represents the area of ​​the kth vegetation area.

[0103] S602: Determine the initial opening parameters of the reservoir ramp gate according to the total water supply:

[0104] β=bA

[0105] Among them, β represents the initial opening parameter of the reservoir ramp, and b represents the opening coefficient.

[0106] S603: Controlling the ramp gate of the water reservoir according to the initial opening parameters of the ramp gate of the water reservoir.

[0107] In the present invention, the initial opening parameter of the ramp gate is determined by calculating the total water supply. This enables the system to automatically adjust the ramp gate opening, avoiding the error and complexity of manual adjustment, and improving the intelligence level of the entire irrigation system. The ramp gate opening is reasonably determined so that water can be supplied to each vegetation area smoothly and efficiently, reducing the resistance and waste that may occur during the water supply process, thereby reducing the energy consumption of equipment such as water pumps.

[0108] S604: During the irrigation process, based on the PID controller, the opening parameters of the ramp gate of the water storage tank are adjusted in real time with the goal of achieving the total water supply.

[0109] Specifically, for the PID controller, the difference value e between the real-time water supply and the total water supply is calculated.

[0110] According to the difference between the hourly water supply and the total water supply, the proportional term, integral term and differential term are calculated:

[0111] [P]=K p e

[0112] [I]=K i ∫edt

[0113]

[0114] Where [P] represents the proportional term, [I] represents the integral term, [D] represents the differential term, and K p Represents the proportional gain coefficient, K i Indicates the integral gain coefficient, K d represents the differential gain coefficient, e represents the difference between the real-time water supply and the total water supply, t represents time, and dt represents the differentiation of t;

[0115] According to the proportional term, integral term and differential term, the PID control parameters are calculated:

[0116]

[0117] Among them, PID represents PID control parameters;

[0118] According to the PID control parameters, the opening parameters of the reservoir ramp gate are adjusted in real time.

[0119] In the present invention, the PID controller can dynamically adjust according to the difference between the real-time water supply and the target total water supply to ensure that the actual water supply keeps up with the demand. This real-time adjustment mechanism enables the system to quickly respond to any changes in water supply demand and ensure the accuracy of water supply.

[0120] S605: Determine the initial opening parameters of the valve according to the water supply per unit area required by each vegetation area and the number of irrigation holes contained:

[0121]

[0122] Among them, λ k represents the initial opening parameter of the valve in the kth vegetation area, c k Represents the number of irrigation holes contained in the kth vegetation area.

[0123] S606: Control the valves in each vegetation area according to the initial opening parameters of the valves.

[0124] In the present invention, by incorporating the number of irrigation holes into the calculation of the opening parameters, the system can evenly distribute water resources according to the actual distribution of irrigation holes, ensuring that the amount of water output by each irrigation hole is consistent, and avoiding the problem of uneven water distribution caused by differences in the number of irrigation holes.

[0125] S607: During the irrigation process, based on the PID controller, the opening parameters of the valves in each vegetation area are adjusted in real time with the goal of achieving the required water supply per unit area in each vegetation area.

[0126] Among them, the principle of the PID controller for controlling the opening parameters of the valves in each vegetation area is consistent with that of the PID controller for controlling the opening parameters of the reservoir ramp gate. To avoid repetition, the present invention will not be described in detail.

[0127] In the present invention, the PID controller can respond to the actual water supply demand changes of each vegetation area in real time during the irrigation process, and ensure that each area always obtains accurate water supply by adjusting the valve opening. This ensures that the vegetation in each area obtains appropriate moisture and prevents over-irrigation or under-irrigation.

[0128] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0129] In the present invention, the required water supply per unit area for each vegetation area is predicted through an artificial neural network according to soil particle size, average temperature, precipitation, air humidity, normalized water index, vegetation type and vegetation growth stage. The vegetation in each vegetation area is intelligently irrigated through the irrigation system in front of the seawall, which reduces the cost of manual operation and management, can provide the best growth conditions for the vegetation in front of the seawall, promote the health and growth of the vegetation, help maintain and improve the stability of the ecosystem in the area in front of the seawall, and support biodiversity.

[0130] Reference Manual Attached Figure 3 , showing a structural schematic diagram of an intelligent irrigation control system for vegetation in front of a seawall provided by the present invention.

[0131] The present invention also provides an intelligent irrigation control system 20 for vegetation in front of a seawall, comprising:

[0132] Processor 201;

[0133] The memory 202 stores computer-readable instructions, and when the computer-readable instructions are executed by the processor 201, the intelligent irrigation control method for vegetation in front of the seawall as described in the method embodiment is implemented.

