Intelligent agricultural equipment and method for improving soil

Abstract

The invention relates to intelligent agricultural equipment and a method for improving soil. The method comprises the steps that environment temperature data are monitored and collected in real time through a temperature sensor, soil surface state images are collected through a camera, and the environment temperature data and the soil surface state images are analyzed through a data processing and image understanding algorithm based on artificial intelligence and deep learning at the rear end; therefore, multi-modal essential fusion features between the soil surface state semantics and the environment temperature time sequence semantics are captured and described, and the spraying amount is recommended and controlled. Therefore, the spraying amount of the modifier can be dynamically adjusted according to the specific condition of the soil and the change of the environment temperature, so that the intelligent degree of control over the agricultural equipment for improving the soil is improved, and better soil improvement quality and efficiency can be achieved.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent agricultural equipment, and in particular, to an intelligent agricultural equipment and method for improving soil. Background Art

[0002] Soil improvement refers to the use of theories and technologies from multiple disciplines such as soil science, biology, and ecology to eliminate or prevent adverse factors that affect crop growth and cause soil degradation, improve soil properties, and increase soil fertility, thereby creating good soil environmental conditions for crops. However, in the process of improving soil, traditional agricultural equipment often mixes soil conditioners and water and applies them directly to the surface of the soil, which needs to slowly penetrate into the ground for absorption. In this process, the conditioners and water are easy to evaporate, and the utilization rate of raw materials is low, which affects the improvement effect of the soil. In addition, traditional agricultural equipment requires people to carry the box on their backs and hold the nozzle for spraying, which is time-consuming and labor-intensive, and has low operating efficiency.

[0003] In response to the above technical problems, Chinese patent CN117882525A proposes an agricultural equipment for improving soil in agriculture, which includes a soil turning mechanism, a soil shoveling mechanism, a transportation mechanism and a crushing mechanism. Specifically, the soil is loosened by the soil turning mechanism, and the soil is shoveled by the soil shoveling mechanism. After the soil is shoveled, the improver is sprayed on the land, and then organic fertilizer is added to the shoveled soil for mixing. Among them, the crushing mechanism can crush the stones or large pieces of soil in the land, making the soil fragmented smaller, which is conducive to making the soil softer after improvement; the transportation mechanism is responsible for transporting the shoveled soil to the next process, and at the same time, the vibration mechanism helps to loosen the soil.

[0004] In the above-mentioned agricultural equipment used for agricultural soil improvement, although a series of improvement work can be completed by turning over, shoveling, transporting and crushing the soil, the equipment lacks a monitoring and response mechanism for environmental variables and the actual state of the soil, and usually relies on manual experience or fixed rules to control the spraying amount of the improver. This means that under different conditions, the application amount of the improver cannot be optimally configured, which will not only lead to low utilization rate of the improver and insufficient improvement, but also may cause environmental pollution due to excessive use, limiting the application of agricultural equipment in diversified agricultural environments.

[0005] Therefore, an optimized agricultural device for improving soil is desired. Summary of the invention

[0006] This summary is provided to introduce concepts in a brief form that will be described in detail in the detailed description below. This summary is not intended to identify key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0007] In a first aspect, the present application provides an intelligent agricultural device for improving soil, the device comprising:

[0008] A vehicle body, a soil turning mechanism, a soil shoveling mechanism, a transport mechanism, a crushing mechanism and an improver spraying mechanism arranged on the vehicle body; wherein the soil turning mechanism is used to loosen the soil to form loose soil and large soil blocks; the soil shoveling mechanism is used to shovel up the loose soil and the large soil blocks, and spray the improver into the loose soil and the large soil blocks through the improver spraying mechanism; the transport mechanism is used to transport the loose soil and the large soil blocks to the crushing mechanism; the crushing mechanism is used to crush the large soil blocks to obtain improved soil;

[0009] The improver spraying mechanism comprises:

[0010] An ambient temperature acquisition module is used to obtain a time queue of the ambient temperature acquired by a temperature sensor;

[0011] A soil surface state image acquisition module is used to acquire a soil surface state image acquired by a camera;

[0012] A soil surface local area state feature extraction module, used for inputting the soil surface state image into a soil surface state feature extractor based on a ViT model to obtain a sequence of soil surface local area state semantic feature vectors;

[0013] A soil surface local area state semantic selection optimization aggregation module is used to perform feature sequence selection optimization aggregation processing based on a feature energy distribution spectrum on the sequence of the soil surface local area state semantic feature vectors to obtain a soil surface state global semantic enhancement feature vector;

[0014] An ambient temperature time series encoding module, used for inputting the time queue of the ambient temperature into an ambient temperature sequence encoder based on an RNN network to obtain an ambient temperature time series associated implicit feature vector;

[0015] An ambient temperature-soil state multimodal essential fusion module is used to input the soil surface state global semantic enhancement feature vector and the ambient temperature time series associated implicit feature vector into a significant fusion network based on feature principal component query matching to obtain an ambient temperature-soil state multimodal essential fusion representation vector;

[0016] The spraying control module is used to input the multi-modal essential fusion representation vector of the ambient temperature-soil state into a spraying intelligent controller based on a decoder to obtain a control instruction, wherein the control instruction is used to represent a decoded value of a recommended spraying amount.

[0017] Optionally, the soil surface local area state semantic selection optimization aggregation module includes: a soil surface local state feature selection optimization unit, which is used to input the sequence of soil surface local area state semantic feature vectors into a feature sequence selection optimization module based on a feature energy distribution spectrum to obtain a sequence of selected soil surface local area state semantic feature vectors; a soil surface state global semantic feature expression unit, which is used to cascade the sequence of selected soil surface local area state semantic feature vectors to obtain the soil surface state global semantic enhancement feature vector.

[0018] Optionally, the soil surface local state feature selection and optimization unit includes: a characteristic energy distribution spectrum vector calculation subunit, which is used to calculate the characteristic energy distribution spectrum vector of each soil surface local area state semantic feature vector in the sequence of the soil surface local area state semantic feature vector to obtain a sequence of soil surface local area state semantic feature energy distribution spectrum vectors; a position mean vector calculation subunit, which is used to calculate the position mean vector of the sequence of the soil surface local area state semantic feature energy distribution spectrum vector as the soil surface local area state semantic energy spectrum distribution center vector; an energy distribution spectrum span factor calculation subunit, which is used to calculate the energy distribution spectrum span factor between the soil surface local area state semantic energy spectrum distribution center vector and each soil surface local area state semantic feature energy distribution spectrum vector in the sequence of the soil surface local area state semantic feature energy distribution spectrum vector to obtain a sequence of soil surface local area state semantic energy distribution spectrum span factors; a span factor and preset threshold comparison subunit, which is used to determine the selected soil based on the comparison between each soil surface local area state semantic energy distribution spectrum span factor in the sequence of the soil surface local area state semantic energy distribution spectrum span factor and a preset threshold. The energy distribution spectrum span factor calculation subunit comprises: a secondary subunit for calculating position difference, which is used to calculate the position difference between the energy distribution spectrum vector of the local area state semantic feature of the soil surface and the central vector of the energy spectrum distribution of the local area state semantic feature of the soil surface to obtain the semantic differential feature vector of the local area state of the soil surface; a secondary subunit for calculating the covariance matrix, which is used to calculate the covariance matrix between the central vector of the energy spectrum distribution of the local area state semantic feature of the soil surface and the energy distribution spectrum vector of the local area state semantic feature of the soil surface, and calculate The inverse of the covariance matrix is ​​calculated to obtain the state semantic inverse covariance matrix of the local area of ​​the soil surface; the state semantic distribution value calculation secondary subunit is used to calculate the transposed vector of the local area state semantic differential feature vector of the soil surface and the product of the local area state semantic inverse covariance matrix of the soil surface and the local area state semantic differential feature vector of the soil surface to obtain the state semantic distribution value of the local area of ​​the soil surface; the semantic energy distribution spectrum span factor calculation secondary subunit is used to calculate the square root of the local area state semantic distribution value of the soil surface to obtain the state semantic energy distribution spectrum span factor of the local area of ​​the soil surface.

