A Permanent Magnet Synchronous Motor Control Method and System Based on Image Processing

Abstract

A permanent magnet synchronous motor control method and system based on image processing provided by the present invention relate to the technical field of motor control. The method includes obtaining lidar data of a field; determining a plurality of initial watering point information and a plurality of initial fertilizing point information based on the lidar data of the field; generating an initial motor control scheme based on the plurality of initial watering point information and the plurality of initial fertilizing point information; controlling a permanent magnet synchronous motor to perform initial watering and fertilizing based on the initial motor control scheme, and obtaining lidar data of the field after the initial watering and fertilizing; generating a target motor control scheme based on the lidar data of the field after the initial watering and fertilizing; and controlling the permanent magnet synchronous motor based on the target motor control scheme. This method can accurately determine the motor control scheme for watering and fertilizing the field.

Description

Technical Field

[0001] The present invention relates to the technical field of motor control, and particularly to a permanent magnet synchronous motor control method and system based on image processing. Background Art

[0002] In modern agricultural production, how to perform precise watering and fertilization management is a key factor in ensuring crop yield and quality. Precise watering and fertilization operations can improve crop yield and quality and avoid waste of resources. Currently, it mainly relies on manual experience judgment and the method of uniform spraying in a fixed mode. Among them, the method of making decisions and performing watering and fertilization by observing the field conditions manually can analyze the state differences in different areas of the field, but there are problems such as low efficiency, time-consuming and laborious, and strong subjectivity in judgment criteria; the method of uniform spraying in a fixed mode improves the operation efficiency, but it cannot perform precise regulation according to the actual differences in water and fertilizer requirements of the field, which may not only cause waste of water resources and fertilizers, but also affect the balanced growth of crops. Although some automated motor devices have been introduced in the prior art, there is still a lack in intelligent decision-making and still cannot perform precise regulation.

[0003] Therefore, how to accurately determine the motor control scheme for watering and fertilizing the field is an urgent problem to be solved currently. Summary of the Invention

[0004] The main technical problem to be solved by the present invention is how to accurately determine the motor control scheme for watering and fertilizing the field.

[0005] According to a first aspect, the present invention provides a permanent magnet synchronous motor control method based on image processing, including: obtaining lidar data of a field; determining a plurality of initial watering point information and a plurality of initial fertilizing point information based on the lidar data of the field; generating an initial motor control scheme based on the plurality of initial watering point information and the plurality of initial fertilizing point information; controlling a permanent magnet synchronous motor to perform initial watering and fertilization based on the initial motor control scheme, and obtaining lidar data of the field after the initial watering and fertilization; generating a target motor control scheme based on the lidar data of the field after the initial watering and fertilization; and controlling the permanent magnet synchronous motor based on the target motor control scheme.

[0006] In a possible implementation manner, the determining a plurality of initial watering point information and a plurality of initial fertilizing point information based on the lidar data of the field includes: generating an initial water content distribution map and an initial fertilizer demand map based on the lidar data of the field; and obtaining a plurality of initial watering point information and a plurality of initial fertilizing point information based on the initial water content distribution map and the initial fertilizer demand map.

[0007] In a possible implementation, generating a target motor control scheme based on the lidar data of the field after the initial watering and fertilization includes: generating a target water content distribution map and a target fertilizer demand map based on the lidar data of the field after the initial watering and fertilization; constructing a knowledge graph, which includes a plurality of initial watering point nodes and a plurality of initial fertilization point nodes. The node features of the initial watering point nodes include initial watering point information and the target water content distribution map, and the node features of the initial fertilization point nodes include initial fertilization point information and the target fertilizer demand map. The edges between different initial watering point nodes are the distances between different initial watering points, the edges between different initial fertilization point nodes are the distances between different initial fertilization points, and the edges between the initial watering point nodes and the initial fertilization point nodes are the distances between the initial watering point and the initial fertilization point; processing the knowledge graph based on a graph neural network model to determine a plurality of target watering point information and a plurality of target fertilization point information; generating a target motor control scheme based on the plurality of target watering point information and the plurality of target fertilization point information.

[0008] In a possible implementation, the initial watering point information includes the initial watering point location and the initial watering amount.