[0134] The intelligent irrigation control system 20 for vegetation in front of a seawall provided by the present invention can execute the above-mentioned intelligent irrigation control method for vegetation in front of a seawall and achieve the same or similar technical effects. To avoid repetition, the present invention will not go into details.

[0135] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0136] In the present invention, the required water supply per unit area for each vegetation area is predicted through an artificial neural network according to soil particle size, average temperature, precipitation, air humidity, normalized water index, vegetation type and vegetation growth stage. Through the irrigation system in front of the seawall, the vegetation in each vegetation area is intelligently irrigated, which reduces the cost of manual operation and management, can provide the best growth conditions for the vegetation in front of the seawall, promote the health and growth of the vegetation, help maintain and improve the stability of the ecosystem in the area in front of the seawall, and support biodiversity.

[0137] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0138] There are a few points to note:

[0139] (1) The drawings of the embodiments of the present invention only relate to the structures related to the embodiments of the present invention, and other structures may refer to the general design.

[0140] (2) For the sake of clarity, the thickness of a film or region in the drawings used to describe the embodiments of the present invention is enlarged or reduced, that is, these drawings are not drawn according to the actual scale. It is understood that when an element such as a film, a region or a substrate is referred to as being "on" or "under" another element, the element may be "directly" "on" or "under" the other element or there may be intermediate elements.

[0141] (3) In the absence of conflict, the embodiments of the present invention and the features therein may be combined with each other to obtain new embodiments.

[0142] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. The protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for controlling intelligent irrigation of vegetation in front of a seawall, characterized in that: include: S1: Construct an irrigation system in front of the seawall; S2: obtain vegetation images; S3: According to the vegetation image, identifying the vegetation growth stage through a convolutional neural network; S4: Divide vegetation areas according to vegetation types and identified vegetation growth stages; S5: Based on soil particle size, average temperature, precipitation, air humidity, normalized water index, vegetation type and vegetation growth stage, the required water supply per unit area of ​​each vegetation area is predicted through artificial neural network; S6: Irrigating the vegetation in each vegetation area through the seawall irrigation system according to the water supply per unit area required by each vegetation area.

2. The intelligent irrigation control method for vegetation in front of a seawall according to claim 1 is characterized in that: The seawall irrigation system specifically includes: a water diversion channel, a sedimentation tank, a water storage tank, a hose and a valve; The water diversion channel outside the seawall is used as a water source. Seawater enters the reservoir after being precipitated in the sedimentation tank. The reservoir is connected to the vegetation planting area through the hose. An irrigation hole is opened on the hose. The valve is set at the irrigation hole. The vegetation is drip-irrigated by opening the valve.

3. The intelligent irrigation control method for vegetation in front of a seawall according to claim 1 is characterized in that: The convolutional neural network includes: a backbone feature extraction network, a feature fusion network and a vegetation growth stage detection head; S3 specifically includes: S301: extracting a multi-scale feature map in the vegetation image through a backbone feature extraction network; S302: fusing the multi-scale feature maps through a feature fusion network to obtain a fused feature map; S303: Determine the vegetation growth stage according to the fused feature map through the vegetation growth stage detection head.

4. The intelligent irrigation control method for vegetation in front of a seawall according to claim 3 is characterized in that: The backbone feature extraction network includes serially connected convolution units, namely, a first convolution unit, a second convolution unit, and a third convolution unit; S301 specifically includes: S3011: taking the vegetation image as input of the first convolution unit, and outputting a small-scale feature map in the vegetation image through the first convolution unit; S3012: using the small-scale feature map as input of the second convolution unit, and outputting a medium-scale feature map in the vegetation image through the second convolution unit; S3013: Taking the medium-scale feature map as the input of the third convolution unit, and outputting a large-scale feature map in the vegetation image through the third convolution unit.

5. The intelligent irrigation control method for vegetation in front of a seawall according to claim 4 is characterized in that: The S3011 specifically includes: The fabric image is convolved through a 3×3 first convolution unit to obtain an initial feature map; Copying the initial feature map to obtain a first initial feature map and a second initial feature map; Separating the first initial feature map according to channels to obtain a first branch feature map and a second branch feature map; Convolutionally processing the second branch feature map through a first grouped convolution unit, a depthwise separable convolution unit, and a second grouped convolution unit to obtain a second branch separation feature map; Connecting the first branch feature map and the second branch separation feature map to obtain a connection feature map; Performing channel shuffling on the connection feature graph to obtain a shuffled feature graph; Symmetrically flipping the second initial feature map to obtain a flipped feature map; The shuffled feature map and the flipped feature map are element-wise multiplied, and convolution is performed through a 3×3 second convolution unit to obtain an enhanced feature map; The enhanced feature map is added to the initial feature map, and convolution processing is performed through a 3×3 third convolution unit to obtain the small-scale feature map.