[0019] Optionally, the feature energy distribution spectrum vector calculation subunit is used for: a feature distribution energy synergistic representation vector calculation secondary subunit, which is used to respectively calculate the feature distribution energy synergistic representation vector between each soil surface local area state semantic feature vector and other soil surface local area state semantic feature vectors in the sequence of soil surface local area state semantic feature vectors to obtain a sequence of multiple soil surface local area state semantic feature distribution energy synergistic representation vectors; a feature distribution energy synergistic factor vector calculation secondary subunit, which is used to calculate the feature distribution energy synergistic factor vector of the sequence of each soil surface local area state semantic feature distribution energy synergistic representation vector in the sequence of multiple soil surface local area state semantic feature distribution energy synergistic representation vectors to obtain the sequence of soil surface local area state semantic feature energy distribution spectrum vectors.

[0020] Optionally, the feature distribution energy collaborative representation vector calculation secondary subunit is used to: extract a predetermined soil surface local area state semantic feature vector from a sequence of feature energy distribution spectrum vectors of the soil surface local area state semantic feature vector; perform element-by-element maximum value extraction on the predetermined soil surface local area state semantic feature vector and each other soil surface local area state semantic feature vector, so as to obtain a sequence of soil surface local area state semantic feature distribution energy collaborative representation vectors corresponding to the predetermined soil surface local area state semantic feature vector.

[0021] Optionally, the feature distribution energy synergy factor vector calculation secondary subunit is used to: respectively extract the maximum value of each soil surface local area state semantic feature distribution energy synergy representation vector in the sequence of the soil surface local area state semantic feature distribution energy synergy representation vector to obtain a sequence of soil surface local area state semantic feature maximum values; respectively calculate the mean and variance of each soil surface local area state semantic feature distribution energy synergy representation vector to obtain a sequence of soil surface local area state semantic feature means and a sequence of soil surface local area state semantic feature variances; positionally add the sequence of soil surface local area state semantic feature variances and preset hyperparameters to obtain a sequence of soil surface local area state semantic feature first energy distribution coefficients; calculate the sequence of soil surface local area state semantic feature maximum values ​​and the sequence of soil surface local area state semantic feature mean values, and the sequence of soil surface local area state semantic feature maximum values ​​and soil surface local area state semantic feature variances for each group. The method comprises the following steps: obtaining a sequence of soil surface local area state semantic difference values ​​by square of the difference between the mean values ​​of the semantic features of the local area state of the soil surface; multiplying the sequence of soil surface local area state semantic feature variances by constant two by position to obtain a sequence of modulated soil surface local area state semantic feature variances; adding each group of corresponding soil surface local area state semantic difference values ​​and modulated soil surface local area state semantic feature variances in the sequence of soil surface local area state semantic difference values ​​and the sequence of modulated soil surface local area state semantic feature variances, and then adding the sum to the preset hyperparameter to obtain a sequence of second energy distribution coefficients of soil surface local area state semantic features; dividing the sequence of first energy distribution coefficients of soil surface local area state semantic features by the sequence of second energy distribution coefficients of soil surface local area state semantic features by position to obtain an energy distribution spectrum vector of soil surface local area state semantic features composed of multiple feature distribution energy synergy factors.

[0022] Optionally, the ambient temperature-soil state multimodal essential fusion module includes: a standardization processing unit, which is used to standardize the global semantic enhancement feature vector of the soil surface state and the ambient temperature time series associated implicit feature vector to obtain a standardized global semantic enhancement feature vector of the soil surface state and a standardized ambient temperature time series associated implicit feature vector; a sample covariance matrix calculation unit, which is used to calculate the sample covariance matrix of the standardized global semantic enhancement feature vector of the soil surface state and the standardized ambient temperature time series associated implicit feature vector to obtain a soil surface state sample covariance matrix and an ambient temperature time series sample covariance matrix; a feature vector extraction calculation unit, which is used to extract the sample covariance matrix of the soil surface state; The matrix and the ambient temperature time series sample covariance matrix are used to perform feature vector extraction based on matrix decomposition to obtain a set of soil surface state principal component feature vectors and a set of ambient temperature time series principal component feature vectors; a maximum approximate query matching unit is used to input the set of soil surface state principal component feature vectors and the set of ambient temperature time series principal component feature vectors into the maximum approximate query matching network to obtain a set of best matching pairs of soil surface state principal component feature vectors and ambient temperature time series principal component feature vectors; a semantic fine-grained gating combination unit is used to combine each soil surface state principal component feature vector and ambient temperature in the set of best matching pairs of soil surface state principal component feature vectors and ambient temperature time series principal component feature vectors. The best matching pair of the time series principal component feature vectors is input into the semantic fine-grained gating joint module to obtain a set of ambient temperature-soil state principal component fusion feature vectors; a vector cascade unit is used to cascade each ambient temperature-soil state principal component fusion feature vector in the set of ambient temperature-soil state principal component fusion feature vectors to obtain the ambient temperature-soil state multimodal essential fusion representation vector; wherein the semantic fine-grained gating joint unit is used to: perform position difference, position dot multiplication and position sum processing on the best matching pairs of each soil surface state principal component feature vector and the ambient temperature time series principal component feature vector respectively to obtain a set of ambient temperature-soil state principal component difference feature vectors, an ambient temperature The method comprises the following steps: obtaining a set of principal component dot product feature vectors of the ambient temperature-soil state principal component and a set of principal component sum feature vectors of the ambient temperature-soil state principal component; cascading the set of principal component difference feature vectors of the ambient temperature-soil state principal component, the set of principal component dot product feature vectors of the ambient temperature-soil state principal component and the set of principal component sum feature vectors of the ambient temperature-soil state principal component by position to obtain a set of principal component multidimensional fusion expression vectors of the ambient temperature-soil state principal component; processing each principal component multidimensional fusion expression vector of the ambient temperature-soil state principal component in the set of principal component multidimensional fusion expression vectors by a one-dimensional convolution layer and then by a maximum pooling layer to obtain a set of principal component fusion feature vectors of the ambient temperature-soil state principal component.

[0023] Optionally, the standardization processing unit is used to: calculate the mean and standard deviation of the global semantic enhancement feature vector of the soil surface state to obtain the mean and standard deviation of the soil surface state; calculate the positional difference between the global semantic enhancement feature vector of the soil surface state and the mean of the soil surface state, and then divide the difference by position with the standard deviation of the soil surface state to obtain the standardized global semantic enhancement feature vector of the soil surface state; calculate the mean and standard deviation of the ambient temperature time series associated implicit feature vector to obtain the ambient temperature time series mean and standard deviation of the ambient temperature time series; calculate the positional difference between the ambient temperature time series associated implicit feature vector and the ambient temperature time series mean, and then divide the difference by position with the standard deviation of the ambient temperature time series to obtain the standardized ambient temperature time series associated implicit feature vector.