[0009] According to a second aspect, the present invention provides a permanent magnet synchronous motor control system based on image processing, including: an acquisition module for acquiring lidar data of a field; an information determination module for determining a plurality of initial watering point information and a plurality of initial fertilization point information based on the lidar data of the field; an initial scheme generation module for generating an initial motor control scheme based on the plurality of initial watering point information and the plurality of initial fertilization point information; an initial control module for controlling a permanent magnet synchronous motor to perform initial watering and fertilization based on the initial motor control scheme and acquiring lidar data of the field after the initial watering and fertilization; a target scheme generation module for generating a target motor control scheme based on the lidar data of the field after the initial watering and fertilization; and a target control module for controlling the permanent magnet synchronous motor based on the target motor control scheme.

[0010] In a possible implementation, the information determination module is further configured to: generate an initial water content distribution map and an initial fertilizer demand map based on the lidar data of the field; and obtain a plurality of initial watering point information and a plurality of initial fertilization point information based on the initial water content distribution map and the initial fertilizer demand map.

[0011] In a possible implementation, the target scheme generation module is further configured to: generate a target water content distribution map and a target fertilizer requirement map based on the lidar data of the field after the initial watering and fertilization; construct a knowledge graph, where the knowledge graph includes a plurality of initial watering point nodes and a plurality of initial fertilization point nodes, the node features of the initial watering point nodes include initial watering point information and the target water content distribution map, the node features of the initial fertilization point nodes include initial fertilization point information and the target fertilizer requirement map, the edges between different initial watering point nodes are the distances between different initial watering points, the edges between different initial fertilization point nodes are the distances between different initial fertilization points, and the edges between the initial watering point nodes and the initial fertilization point nodes are the distances between the initial watering points and the initial fertilization points; process the knowledge graph based on a graph neural network model to determine a plurality of target watering point information and a plurality of target fertilization point information; and generate a target motor control scheme based on the plurality of target watering point information and the plurality of target fertilization point information.

[0012] In a possible implementation, the initial watering point information includes the initial watering point location and the initial watering amount.

[0013] A permanent magnet synchronous motor control method and system based on image processing provided by the present invention, the method includes obtaining lidar data of a field; determining a plurality of initial watering point information and a plurality of initial fertilization point information based on the lidar data of the field; generating an initial motor control scheme based on the plurality of initial watering point information and the plurality of initial fertilization point information; controlling a permanent magnet synchronous motor to perform initial watering and fertilization based on the initial motor control scheme, and obtaining lidar data of the field after the initial watering and fertilization; generating a target motor control scheme based on the lidar data of the field after the initial watering and fertilization; and controlling the permanent magnet synchronous motor based on the target motor control scheme, and this method can accurately determine the motor control scheme for watering and fertilizing the field. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic flowchart of a permanent magnet synchronous motor control method based on image processing provided by an embodiment of the present invention;

[0015] Figure 2 It is a schematic flowchart of determining a plurality of initial watering point information and a plurality of initial fertilization point information provided by an embodiment of the present invention;

[0016] Figure 3 It is a schematic flowchart of generating a target motor control scheme provided by an embodiment of the present invention;

[0017] Figure 4 It is a schematic diagram of a permanent magnet synchronous motor control system based on image processing provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In the embodiments of the present invention, a permanent magnet synchronous motor control method based on image processing as shown in Figure 1 is provided. The permanent magnet synchronous motor control method based on image processing includes steps S1 to S6:

[0019] Step S1: Obtain the LiDAR data of the field.

[0020] LiDAR (Light Detection and Ranging) is an advanced active remote sensing device. LiDAR can accurately measure the distance of the target object by emitting a laser beam to the target and receiving the reflected laser signal, and based on the round-trip time of the laser pulse. In some embodiments, LiDAR can be used to scan the field in all directions and obtain high-precision three-dimensional information on the field surface, including information such as terrain, crop height, and canopy structure, so as to analyze information such as the soil conditions and crop growth status of the field. [[ID=!1]]

[0021] For example, multiple LiDAR devices can be installed in the field. These devices are accurately calibrated and set to monitor the entire field. When the laser beam emitted by the LiDAR irradiates the crops and soil surface in the field, reflections will occur. The LiDAR device can calculate the distance between each measurement point and the device by analyzing the intensity and time delay of the reflected laser signal, and combine the position information and scanning angle of the device, so as to construct the LiDAR data of the field.

[0022] LiDAR data is a digital data set obtained by LiDAR through emitting laser beams, receiving reflected signals, and calculation, which is used to characterize information such as the terrain and crop growth status of the field. Based on the LiDAR data, the crop growth differences in different regions of the field and the soil environment differences in different regions can be accurately analyzed.