6. The intelligent irrigation control method for vegetation in front of a seawall according to claim 4 is characterized in that: The feature fusion network is specifically a feature pyramid network, and S302 specifically includes: Upsampling the large-scale feature map so that the large-scale feature map matches the size of the medium-scale feature map; Performing weighted fusion on the upsampled large-scale feature map and the medium-scale feature map to obtain a preliminary fused feature map; Upsampling the preliminary fused feature map so that the size of the preliminary fused feature map matches that of the small-scale feature map; The upsampled preliminary fusion feature map is weightedly fused with the small-scale feature map to obtain a fusion feature map.

7. The intelligent irrigation control method for vegetation in front of a seawall according to claim 4 is characterized in that: The S303 is specifically as follows: According to the fused feature map, the probability that the vegetation image belongs to each growth stage is calculated, and the growth stage with the largest probability value is taken as the detection result: P=Softmax(WV r +b) P=[p1,…,p i ,…,p n ] Where P represents the probability that the vegetation image belongs to each growth stage, p i represents the probability that the vegetation image belongs to the i-th growth stage, Softmax represents the Softmax activation function, W represents the weight matrix, V r represents the fused feature map, and b represents the bias term.

8. The intelligent irrigation control method for vegetation in front of a seawall according to claim 1 is characterized in that: The artificial neural network includes an input layer, a hidden layer, an output layer and a prediction layer; S5 specifically includes: S501: In the input layer, input a state vector, the state vector includes a plurality of state parameters, the state parameters specifically include: soil particle size, average temperature, precipitation, air humidity, normalized water index, vegetation type and vegetation growth stage; S502: Each neuron in the hidden layer performs weighted summation on the input state vector to obtain a hidden state vector: Among them, y j represents the hidden state of the jth hidden layer neuron output, σ1 represents the hidden layer activation function, W j represents the weight vector of the jth hidden layer neuron, T represents the transpose operation, X represents the input state vector, and b j represents the bias term of the jth hidden layer neuron, ω ij represents the weight of the i-th state parameter in the j-th hidden layer neuron, x i represents the i-th state parameter value, and n represents the total number of state parameters; S503: Each neuron in the output layer outputs a predicted value of water supply per unit area: Among them, a represents the output unit area water supply prediction value, σ2 represents the output layer activation function, ω j represents the connection weight between the jth hidden layer neuron and the output layer, b f represents the bias term of the output layer, and m represents the total number of neurons in the hidden layer.

9. The intelligent irrigation control method for vegetation in front of a seawall according to claim 2 is characterized in that: The S6 specifically includes: S601: Calculate the total water supply according to the water supply per unit area required by each vegetation area: Among them, A represents the total water supply, a k represents the water supply per unit area required by the kth vegetation area, S k represents the area of ​​the kth vegetation region; S602: Determine the initial opening parameters of the ramp gate of the reservoir according to the total water supply: β=bA Among them, β represents the initial opening parameter of the reservoir ramp, and b represents the opening coefficient; S603: controlling the ramp gate of the water reservoir according to the initial opening parameter of the ramp gate of the water reservoir; S604: During the irrigation process, based on the PID controller, the opening parameters of the ramp gate of the water storage tank are adjusted in real time with the goal of achieving the total water supply; S605: Determine the initial opening parameters of the valve according to the water supply per unit area required by each vegetation area and the number of irrigation holes contained therein: Among them, λ k represents the initial opening parameter of the valve in the kth vegetation area, c k represents the number of irrigation holes contained in the kth vegetation area; S606: Controlling the valves in each vegetation area according to the initial opening parameters of the valves; S607: During the irrigation process, based on the PID controller, the opening parameters of the valves in each vegetation area are adjusted in real time with the goal of achieving the required water supply per unit area in each vegetation area.

10. An intelligent irrigation control system for vegetation in front of a seawall, characterized in that: include: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the intelligent irrigation control method for vegetation in front of a seawall as described in any one of claims 1 to 9 is implemented.

Images