[0024] Optionally, the maximum approximate query matching unit is used to: take the soil surface state principal component feature vector at a predetermined position in the set of soil surface state principal component feature vectors to obtain a predetermined soil surface state principal component feature vector; calculate the inner product between the predetermined soil surface state principal component feature vector and each ambient temperature time series principal component feature vector in the set of ambient temperature time series principal component feature vectors to obtain a sequence of ambient temperature-soil state inner product values; calculate a norm of the predetermined soil surface state principal component feature vector to obtain a norm of the predetermined soil surface state principal component feature; calculate the ambient temperature time series principal component feature vector The one-norm of each ambient temperature time series principal component eigenvector in the set is used to obtain a sequence of the one-norm of the ambient temperature time series principal component characteristics; the one-norm of the predetermined soil surface state principal component characteristics and the sequence of the one-norm of the ambient temperature time series principal component characteristics are multiplied by the position point to obtain a sequence of ambient temperature-soil state semantic interaction values; after calculating the point-by-position division between the sequence of ambient temperature-soil state inner product values ​​and the sequence of ambient temperature-soil state semantic interaction values, the ambient temperature time series principal component eigenvector corresponding to the maximum eigenvalue in the sequence of eigenvalues ​​obtained and the predetermined soil surface state principal component eigenvector form a set of best matching pairs.

[0025] In a second aspect, the present application provides a smart agricultural method for improving soil, the method comprising:

[0026] Using a soil turning mechanism to loosen the soil to form loose soil and bulk soil; using a shoveling mechanism to scoop up the loose soil and bulk soil, and spraying an improver into the loose soil and bulk soil through an improver spraying mechanism; using a transport mechanism to transport the loose soil and bulk soil to a crushing mechanism; using the crushing mechanism to crush the bulk soil to obtain improved soil;

[0027] The method comprises spraying the improver into the loose soil and the bulk soil by an improver spraying mechanism, comprising:

[0028] Get the time queue of the ambient temperature collected by the temperature sensor;

[0029] Acquire a soil surface state image collected by a camera;

[0030] Inputting the soil surface state image into a soil surface state feature extractor based on the ViT model to obtain a sequence of semantic feature vectors of local area states of the soil surface;

[0031] Performing feature sequence selection optimization aggregation processing based on feature energy distribution spectrum on the sequence of semantic feature vectors of local area state of soil surface to obtain a global semantic enhancement feature vector of soil surface state;

[0032] Inputting the time queue of the ambient temperature into an ambient temperature sequence encoder based on an RNN network to obtain an ambient temperature time series associated implicit feature vector;

[0033] Inputting the soil surface state global semantic enhancement feature vector and the ambient temperature time series associated implicit feature vector into a significant fusion network based on feature principal component query matching to obtain an ambient temperature-soil state multimodal essential fusion representation vector;

[0034] The multi-modal essential fusion representation vector of the ambient temperature and soil state is input into a spraying intelligent controller based on a decoder to obtain a control instruction, wherein the control instruction is used to represent a decoded value of a recommended spraying amount.

[0035] The above technical solution is adopted to collect ambient temperature data in real time through temperature sensors, and collect soil surface state images through cameras. The data processing and image understanding algorithms based on artificial intelligence and deep learning are used at the back end to analyze these ambient temperature data and soil surface state images, so as to capture and characterize the multimodal essential fusion characteristics between the semantics of soil surface state and the time series semantics of ambient temperature, so as to recommend and control the spraying amount. In this way, the spraying amount of the improver can be dynamically adjusted according to the specific conditions of the soil and the changes in ambient temperature, thereby improving the intelligence level of soil improvement agricultural equipment control and helping to achieve better soil improvement quality and efficiency.

[0036] Other features and advantages of the present application will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The above and other features, advantages and aspects of the embodiments of the present application will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and the originals and elements are not necessarily drawn to scale. In the drawings:

[0038] Figure 1 It is a schematic diagram of the structure of an intelligent agricultural device for improving soil according to an exemplary embodiment.

[0039] Figure 2 It is a block diagram of the improver spraying mechanism of an intelligent agricultural equipment for improving soil according to an exemplary embodiment.

[0040] Figure 3 The invention is a flow chart of a smart agricultural method for improving soil according to an exemplary embodiment.

[0041] Figure 4 It is a block diagram of an electronic device according to an exemplary embodiment.

[0042] Figure 5 This is a diagram of an application scenario of a smart agricultural device for improving soil according to an exemplary embodiment. DETAILED DESCRIPTION

[0043] The embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be construed as being limited to the embodiments described herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present application. It should be understood that the drawings and embodiments of the present application are only for exemplary purposes and are not intended to limit the scope of protection of the present application.

[0044] It should be understood that the various steps described in the method implementation of the present application can be performed in different orders and / or performed in parallel. In addition, the method implementation may include additional steps and / or omit the steps shown. The scope of the present application is not limited in this respect.

[0045] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0046] It should be noted that the concepts such as "first" and "second" mentioned in this application are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0047] It should be noted that the modifications of "one" and "plurality" mentioned in the present application are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0048] The names of the messages or information exchanged between multiple devices in the embodiments of the present application are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0049] Accordingly, considering that the existing method of spraying the improver does not take into account the actual state of the soil and the influence of the surrounding environment, such as temperature changes may cause changes in the activity or volatility of the improver, thereby affecting the improvement effect, and the actual state of the soil also requires different spraying amounts of improvers. This method that relies on manual operation or a fixed spraying amount is likely to result in the improver not being effectively utilized, or the improvement effect is reduced due to inappropriate dosage.

[0050] Based on this, the technical concept of this application is to monitor and collect ambient temperature data in real time through temperature sensors during the operation of the improver spraying mechanism, and collect soil surface state images through cameras, and use data processing and image understanding algorithms based on artificial intelligence and deep learning to analyze these ambient temperature data and soil surface state images at the back end, so as to capture and characterize the multimodal essential fusion characteristics between the soil surface state semantics and the ambient temperature time series semantics, so as to recommend and control the spraying amount. In this way, the spraying amount of the improver can be dynamically adjusted according to the specific conditions of the soil and the changes in ambient temperature, thereby improving the intelligence level of soil improvement agricultural equipment control and helping to achieve better soil improvement quality and efficiency.

[0051] The specific implementation of the present application is described in detail below with reference to the accompanying drawings.

[0052] Figure 1 is a schematic diagram of a structure of an intelligent agricultural device for improving soil according to an exemplary embodiment. Figure 1 As shown, Figure 1 As shown, the smart agricultural equipment includes:

[0053] A vehicle body 1, a soil turning mechanism 2, a soil shoveling mechanism 3, a transportation mechanism 4, a crushing mechanism 5 and an improver spraying mechanism 6 are arranged on the vehicle body 1; wherein the soil turning mechanism 2 is used to loosen the soil to form loose soil and large soil blocks; the soil shoveling mechanism 3 is used to shovel up the loose soil and the large soil blocks, and spray the improver into the loose soil and the large soil blocks through the improver spraying mechanism 6; the transportation mechanism 4 is used to transport the loose soil and the large soil blocks to the crushing mechanism 5; the crushing mechanism 5 is used to crush the large soil blocks to obtain improved soil.