[0023] Step S2: Determine multiple initial watering point information and multiple initial fertilization point information based on the LiDAR data of the field.

[0024] In some embodiments, multiple initial watering point information and multiple initial fertilization point information can be determined through the Figure 2 process. Figure 2 FIG. is a schematic flow chart of a process for determining multiple initial watering point information and multiple initial fertilization point information provided by the embodiments of the present invention. The determination of multiple initial watering point information and multiple initial fertilization point information includes steps S21 to S22:

[0025] Step S21: Generate an initial water content distribution map and an initial fertilizer requirement map based on the LiDAR data of the field.

[0026] In some embodiments, an initial water content distribution map and an initial fertilizer requirement map are generated using a first generative adversarial network based on the lidar data of the field. The input of the first generative adversarial network is the lidar data of the field, and the outputs of the first generative adversarial network are the initial water content distribution map and the initial fertilizer requirement map.

[0027] Generative adversarial networks (GANs) consist of a generator and a discriminator. The generator is responsible for generating new data samples based on the input data, and the discriminator is used to judge the difference between the data samples generated by the generator and the real data, and evaluate the authenticity of the generated samples. The two compete with each other and make progress together.

[0028] The initial water content distribution map is a visualization chart of the water content distribution at different positions in the field generated by generative adversarial networks (GANs). The initial water content distribution map clearly presents the difference distribution of water content between different regions of the field.

[0029] The initial fertilizer requirement map is a visualization chart generated by generative adversarial networks (GANs) for showing the degree of fertilizer requirement in different regions of the field. Through the initial fertilizer requirement map, the demand distribution of fertilizer in each region of the field can be intuitively presented.

[0030] The lidar data of the field contains rich information such as field terrain, crop height, vegetation coverage and other data. The generator in the generative adversarial network can learn and analyze these data, and find out the laws and data characteristics closely related to the water content and fertilizer requirement of the field. The discriminator continuously evaluates the results generated by the generator, and prompts the generator to improve the generated images, so that they are more in line with the real situation. Through this adversarial training mechanism, the generative adversarial network can gradually master the corresponding conversion method from the lidar data of the field to the initial water content distribution map and the initial fertilizer requirement map, and then generate the initial water content distribution map and the initial fertilizer requirement map more accurately.

[0031] Step S22, obtaining a plurality of initial watering point information and a plurality of initial fertilization point information based on the initial water content distribution map and the initial fertilizer requirement map.

[0032] In some embodiments, a plurality of initial watering point information and a plurality of initial fertilization point information can be obtained using a deep neural network model based on the initial water content distribution map and the initial fertilizer requirement map. The input of the deep neural network model is the initial water content distribution map and the initial fertilizer requirement map, and the outputs of the deep neural network model are a plurality of initial watering point information and a plurality of initial fertilization point information.

[0033] The deep neural network model includes a Deep Neural Network (DNN). A deep neural network can be composed of multiple processing layers, and each processing layer contains a large number of neurons. Neurons process information by performing matrix transformations on the input data, and the parameters relied on for matrix operations are continuously obtained during the model training process. By adjusting and optimizing these parameters, the deep neural network model can possess a powerful non-linear fitting ability, thereby being able to process complex data patterns and correlation relationships.

[0034] The initial watering point information includes the initial watering point location and the initial watering amount.

[0035] The multiple initial fertilization point information is a set of information determined by the deep neural network model based on the initial water content distribution map and the initial fertilizer requirement map, and includes information such as the fertilization location, fertilizer type, and fertilization amount.

[0036] The deep neural network model has powerful learning and analysis capabilities. It can automatically extract the feature information in the initial water content distribution map and the initial fertilizer requirement map, and mine the patterns and rules in the feature information. The deep neural network model can also be trained with a large amount of data, thereby learning the complex relationships between different water content and fertilizer requirement patterns and the actual watering and fertilization point locations. For example, the deep neural network model can discover the optimal watering and fertilization locations corresponding to specific water content and fertilizer requirement combinations, and convert the input image information into specific multiple initial watering point information and multiple initial fertilization point information, thereby achieving precise positioning.

[0037] In some embodiments, obtaining the multiple initial watering point information and the multiple initial fertilization point information based on the initial water content distribution map and the initial fertilizer requirement map includes:

[0038] Using the deep neural network model based on the initial water content distribution map and the initial fertilizer requirement map to determine multiple water shortage point information, multiple fertilizer shortage point information, and the K value of K-means clustering;

[0039] Based on the K value, clustering the multiple water shortage point information and the multiple fertilizer shortage point information respectively to obtain multiple water shortage point clusters and multiple fertilizer shortage point clusters, taking the water shortage points corresponding to the centers in each water shortage point cluster as the initial watering points, and taking the fertilizer shortage points corresponding to the centers in each fertilizer shortage point cluster as the initial fertilization points.

[0040] Through K-means clustering, neighboring water shortage points / fertilizer shortage points are merged into clusters, and each cluster corresponds to a watering point or a fertilization point, avoiding repeated operations and reducing resource waste. The cluster center represents the "average demand" of the area, ensuring that the watering or fertilization amount matches the actual demand of the area.

[0041] Step S3: Generate an initial motor control scheme based on the multiple initial watering point information and the multiple initial fertilizing point information.

[0042] In some embodiments, a second generative adversarial network is used to generate an initial motor control scheme based on the multiple initial watering point information and the multiple initial fertilizing point information. The input of the second generative adversarial network is the multiple initial watering point information, and the output of the second generative adversarial network is the initial motor control scheme.

[0043] The initial motor control scheme is a set of operation instructions for controlling the operation of the motor generated by the second generative adversarial network based on the multiple initial watering point information and the multiple initial fertilizing point information. By controlling the motor to drive the watering and fertilizing equipment to carry out operations at the preset points and observing the preliminary operation effects, the scheme can be further optimized to provide more precise watering and fertilizing control.

[0044] Generative adversarial networks (GANs) have their unique structures and training mechanisms and possess powerful capabilities for learning and generating complex patterns. When generating the initial motor control scheme, the generator can deeply mine and learn various data features in the multiple initial watering point information and the multiple initial fertilizing point information. For example, the generator understands the path and distance that the equipment needs to move from the position information of the watering points and fertilizing points, and then generates control instructions such as the rotation direction and operation duration of the motor; the generator can also deduce the power required by the motor from the expected watering volume and fertilizing volume information to control the output of the watering and fertilizing equipment. At the same time, the discriminator can continuously evaluate the control scheme generated by the generator, compare it with the ideal control scheme that meets the actual operation requirements, and feedback the differences to the generator to prompt the generator to continuously improve and optimize the generated scheme. Through the training of a large amount of data and this adversarial learning process, the generative adversarial network gradually masters the internal mapping relationship from the initial watering and fertilizing point information to the motor control scheme, so as to generate an initial motor control scheme that better meets the actual needs.

[0045] Step S4: Control the permanent magnet synchronous motor to perform initial watering and fertilizing based on the initial motor control scheme, and obtain the lidar data of the field after the initial watering and fertilizing.

[0046] The lidar data of the field after the initial watering and fertilizing refers to the three-dimensional space information of the field after the initial watering and fertilizing obtained by lidar scanning. The lidar data includes information such as crop height, soil topography, and vegetation density, and can be used to evaluate the actual effect of the current initial motor control scheme to ensure that the final watering and fertilizing operations are more accurate and effective.

[0047] Step S5: Generate a target motor control scheme based on the lidar data of the field after the initial watering and fertilizing.

[0048] In some embodiments, it is also possible to Figure 3 generate the target motor control scheme through the Figure 3 flow shown in the flowchart of generating the target motor control scheme provided by the embodiment of the present invention. The generation of the target motor control scheme includes steps S31 to S34:

[0049] Step S31, generate a target water content distribution map and a target fertilizer demand map based on the lidar data of the field after the initial watering and fertilization.

[0050] In some embodiments, a first generative adversarial network is used to generate a target water content distribution map and a target fertilizer demand map based on the lidar data of the field after the initial watering and fertilization. The input of the first generative adversarial network is the lidar data of the field after the initial watering and fertilization, and the output of the first generative adversarial network is the target water content distribution map and the target fertilizer demand map.

[0051] The target water content distribution map is a visual chart of the water distribution generated by processing the lidar data of the field after the initial watering and fertilization through the first generative adversarial network. This chart calculates the actual penetration depth and lateral diffusion range of water in the soil by analyzing characteristic parameters such as the soil surface reflectivity and vegetation canopy water content collected by the lidar. The target water content distribution map can intuitively display the actual water distribution status of the field after the initial watering and fertilization.