[0054] Figure 2 is a block diagram of the improver spraying mechanism of an intelligent agricultural device for improving soil according to an exemplary embodiment, such as Figure 2 As shown, the improver spraying mechanism 6 comprises:

[0055] The ambient temperature acquisition module 101 is used to acquire the time queue of the ambient temperature acquired by the temperature sensor;

[0056] The soil surface state image acquisition module 102 is used to acquire the soil surface state image acquired by the camera;

[0057] The soil surface local area state feature extraction module 103 is used to input the soil surface state image into a soil surface state feature extractor based on the ViT model to obtain a sequence of soil surface local area state semantic feature vectors;

[0058] The soil surface local area state semantic selection optimization aggregation module 104 is used to perform feature sequence selection optimization aggregation processing based on the feature energy distribution spectrum on the sequence of the soil surface local area state semantic feature vectors to obtain the soil surface state global semantic enhancement feature vector;

[0059] The ambient temperature time series encoding module 105 is used to input the time queue of the ambient temperature into an ambient temperature sequence encoder based on an RNN network to obtain an ambient temperature time series associated implicit feature vector;

[0060] The environment temperature-soil state multimodal essential fusion module 106 is used to input the soil surface state global semantic enhancement feature vector and the environment temperature time series associated implicit feature vector into a significant fusion network based on feature principal component query matching to obtain an environment temperature-soil state multimodal essential fusion representation vector;

[0061] The spraying control module 107 is used to input the ambient temperature-soil state multimodal essential fusion representation vector into the decoder-based spraying intelligent controller to obtain a control instruction, and the control instruction is used to represent the decoded value of the recommended spraying amount.

[0062] Specifically, in the technical solution of the present application, first, the time queue of the ambient temperature collected by the temperature sensor is obtained, and the soil surface state image collected by the camera is obtained. Then, in order to effectively extract the semantic features of the soil surface state from the soil surface state image, realize the accurate analysis and understanding of the soil state, and thus help to intelligently control the spraying amount of the modifier, in the technical solution of the present application, the soil surface state image is further input into the soil surface state feature extractor based on the ViT model to obtain a sequence of semantic feature vectors of the local area state of the soil surface. It should be understood that the ViT model is a visual recognition model based on the Transformer architecture, which can divide the image into a series of local image blocks, and then regard these local image blocks as elements in the sequence, and capture the semantic association relationship between these local image blocks through the self-attention mechanism, so as to better understand the content of the soil surface state image. Therefore, through the processing of the soil surface state feature extractor based on the ViT model, the color, texture, structure and other local area state features of the soil in each local image block can be effectively identified, and information that can reflect the physical and chemical properties of the soil can be extracted. And learn and capture the implicit associations between the semantics of these local image blocks, providing a basis for deciding the amount of the improver to be sprayed.

[0063] In one embodiment of the present application, the soil surface local area state semantic selection optimization aggregation module includes: a soil surface local state feature selection optimization unit, which is used to input the sequence of the soil surface local area state semantic feature vectors into a feature sequence selection optimization module based on a feature energy distribution spectrum to obtain a sequence of selected soil surface local area state semantic feature vectors; a soil surface state global semantic feature expression unit, which is used to cascade the sequence of selected soil surface local area state semantic feature vectors to obtain the soil surface state global semantic enhancement feature vector.

[0064] Furthermore, considering that in the actual process of feature analysis and semantic association of the semantics of each local image block in the soil surface state image, different local image block semantics have different importance and contribution to the subsequent decision-making task of the amount of modifier sprayed, that is, not all semantic features of the local area of ​​the soil surface are equally important for the subsequent tasks, and some areas in the image contain a large amount of redundant information and interference information. Therefore, in order to feature select the image block semantic features with higher relevance to the subsequent tasks from the sequence of the semantic feature vectors of the local area state of the soil surface, and enhance these features to ensure that the selected features are more representative and more in line with the actual situation of the current soil state, in the technical solution of the present application, the sequence of the semantic feature vectors of the local area state of the soil surface is further input into the feature sequence selection optimization module based on the feature energy distribution spectrum to obtain the sequence of the selected semantic feature vectors of the local area state of the soil surface. The feature sequence selection optimization module based on the feature energy distribution spectrum is a feature selection technology based on energy distribution spectrum analysis, which aims to optimize the expression of the feature vector sequence by quantifying the energy distribution of the semantic features of each local image block. Specifically, the feature sequence selection optimization module based on the characteristic energy distribution spectrum can evaluate the energy distribution characteristics of each soil surface local area state semantic feature vector by calculating the characteristic energy distribution spectrum, which helps to determine which soil surface local area state semantic feature vectors carry more valuable information. The determination of the center vector of the energy spectrum distribution provides a reference point for the selection and optimization of feature vectors, which can be used to measure the importance of each soil surface local area state semantic feature vector. The calculation of the energy distribution spectrum span factor further reveals the distribution between feature vectors, thereby helping to understand the differences and importance between different soil surface local area state semantic features. The final feature selection process is a screening process that ensures that only those key features that best characterize the soil surface state are retained. Through this series of operations, a sequence of selected soil surface local area state semantic feature vectors can be obtained. Each feature vector in the sequence focuses more on the expression of important local area semantic features of the soil surface state, removes redundant information, and improves the quality of soil surface state feature expression. The advantage of this is that the subsequent model can make decisions about soil improvement strategies more accurately based on the optimized selected feature vectors, such as determining the amount of modifier to be sprayed, thereby achieving a more accurate and efficient soil improvement effect.

[0065] In one embodiment of the present application, the soil surface local state feature selection optimization unit includes: a characteristic energy distribution spectrum vector calculation subunit, which is used to calculate the characteristic energy distribution spectrum vector of each soil surface local area state semantic feature vector in the sequence of the soil surface local area state semantic feature vector to obtain a sequence of soil surface local area state semantic feature energy distribution spectrum vectors; a position mean vector calculation subunit, which is used to calculate the position mean vector of the sequence of the soil surface local area state semantic feature energy distribution spectrum vector as the soil surface local area state semantic energy spectrum distribution center vector; an energy distribution spectrum span factor calculation subunit, which is used to calculate the energy distribution spectrum span factor between the soil surface local area state semantic energy spectrum distribution center vector and each soil surface local area state semantic feature energy distribution spectrum vector in the sequence of the soil surface local area state semantic feature energy distribution spectrum vector to obtain a sequence of soil surface local area state semantic energy distribution spectrum span factors; a span factor and preset threshold comparison subunit, which is used to determine the selection based on the comparison between each soil surface local area state semantic energy distribution spectrum span factor in the sequence of the soil surface local area state semantic energy distribution spectrum span factor and a preset threshold. The sequence of selected semantic feature vectors of the local area state of the soil surface; wherein the energy distribution spectrum span factor calculation subunit includes: a secondary subunit for calculating positional difference, which is used to calculate the positional difference between the semantic feature energy distribution spectrum vector of the local area state of the soil surface and the central vector of the semantic energy spectrum distribution of the local area state of the soil surface to obtain the semantic differential feature vector of the local area state of the soil surface; a secondary subunit for calculating the covariance matrix, which is used to calculate the covariance matrix between the central vector of the semantic energy spectrum distribution of the local area state of the soil surface and the semantic feature energy distribution spectrum vector of the local area state of the soil surface. And calculate the inverse of the covariance matrix to obtain the state semantic inverse covariance matrix of the local area of ​​the soil surface; the state semantic distribution value calculation secondary subunit is used to calculate the transposed vector of the local area state semantic differential feature vector of the soil surface and the product of the local area state semantic inverse covariance matrix of the soil surface and the local area state semantic differential feature vector of the soil surface to obtain the state semantic distribution value of the local area of ​​the soil surface; the semantic energy distribution spectrum span factor calculation secondary subunit is used to calculate the square root of the local area state semantic distribution value of the soil surface to obtain the state semantic energy distribution spectrum span factor of the local area of ​​the soil surface.

[0066] Furthermore, in one embodiment of the present application, the feature energy distribution spectrum vector calculation subunit is used for: a feature distribution energy synergistic representation vector calculation secondary subunit, which is used to respectively calculate the feature distribution energy synergistic representation vector between each of the soil surface local area state semantic feature vectors and other soil surface local area state semantic feature vectors in the sequence of the soil surface local area state semantic feature vectors to obtain a sequence of multiple soil surface local area state semantic feature distribution energy synergistic representation vectors; a feature distribution energy synergistic factor vector calculation secondary subunit, which is used to calculate the feature distribution energy synergistic factor vector of the sequence of each soil surface local area state semantic feature distribution energy synergistic representation vector in the sequence of the multiple soil surface local area state semantic feature distribution energy synergistic representation vectors to obtain the sequence of the soil surface local area state semantic feature energy distribution spectrum vectors.

[0067] Furthermore, in one embodiment of the present application, the feature distribution energy collaborative representation vector calculation secondary subunit is used to: extract a predetermined soil surface local area state semantic feature vector from a sequence of feature energy distribution spectrum vectors of the soil surface local area state semantic feature vector; perform element-by-element maximum value extraction on the predetermined soil surface local area state semantic feature vector and each other soil surface local area state semantic feature vector, so as to obtain a sequence of soil surface local area state semantic feature distribution energy collaborative representation vectors corresponding to the predetermined soil surface local area state semantic feature vector.

[0068] Furthermore, in one embodiment of the present application, the feature distribution energy synergy factor vector calculation secondary subunit is used to: respectively extract the maximum value of each soil surface local area state semantic feature distribution energy synergy representation vector in the sequence of the soil surface local area state semantic feature distribution energy synergy representation vector to obtain a sequence of soil surface local area state semantic feature maximum values; respectively calculate the mean and variance of each soil surface local area state semantic feature distribution energy synergy representation vector to obtain a sequence of soil surface local area state semantic feature means and a sequence of soil surface local area state semantic feature variances; add the sequence of soil surface local area state semantic feature variances and preset hyperparameters by position to obtain a sequence of soil surface local area state semantic feature first energy distribution coefficients; calculate the maximum value of each group of soil surface local area state semantic feature corresponding to the sequence of the soil surface local area state semantic feature maximum values ​​and the sequence of the soil surface local area state semantic feature mean values. The square of the difference between the value and the mean value of the semantic feature of the local area state of the soil surface is obtained to obtain a sequence of semantic difference values ​​of the local area state of the soil surface; the sequence of variances of the semantic features of the local area state of the soil surface is multiplied by constant two by position to obtain a sequence of variances of the modulated semantic features of the local area state of the soil surface; after adding each group of corresponding semantic difference values ​​of the local area state of the soil surface and the variance of the modulated semantic features of the local area state of the soil surface, the sum is added to the preset hyperparameter to obtain a sequence of second energy distribution coefficients of the semantic features of the local area state of the soil surface; the sequence of the first energy distribution coefficients of the semantic features of the local area state of the soil surface is divided by the sequence of the second energy distribution coefficients of the semantic features of the local area state of the soil surface by position to obtain the energy distribution spectrum vector of the semantic features of the local area state of the soil surface composed of multiple feature distribution energy synergy factors.

[0069] Specifically, the sequence of the semantic feature vectors of the local area state of the soil surface is input into the feature sequence selection optimization module based on the characteristic energy distribution spectrum for processing using the following feature selection optimization formula to obtain the sequence of the semantic feature vectors of the local area state of the soil surface after the selection;

[0070] Wherein, the feature selection optimization formula is:

[0071] V co,ij =max(V i ,V j )

[0072]

[0073] P i={e i1 ,e i2 ,...,e ij ,...,e in}

[0074] P={P 1 ,P 2 ,...,P i ,...,P N}

[0075]

[0076] K={f i |Dist(c,P i )≤θ}

[0077] Among them, V i and V j denote the i-th and j-th soil surface local area state semantic feature vectors in the sequence of the soil surface local area state semantic feature vectors, max(·,·) denotes taking the maximum value between the two vectors element by element, V co,ij is the soil surface local area state semantic feature distribution energy synergistic representation vector between the i-th and j-th soil surface local area state semantic feature vectors, max(·) is the maximum value of the extracted vector, μ ij and σ ij 2 V co,ij The mean and variance of , λ is the preset hyperparameter, e ij is the jth characteristic distribution energy synergy factor in the i-th soil surface local area state semantic characteristic energy distribution spectrum vector in the sequence of the soil surface local area state semantic characteristic energy distribution spectrum vector, e in is the nth characteristic distribution energy synergy factor in the i-th soil surface local area state semantic feature energy distribution spectrum vector, P represents the sequence of the soil surface local area state semantic feature energy distribution spectrum vector, P i is the i-th soil surface local area state semantic feature energy distribution spectrum vector in the sequence of soil surface local area state semantic feature energy distribution spectrum vectors, N is the number of vectors in the sequence of soil surface local area state semantic feature energy distribution spectrum vectors, c is the soil surface local area state semantic energy spectrum distribution center vector, (·) T is the transpose of the vector, S i is the covariance matrix between the i-th soil surface local area state semantic feature energy distribution spectrum vector and the soil surface local area state semantic energy spectrum distribution center vector, Dist(c,P i) is the span factor of the soil surface local area state semantic energy distribution spectrum corresponding to the i-th soil surface local area state semantic feature energy distribution spectrum vector, θ is the preset threshold, f i is the i-th selected soil surface local area state semantic feature vector in the sequence of selected soil surface local area state semantic feature vectors, and K is the sequence of selected soil surface local area state semantic feature vectors.

[0078] Then, the selected sequence of the semantic feature vectors of the local area state of the soil surface is cascaded to obtain the global semantic enhanced feature vector of the soil surface state. The semantic feature vector of the local area state of the soil surface optimized by cascade feature selection can better reflect the overall condition of the entire soil surface, so that agricultural equipment can formulate a more reasonable improvement strategy according to the actual situation of the entire soil surface.

[0079] Then, the time queue of the ambient temperature is input into an ambient temperature sequence encoder based on a RNN network for encoding, so as to extract the temporal implicit correlation feature information of the ambient temperature in the time dimension, thereby obtaining an ambient temperature temporal correlation implicit feature vector.