[0052] The target fertilizer demand map is a visual chart of the fertilizer demand generated by the first generative adversarial network based on the lidar data after the initial watering and fertilization. The target fertilizer demand map shows the real-time fertilizer demand distribution of crops in different regions, thus providing a basis for determining the final fertilization strategy.

[0053] Step S32, construct a knowledge graph, the knowledge graph includes a plurality of initial watering point nodes and a plurality of initial fertilization point nodes. The node features of the initial watering point nodes include initial watering point information and the target water content distribution map. The node features of the initial fertilization point nodes include initial fertilization point information and the target fertilizer demand map. The edges between different initial watering point nodes are the distances between different initial watering points. The edges between different initial fertilization point nodes are the distances between different initial fertilization points. The edges between the initial watering point nodes and the initial fertilization point nodes are the distances between the initial watering points and the initial fertilization points.

[0054] A knowledge graph is a data structure that can be used to represent the relationships between nodes (vertices) and the edges between nodes. Multiple nodes include an initial watering point node and an initial fertilization point node. The node features of the initial watering point node include the initial watering point information and the target water content distribution map, and the node features of the initial fertilization point node include the initial fertilization point information and the target fertilizer demand map. The features of the edges between nodes include the distances between different initial watering points, the distances between different initial fertilization points, and the distances between the initial watering points and the initial fertilization points. By constructing a knowledge graph, the spatial relationships between the initial watering points and the initial fertilization points in the field can be visually displayed.

[0055] Step S33: Process the knowledge graph based on the graph neural network model to determine multiple target watering point information and multiple target fertilization point information.

[0056] A graph neural network model (Graph Neural Network, GNN) is a neural network that can directly operate on knowledge graph data. The graph neural network model consists of a graph neural network layer and a fully connected layer. The graph neural network layer is responsible for processing the information of nodes and edges in the knowledge graph and capturing the complex relationships between nodes, while the fully connected layer can further integrate and classify the features output by the graph neural network layer. The input of the graph neural network model is the knowledge graph, and the output of the graph neural network model is multiple target watering point information and multiple target fertilization point information.

[0057] The graph structure naturally represents the spatial coordination relationship between the watering operation and the fertilization operation. By taking the initial watering point and the initial fertilization point as nodes and using distance features to construct edges, this relationship can intuitively reflect the actual distribution pattern of each operation point in the field. By organizing the discrete operation points into a graph through spatial relationships, both the accuracy of the original data is retained, and a semantic relationship network that can be processed by the graph neural network is constructed.

[0058] The features of the initial watering point node include the initial watering point location, the initial watering amount, and the target water content distribution map. The features of the initial fertilization point node include the initial fertilization point location, the initial fertilization amount, and the target fertilizer demand map. These features together provide the geographical location of the fertilization point, the previous fertilizer application amount, and the current nutrient demand distribution of the crop. Through this strictly corresponding combination of node features, the graph neural network model can make collaborative decisions based on the complete watering and fertilization operation history data and real-time monitoring data to obtain accurate results.

[0059] The edges between the initial watering point nodes can calculate the spatial correlation through the graph neural network model and determine which watering points need to be merged and optimized. The edges between the initial fertilization point nodes can be used to evaluate the regional coverage density of the fertilization points, so as to calculate whether the distribution of the fertilization points is uniform. The edges between the initial watering point nodes and the initial fertilization point nodes can be used to establish the association rules between the watering points and the fertilization points, and ensure that the determined target points meet the requirements of collaborative operations through the connection relationship of the edges.

[0060] The information of multiple target watering points is a data set of specific watering points in the field determined by the graph neural network model. Each piece of target watering point information includes the point coordinates of the watering point and the amount of water to be poured corresponding to each watering point.

[0061] The information of multiple target fertilization points is a data set of specific fertilization points in the field determined by the graph neural network model. The information of multiple target fertilization points includes the point coordinates of the fertilization points, the type of fertilizer and the amount of fertilizer applied corresponding to each fertilization point.

[0062] The core advantage of the graph neural network model in processing the knowledge graph lies in its unique spatial relationship modeling ability. Through the message passing mechanism of the graph neural network layer, the graph neural network model can perform feature interaction learning on the initial watering point nodes and the initial fertilization point nodes. In this process, each node will dynamically adjust the receiving weight of the information of adjacent nodes according to the distance characteristics of its connected edges. For example, a fertilization point node with a shorter distance will have a greater impact on the feature update of the watering point node. The fully connected layer then converts these learned spatial association features into executable operation parameters, including determining the specific position and optimal watering amount of each target watering point. Through this learning method based on the knowledge graph, the finally generated target watering point information and target fertilization point information can not only meet the needs of crop growth, but also achieve precise coordination between the watering operation and the fertilization operation.