[0080] It should be understood that the global semantic enhancement feature vector of the soil surface state contains the global semantic selection expression enhancement features related to the soil surface state, such as texture, humidity, nutrients, etc., and the time-series associated implicit feature vector of the ambient temperature contains the time-series associated implicit features of the ambient temperature, reflecting the trend and pattern of ambient temperature changes. These two feature vectors each carry important information about the health of the soil and the external environmental conditions. Therefore, in order to combine the data of the two different modes of soil state and ambient temperature, extract the important associated information between the two, and thus form a more representative and explanatory semantic representation to make decisions and control the spraying amount, in the technical solution of the present application, the global semantic enhancement feature vector of the soil surface state and the time-series associated implicit feature vector of the ambient temperature are further input into the significant fusion network based on feature principal component query matching to obtain the multi-modal essential fusion representation vector of ambient temperature-soil state. By inputting these two feature vectors into the significant fusion network based on feature principal component query matching, the network's feature sparse processing, principal component analysis and significant fusion capabilities can be used to reveal the deep connection between soil state and ambient temperature. Specifically, the significant fusion network based on feature principal component query matching can remove the dimensional influence of the feature vectors of the global semantics of the soil surface state and the time series semantics of the ambient temperature through standardization, and can capture the correlation between the two features through the calculation of the covariance matrix. The dimension of the feature space is reduced through principal component analysis and the most critical information in the soil surface state and the ambient temperature time series is retained. The matching network further strengthens the significant association between features, and the gating mechanism helps to screen out those feature combinations that have the greatest impact on the final goal. After such processing, the generated ambient temperature-soil state multimodal essential fusion representation vector contains the closest and most meaningful relationship between soil state and ambient temperature, providing strong support for subsequent spraying amount decisions.

[0081] In one embodiment of the present application, the ambient temperature-soil state multimodal essential fusion module includes: a standardization processing unit, which is used to standardize the global semantic enhancement feature vector of the soil surface state and the ambient temperature time series associated implicit feature vector to obtain a standardized global semantic enhancement feature vector of the soil surface state and a standardized ambient temperature time series associated implicit feature vector; a sample covariance matrix calculation unit, which is used to calculate the sample covariance matrix of the standardized global semantic enhancement feature vector of the soil surface state and the standardized ambient temperature time series associated implicit feature vector to obtain a soil surface state sample covariance matrix and an ambient temperature time series sample covariance matrix; a feature vector extraction calculation unit, which is used to extract the soil surface state sample covariance matrix from the soil surface state sample covariance matrix. The sample covariance matrix and the ambient temperature time series sample covariance matrix are subjected to matrix decomposition-based feature vector extraction to obtain a set of soil surface state principal component feature vectors and a set of ambient temperature time series principal component feature vectors; a maximum approximate query matching unit is used to input the set of soil surface state principal component feature vectors and the set of ambient temperature time series principal component feature vectors into a maximum approximate query matching network to obtain a set of best matching pairs of soil surface state principal component feature vectors and ambient temperature time series principal component feature vectors; a semantic fine-grained gating combination unit is used to combine each soil surface state principal component feature vector and the set of best matching pairs of soil surface state principal component feature vectors and ambient temperature time series principal component feature vectors. The best matching pair of the ambient temperature time series principal component feature vectors is input into the semantic fine-grained gating joint module to obtain a set of ambient temperature-soil state principal component fusion feature vectors; a vector cascade unit is used to cascade each ambient temperature-soil state principal component fusion feature vector in the set of ambient temperature-soil state principal component fusion feature vectors to obtain the ambient temperature-soil state multimodal essential fusion representation vector; wherein the semantic fine-grained gating joint unit is used to: perform position difference, position dot multiplication and position sum processing on the best matching pairs of each soil surface state principal component feature vector and the ambient temperature time series principal component feature vector respectively to obtain a set of ambient temperature-soil state principal component difference feature vectors, a circumferential difference feature vector, and a circumferential difference feature vector. The method comprises the following steps: obtaining a set of ambient temperature-soil state principal component dot product feature vectors and a set of ambient temperature-soil state principal component sum feature vectors; cascading the set of ambient temperature-soil state principal component difference feature vectors, the set of ambient temperature-soil state principal component dot product feature vectors and the set of ambient temperature-soil state principal component sum feature vectors by position to obtain a set of ambient temperature-soil state principal component multidimensional fusion expression vectors; processing each ambient temperature-soil state principal component multidimensional fusion expression vector in the set of ambient temperature-soil state principal component multidimensional fusion expression vectors by a one-dimensional convolution layer and then by a maximum pooling layer to obtain the set of ambient temperature-soil state principal component fusion feature vectors.

[0082] Furthermore, in one embodiment of the present application, the standardization processing unit is used to: calculate the mean and standard deviation of the global semantic enhancement feature vector of the soil surface state to obtain the mean and standard deviation of the soil surface state; calculate the positional difference between the global semantic enhancement feature vector of the soil surface state and the mean of the soil surface state, and then divide the difference by position with the standard deviation of the soil surface state to obtain the standardized global semantic enhancement feature vector of the soil surface state; calculate the mean and standard deviation of the ambient temperature time series associated implicit feature vector to obtain the ambient temperature time series mean and standard deviation of the ambient temperature time series; calculate the positional difference between the ambient temperature time series associated implicit feature vector and the ambient temperature time series mean, and then divide the difference by position with the standard deviation of the ambient temperature time series to obtain the standardized ambient temperature time series associated implicit feature vector.

[0083] Furthermore, in one embodiment of the present application, the maximum approximate query matching unit is used to: take the soil surface state principal component feature vector at a predetermined position in the set of the soil surface state principal component feature vectors to obtain a predetermined soil surface state principal component feature vector; calculate the inner product between the predetermined soil surface state principal component feature vector and each ambient temperature time series principal component feature vector in the set of the ambient temperature time series principal component feature vector to obtain a sequence of ambient temperature-soil state inner product values; calculate the -norm of the predetermined soil surface state principal component feature vector to obtain the -norm of the predetermined soil surface state principal component feature vector; calculate the ambient temperature time series principal component feature vector. The one-norm of each ambient temperature time series principal component feature vector in the set of component feature vectors is obtained to obtain a sequence of the one-norm of the ambient temperature time series principal component feature; the one-norm of the predetermined soil surface state principal component feature and the sequence of the one-norm of the ambient temperature time series principal component feature are multiplied by the position point to obtain a sequence of ambient temperature-soil state semantic interaction values; after calculating the point-by-position division between the sequence of ambient temperature-soil state inner product values ​​and the sequence of ambient temperature-soil state semantic interaction values, the ambient temperature time series principal component feature vector corresponding to the maximum eigenvalue in the obtained sequence of eigenvalues ​​and the predetermined soil surface state principal component feature vector form a set of best matching pairs.

[0084] Specifically, the soil surface state global semantic enhancement feature vector and the ambient temperature time series associated implicit feature vector are input into the feature principal component query matching-based saliency fusion network and processed using the following feature principal component query matching formula to obtain the ambient temperature-soil state multimodal essential fusion representation vector;

[0085] The characteristic principal component query matching formula is:

[0086]

[0087] C1=U1Λ1U1 T

[0088] C2=U2Λ2U2 T

[0089] U1=[v 11 ,v 12 ,…,v 1m ]

[0090] Λ1=diag(λ 11 ,λ 12 ,…,λ 1m )

[0091] U2=[v 21 ,v 22 ,…,v 2m ]

[0092] Λ2=diag(λ 21 ,λ 22 ,…,λ 2m )

[0093]

[0094] v f =[v p1 ;v p2 ;...v pm ]

[0095] Among them, V1 is the global semantic enhancement feature vector of the soil surface state, V2 is the implicit feature vector associated with the ambient temperature time series, μ1 is the mean of the global semantic enhancement feature vector of the soil surface state, σ1 is the standard deviation of the global semantic enhancement feature vector of the soil surface state, μ2 is the mean of the implicit feature vector associated with the ambient temperature time series, σ2 is the standard deviation of the implicit feature vector associated with the ambient temperature time series, Indicates position difference, V 1norm is the global semantic enhancement feature vector of the standardized soil surface state, V 2norm is the standardized ambient temperature time series associated implicit feature vector, n1 is the number of eigenvalues ​​in the standardized soil surface state global semantic enhanced feature vector, n2 is the number of eigenvalues ​​in the standardized ambient temperature time series associated implicit feature vector, C1 is the soil surface state sample covariance matrix, C2 is the ambient temperature time series sample covariance matrix, v 11 ,v 12 ,…,v 1mis the principal component eigenvector of each soil surface state in the set of principal component eigenvectors of soil surface state, U1 is the orthogonal matrix of principal component of soil surface state, diag(λ 11 ,λ 12 ,…,λ 1m ) indicates that the diagonal elements in the matrix are λ 11 ,λ 12 ,…,λ 1m The diagonal matrix of 11 ,λ 12 ,…,λ 1m represents the weight parameters of the principal component eigenvectors of each soil surface state, Λ1 is the diagonal matrix of soil surface state, v 21 ,v 22 ,…,v 2m is the principal component eigenvector of each ambient temperature time series in the set of principal component eigenvectors of ambient temperature time series, U2 is the orthogonal matrix of principal components of ambient temperature time series, diag(λ 21 ,λ 22 ,…,λ 2m ) indicates that the diagonal elements in the matrix are λ 21 ,λ 22 ,…,λ 2m The diagonal matrix of the ambient temperature, Λ2 is the diagonal matrix of the ambient temperature time series, λ 21 ,λ 22 ,…,λ 2m Respectively represent the weight parameters of the principal component eigenvectors of each ambient temperature time series, v 1i is the i-th soil surface state principal component eigenvector in the set of soil surface state principal component eigenvectors, v 2i is the i-th ambient temperature time series principal component eigenvector in the set of the ambient temperature time series principal component eigenvectors, <·> represents the inner product of the vectors, ||·|| represents the one-norm of the vector, Indicates that the j value corresponding to the maximum value point is returned, k is the maximum approximate query matching value, ⊙ and They represent positional multiplication and positional addition, conv1D represents a one-dimensional convolutional layer, MaxPool represents maximum pooling, and v pi is the ith ambient temperature-soil state principal component fusion feature vector in the set of ambient temperature-soil state principal component fusion feature vectors, m is the number of feature vectors in the set of ambient temperature-soil state principal component fusion feature vectors, v f It is the multi-modal essential fusion representation vector of the ambient temperature-soil state.

[0096] Then, the multimodal essential fusion representation vector of ambient temperature and soil state is input into the spraying intelligent controller based on the decoder to obtain a control instruction, which is used to represent the decoded value of the recommended spraying amount. In other words, the essential fusion feature information between the ambient temperature time series features and the global semantic reinforcement features of the soil state is used for decoding regression to recommend and control the spraying amount. In this way, the spraying amount of the improver can be dynamically adjusted according to the specific conditions of the soil and the changes in the ambient temperature, thereby improving the intelligence level of the control of the soil improvement agricultural equipment and helping to achieve better soil improvement quality and efficiency.

[0097] In a preferred example, considering that the global semantic enhancement feature vector of the soil surface state and the ambient temperature time series association implicit feature vector respectively represent the global image association features of the soil state image based on the converter structure and the local time series association features of the ambient temperature, when performing significant fusion based on feature principal component query matching, due to the feature modality differences caused by the source data modality differences and the feature encoding method differences, the ambient temperature-soil state multimodal essential fusion representation vector will also have a feature modality diversified set expression distribution. Therefore, it is expected to improve the balance between the regression mapping accuracy and completeness of the ambient temperature-soil state multimodal essential fusion representation vector when it is input into the decoder-based spraying intelligent controller for decoding regression, thereby improving the accuracy of the obtained control instructions.

[0098] Based on this, in the preferred example, inputting the ambient temperature-soil state multimodal essential fusion representation vector into the decoder-based spraying intelligent controller to obtain the control instruction specifically includes:

[0099] Determine the maximum eigenvalue of the environment temperature-soil state multimodal essence fusion and the minimum eigenvalue of the environment temperature-soil state multimodal essence fusion representation vector, and calculate the mean value of the environment temperature-soil state multimodal essence fusion and the standard deviation of the environment temperature-soil state multimodal essence fusion of the feature set of the environment temperature-soil state multimodal essence fusion representation vector;

[0100] Calculating the quotient of the ambient temperature-soil state multimodal essential fusion mean and the ambient temperature-soil state multimodal essential fusion standard deviation to obtain an ambient temperature-soil state multimodal essential fusion statistical normalization value;

[0101] Calculate the inverse of each eigenvalue of the ambient temperature-soil state multimodal essence fusion representation vector, multiply it by the difference between the ambient temperature-soil state multimodal essence fusion maximum eigenvalue and the ambient temperature-soil state multimodal essence fusion minimum eigenvalue, and then subtract it from the ambient temperature-soil state multimodal essence fusion statistical normalization value to obtain an ambient temperature-soil state multimodal essence fusion distribution approximation vector;

[0102] Calculate an exponential function with a natural constant as the base, where each eigenvalue of the ambient temperature-soil state multimodal essential fusion distribution approximation vector is an exponential function, and perform point addition with the ambient temperature-soil state multimodal essential fusion statistical normalization value to obtain an ambient temperature-soil state multimodal essential fusion probability approximation vector;

[0103] Calculating the base-two logarithm of the absolute value of each eigenvalue of the ambient temperature-soil state multimodal essential fusion probability approximation vector to obtain an optimized ambient temperature-soil state multimodal essential fusion representation vector; and

[0104] The optimized ambient temperature-soil state multimodal essential fusion representation vector is input into a decoder-based spraying intelligent controller to obtain a control instruction.

[0105] The optimal representation of the multi-modal essential fusion representation vector of the ambient temperature and soil state, denoted as V, is:

[0106]

[0107] α, β, v min and v max are the feature set v of the multi-modal essential fusion representation vector of the ambient temperature and soil state. i ∈V, and log is the logarithm with base 2, V ⊙-1 is the reciprocal of each eigenvalue of the multi-modal essential fusion representation vector of the ambient temperature-soil state, exp[·] is an exponential function calculated with a natural constant as the base and each eigenvalue of the vector as the exponent, V' is the optimized multi-modal essential fusion representation vector of the ambient temperature-soil state, It is added by location point. is point-wise subtraction, and ⊙ is point-wise multiplication.