[0063] Step S34, generating a target motor control scheme based on the information of the multiple target watering points and the information of the multiple target fertilization points.

[0064] In some embodiments, a third generative adversarial network is used to generate a target motor control scheme based on the information of the multiple target watering points and the information of the multiple target fertilization points. The input of the third generative adversarial network is the information of the multiple target watering points and the information of the multiple target fertilization points, and the output of the third generative adversarial network is the target motor control scheme.

[0065] The target motor control scheme is the final motor control instruction set generated by the third generative adversarial network based on the optimized information of multiple target watering points and multiple target fertilization points. The target motor control scheme details the operating parameters of the target motor, including core operation instructions such as path planning for driving the watering equipment and fertilization device to move to the specified coordinates, the residence duration at each operation point, and power adjustment parameters for controlling the output of watering and fertilization. These instructions can ensure the precise implementation of watering and fertilization operations in the field, thereby achieving the optimal distribution of water and nutrients in the field in terms of space and time dimensions.

[0066] Through its unique adversarial training mechanism, the generative adversarial network can effectively learn the complex mapping relationship from spatial point data to motor control parameters. The generator analyzes the spatial distribution characteristics of the target watering points and fertilization points and derives the optimal equipment movement trajectory and operation timing arrangement. At the same time, the discriminator evaluates the actual feasibility of the generated scheme to ensure that the output control parameters meet both agronomic requirements and equipment operation constraints, so as to generate the target motor control scheme with the best effect and executability.

[0067] In some embodiments, the third generative adversarial network includes a regional division layer, a scheme simulation layer, and an optimal scheme decision layer. The input of the regional division layer is the information of multiple target watering points and multiple target fertilization points, and the output of the regional division layer is multiple divided regions of the field, the water demand index of each region, and the fertilizer demand index of each region. The input of the scheme simulation layer is multiple divided regions of the field, the water demand index of each region, and the fertilizer demand index of each region, and the output of the scheme simulation layer is multiple sets of simulated motor control schemes, the simulated execution effect of each scheme, and the operation score of each scheme. The input of the optimal scheme decision layer is multiple sets of simulated motor control schemes, the simulated execution effect of each scheme, and the operation score of each scheme, and the output of the optimal scheme decision layer is the target motor control scheme.

[0068] The multiple divided regions of the field are multiple regional operation units obtained by dividing the field through a spatial clustering algorithm, and each unit contains multiple watering points and fertilization points. By dividing the field into multiple regions, the water and fertilizer requirements of each independent region can be accurately quantified.

[0069] The water demand index of each region is used to evaluate the degree of water demand of crops in the region. The water demand index is an integer quantification index evaluated by analyzing all the watering point information contained in the region and according to a reference value (usually 200 ml).

[0070] The fertilizer demand index of each region is specifically used to quantify the degree of fertilizer demand of crops. The higher the fertilizer demand index, the more serious the lack of fertilizer in the crops in this region, and it needs to be prioritized in the control scheme.

[0071] Multiple sets of simulated motor control schemes are multiple differentiated control strategies initially simulated, including multiple control schemes such as simultaneous execution of water and fertilizer, priority execution of water, and alternating execution by zones. Each scheme contains detailed parameters such as complete path planning, moving speed, and operation time.

[0072] The simulated execution effects of each scheme mainly include two percentage indicators, namely the water compliance rate and the fertilizer compliance rate, which can intuitively reflect the actual execution effects of each scheme. For example, the simulated execution effect of an excellent scheme shows that the final water compliance rate is 95% and the fertilizer compliance rate is 90%.

[0073] The operation score of each scheme is a comprehensive evaluation index from 1 to 10. The determination of the operation score is to quantitatively evaluate the scheme from three dimensions: energy consumption, time consumption, and coverage uniformity. The score comprehensively considers the execution efficiency (40% weight), energy consumption (30% weight), and operation coverage quality (30% weight) of the scheme, providing an objective basis for the selection of the optimal scheme. For example, a scheme with low energy consumption, short time consumption, and uniform coverage may obtain a score of 9.

[0074] By dividing the third generative adversarial network into three layers: the regional division layer, the scheme simulation layer, and the optimal scheme decision layer, each layer can focus on performing specific tasks. The regional division layer is responsible for dividing the field into multiple operation areas according to the spatial distribution characteristics of the watering points and fertilizing points, and calculating the water demand index and fertilizer demand index of each area, thus providing a data basis for subsequent processing. The scheme simulation layer generates multiple sets of differentiated motor control schemes based on the demand indexes of the divided areas, and evaluates the simulated execution effects and operation scores of each scheme, thus providing alternative schemes for decision-making. The optimal scheme decision layer comprehensively compares the execution effects and operation scores of each scheme, and then selects the optimal control scheme. This hierarchical design enables each layer to be optimized for its specific task. Through gradual refinement and progressive processing, it not only improves the processing efficiency of the system but also ensures the rationality and reliability of the control scheme.

[0075] Step S6, controlling the permanent magnet synchronous motor based on the target motor control scheme.

[0076] After determining the target motor control scheme, the permanent magnet synchronous motor is controlled according to the instructions and parameters set in the target motor control scheme to complete the watering and fertilizing of the field.

[0077] Based on the same inventive concept, Figure 4 FIG. is a schematic diagram of a permanent magnet synchronous motor control system based on image processing provided by an embodiment of the present invention. The permanent magnet synchronous motor control system based on image processing includes:

[0078] An acquisition module 41 for acquiring lidar data of a field;

[0079] An information determination module 42 for determining a plurality of initial watering point information and a plurality of initial fertilization point information based on the lidar data of the field;

[0080] An initial scheme generation module 43 for generating an initial motor control scheme based on the plurality of initial watering point information and the plurality of initial fertilization point information;

[0081] An initial control module 44 for controlling a permanent magnet synchronous motor to perform initial watering and fertilization based on the initial motor control scheme, and acquiring lidar data of the field after the initial watering and fertilization;

[0082] A target scheme generation module 45 for generating a target motor control scheme based on the lidar data of the field after the initial watering and fertilization;

[0083] A target control module 46 for controlling the permanent magnet synchronous motor based on the target motor control scheme.

[0084] Similarly, it should be noted that, in order to simplify the presentation disclosed in this specification and thus help the understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of this specification, sometimes multiple features are merged into one embodiment, drawing, or description thereof. However, this disclosure method does not mean that the features required by the subject matter of this specification are more than those mentioned in the claims. In fact, the features of the embodiments are less than all the features of the individual embodiments disclosed above.

[0085] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, by way of example and not limitation, alternative configurations of the embodiments of this specification may be regarded as consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to those explicitly introduced and described in this specification.

Claims

1. A permanent magnet synchronous motor control method based on image processing, characterized in that include: Obtain LiDAR data of the fields; determining a plurality of initial watering point information and a plurality of initial fertilizing point information based on the laser radar data of the field; generating an initial motor control scheme based on the plurality of initial watering point information and the plurality of initial fertilizing point information; Controlling the permanent magnet synchronous motor to perform initial watering and fertilizing based on the initial motor control scheme, and acquiring laser radar data of the field after the initial watering and fertilizing; Generating a target motor control scheme based on the laser radar data of the field after the initial watering and fertilization, wherein generating a target motor control scheme based on the laser radar data of the field after the initial watering and fertilization comprises: generating a target moisture content distribution map and a target fertilizer requirement map based on the lidar data of the field after the initial watering and fertilization; Constructing a knowledge graph, the knowledge graph including a plurality of initial watering point nodes and a plurality of initial fertilizing point nodes, the node features of the initial watering point nodes including initial watering point information and a target moisture content distribution map, the node features of the initial fertilizing point nodes including initial fertilizing point information and a target fertilizer demand map, the edges between different initial watering point nodes being the distances between different initial watering points, the edges between different initial fertilizing point nodes being the distances between different initial fertilizing points, and the edges between the initial watering point node and the initial fertilizing point node being the distances between the initial watering point and the initial fertilizing point; Processing the knowledge graph based on a graph neural network model to determine multiple target watering point information and multiple target fertilizing point information; Generate a target motor control scheme based on the multiple target watering point information and the multiple target fertilizing point information, the generating the target motor control scheme based on the multiple target watering point information and the multiple target fertilizing point information comprising: using a third generative adversarial network to generate a target motor control scheme based on the multiple target watering point information and the multiple target fertilizing point information, the third generative adversarial network comprising a region division layer, a scheme simulation layer, and an optimal scheme decision layer, the input of the region division layer being the multiple target watering point information and the multiple target fertilizing point information, the output of the region division layer being the multiple divided regions of the field, the water demand index of each region, and the fertilizer demand index of each region, the input of the scheme simulation layer being the multiple divided regions of the field, the water demand index of each region, and the fertilizer demand index of each region, the output of the scheme simulation layer being multiple sets of simulated motor control schemes, the simulated execution effect of each set of schemes, and the operation score of each set of schemes, the input of the optimal scheme decision layer being the multiple sets of simulated motor control schemes, the simulated execution effect of each set of schemes, and the operation score of each set of schemes, and the output of the optimal scheme decision layer being the target motor control scheme; The permanent magnet synchronous motor is controlled based on the target motor control scheme.

2. The permanent magnet synchronous motor control method based on image processing according to claim 1, characterized in that: The determining of a plurality of initial watering point information and a plurality of initial fertilizing point information based on the laser radar data of the field comprises: generating an initial moisture content distribution map and an initial fertilizer requirement map based on the LiDAR data of the field; Based on the initial water content distribution map and the initial fertilizer requirement map, multiple initial watering point information and multiple initial fertilizing point information are obtained.

3. The permanent magnet synchronous motor control method based on image processing according to claim 1, characterized in that The initial watering point information includes the initial watering point location and the initial watering amount.

4. A permanent magnet synchronous motor control system based on image processing, characterized in that: include: Acquisition module, used to obtain LiDAR data of the field; an information determination module, configured to determine a plurality of initial watering point information and a plurality of initial fertilization point information based on the laser radar data of the field; An initial scheme generating module, configured to generate an initial motor control scheme based on the plurality of initial watering point information and the plurality of initial fertilizing point information; an initial control module, configured to control the permanent magnet synchronous motor to perform initial watering and fertilizing based on the initial motor control scheme, and to obtain lidar data of the field after the initial watering and fertilizing; A target solution generation module is used to generate a target motor control solution based on the laser radar data of the field after the initial watering and fertilization, and the target solution generation module is further used to: generating a target moisture content distribution map and a target fertilizer requirement map based on the lidar data of the field after the initial watering and fertilization; Constructing a knowledge graph, the knowledge graph including a plurality of initial watering point nodes and a plurality of initial fertilizing point nodes, the node features of the initial watering point nodes including initial watering point information and a target moisture content distribution map, the node features of the initial fertilizing point nodes including initial fertilizing point information and a target fertilizer demand map, the edges between different initial watering point nodes being the distances between different initial watering points, the edges between different initial fertilizing point nodes being the distances between different initial fertilizing points, and the edges between the initial watering point node and the initial fertilizing point node being the distances between the initial watering point and the initial fertilizing point; Processing the knowledge graph based on a graph neural network model to determine multiple target watering point information and multiple target fertilizing point information; Generate a target motor control scheme based on the multiple target watering point information and the multiple target fertilizing point information, the generating the target motor control scheme based on the multiple target watering point information and the multiple target fertilizing point information comprising: using a third generative adversarial network to generate a target motor control scheme based on the multiple target watering point information and the multiple target fertilizing point information, the third generative adversarial network comprising a region division layer, a scheme simulation layer, and an optimal scheme decision layer, the input of the region division layer being the multiple target watering point information and the multiple target fertilizing point information, the output of the region division layer being the multiple divided regions of the field, the water demand index of each region, and the fertilizer demand index of each region, the input of the scheme simulation layer being the multiple divided regions of the field, the water demand index of each region, and the fertilizer demand index of each region, the output of the scheme simulation layer being multiple sets of simulated motor control schemes, the simulated execution effect of each set of schemes, and the operation score of each set of schemes, the input of the optimal scheme decision layer being the multiple sets of simulated motor control schemes, the simulated execution effect of each set of schemes, and the operation score of each set of schemes, and the output of the optimal scheme decision layer being the target motor control scheme; A target control module is used to control the permanent magnet synchronous motor based on the target motor control scheme.

5. The permanent magnet synchronous motor control system based on image processing according to claim 4, characterized in that: The information determination module is further configured to: generating an initial moisture content distribution map and an initial fertilizer requirement map based on the LiDAR data of the field; Based on the initial water content distribution map and the initial fertilizer requirement map, multiple initial watering point information and multiple initial fertilizing point information are obtained.

6. The permanent magnet synchronous motor control system based on image processing according to claim 4, characterized in that, The initial watering point information includes the initial watering point location and the initial watering amount.

Images