[0108] That is, by performing random statistical normalization on the diversified feature set of the multimodal fusion representation vector of the ambient temperature-soil state, the confidence space constructed based on the overall eigenvalue of the multimodal fusion representation vector of the ambient temperature-soil state is approximated by the standardized continuous probability density distribution relative to the response hypothesis test of each eigenvalue of the multimodal fusion representation vector of the ambient temperature-soil state, thereby establishing the target reachability of the discrete feature distribution of the multimodal fusion representation vector of the ambient temperature-soil state to the unified regression target, so as to achieve the balanced executableness between mapping accuracy and mapping integrity in the decoding regression process based on the discrete feature distribution of the multimodal fusion representation vector of the ambient temperature-soil state, and improve the accuracy of the control instructions obtained by the spraying intelligent controller based on the decoder of the multimodal fusion representation vector of the ambient temperature-soil state. In this way, the spraying amount of the improver can be adjusted more accurately according to the specific conditions of the soil and the changes in the ambient temperature, thereby improving the intelligence level of the control of the agricultural equipment for improving the soil, and helping to achieve better soil improvement quality and efficiency.

[0109] In summary, the above scheme is adopted. During the operation of the improver spraying mechanism, the ambient temperature data is collected in real time through the temperature sensor, and the soil surface state image is collected through the camera. The data processing and image understanding algorithms based on artificial intelligence and deep learning are used in the back end to analyze these ambient temperature data and soil surface state images, so as to capture and characterize the multimodal essential fusion characteristics between the soil surface state semantics and the ambient temperature time series semantics, so as to recommend and control the spraying amount. In this way, the spraying amount of the improver can be dynamically adjusted according to the specific conditions of the soil and the changes in the ambient temperature, thereby improving the intelligence level of the control of the soil improvement agricultural equipment and helping to achieve better soil improvement quality and efficiency.

[0110] Figure 3 is a flow chart of a smart agricultural method for improving soil according to an exemplary embodiment. Figure 3 As shown, the method includes:

[0111] Step S201, using a soil turning mechanism to loosen the soil to form loose soil and large soil blocks; step S202, using a shoveling mechanism to shovel up the loose soil and the large soil blocks; step S203, spraying an improver into the loose soil and the large soil blocks through an improver spraying mechanism; step S204, using a transport mechanism to transport the loose soil and the large soil blocks to a crushing mechanism; step S205, using a crushing mechanism to crush the large soil blocks to obtain improved soil;

[0112] Wherein, step S203, spraying the improver into the loose soil and the bulk soil by the improver spraying mechanism, comprises:

[0113] Step S2031, obtaining a time queue of the ambient temperature collected by the temperature sensor;

[0114] Step S2032, obtaining a soil surface state image collected by a camera;

[0115] Step S2033, inputting the soil surface state image into a soil surface state feature extractor based on the ViT model to obtain a sequence of semantic feature vectors of local area states of the soil surface;

[0116] Step S2034, performing feature sequence selection optimization aggregation processing based on feature energy distribution spectrum on the sequence of semantic feature vectors of the local area state of the soil surface to obtain a global semantic enhancement feature vector of the soil surface state;

[0117] Step S2035, inputting the time queue of the ambient temperature into the ambient temperature sequence encoder based on the RNN network to obtain an ambient temperature time series associated implicit feature vector;

[0118] Step S2036: input the soil surface state global semantic enhancement feature vector and the ambient temperature time series associated implicit feature vector into a significant fusion network based on feature principal component query matching to obtain an ambient temperature-soil state multimodal essential fusion representation vector;

[0119] Step S2037: input the ambient temperature-soil state multimodal essential fusion representation vector into the decoder-based spraying intelligent controller to obtain a control instruction, wherein the control instruction is used to represent a decoded value of the recommended spraying amount.

[0120] Reference below Figure 4 , which shows a schematic diagram of the structure of an electronic device 600 suitable for implementing the embodiment of the present application. The terminal device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 4 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0121] like Figure 4As shown, the electronic device 600 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0122] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 4 The electronic device 600 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0123] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 609, or installed from the storage device 608, or installed from the ROM 602. When the computer program is executed by the processing device 601, the above-mentioned functions defined in the method of the embodiment of the present application are executed.

[0124] It should be noted that the computer-readable medium mentioned above in the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0125] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0126] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0127] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including, but not limited to, object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0128] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0129] The modules involved in the embodiments described in this application can be implemented by software or hardware. The name of the module does not limit the module itself in some cases. For example, the test parameter acquisition module can also be described as a "module for acquiring device test parameters corresponding to the target device".

[0130] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0131] In the context of the present application, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0132] Figure 5 FIG. 1 is an application scenario diagram of an intelligent agricultural device for improving soil according to an exemplary embodiment. Figure 5 As shown, in this application scenario, first, the time queue of the ambient temperature collected by the temperature sensor is obtained (for example, Figure 5 C1 shown in the figure); obtaining a soil surface state image collected by a camera (for example, Figure 5 Then, the acquired time sequence of the ambient temperature and the soil surface state image are input to a server (for example, Figure 5 In S) shown in , the server is capable of processing the time queue of the ambient temperature and the soil surface state image based on the smart agricultural equipment algorithm for improving soil to obtain a control instruction, and the control instruction is used to represent a decoded value of the recommended spraying amount.

[0133] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other to form a technical solution.

[0134] In addition, although each operation is described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or to be performed in a sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the application. Some features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination mode.

[0135] Although the subject matter has been described in language specific to structural features and / or method logic actions, it should be understood that the subject matter defined is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely exemplary forms of implementation. With respect to the apparatus in the above-mentioned embodiments, the specific manner in which each module performs the operation has been described in detail in the embodiments related to the method, and will not be described in detail here.

Claims

1. An intelligent agricultural device for improving soil, comprising: A vehicle body, a soil turning mechanism, a soil shoveling mechanism, a transport mechanism, a crushing mechanism and an improver spraying mechanism arranged on the vehicle body; wherein the soil turning mechanism is used to loosen the soil to form loose soil and large soil blocks; the soil shoveling mechanism is used to shovel up the loose soil and the large soil blocks, and spray the improver into the loose soil and the large soil blocks through the improver spraying mechanism; the transport mechanism is used to transport the loose soil and the large soil blocks to the crushing mechanism; the crushing mechanism is used to crush the large soil blocks to obtain improved soil; Characterized in that, the improver spraying mechanism comprises: An ambient temperature acquisition module is used to obtain a time queue of the ambient temperature acquired by a temperature sensor; A soil surface state image acquisition module is used to acquire a soil surface state image acquired by a camera; A soil surface local area state feature extraction module, used for inputting the soil surface state image into a soil surface state feature extractor based on a ViT model to obtain a sequence of soil surface local area state semantic feature vectors; A soil surface local area state semantic selection optimization aggregation module is used to perform feature sequence selection optimization aggregation processing based on a feature energy distribution spectrum on the sequence of the soil surface local area state semantic feature vectors to obtain a soil surface state global semantic enhancement feature vector; An ambient temperature time series encoding module, used for inputting the time queue of the ambient temperature into an ambient temperature sequence encoder based on an RNN network to obtain an ambient temperature time series associated implicit feature vector; An ambient temperature-soil state multimodal essential fusion module is used to input the soil surface state global semantic enhancement feature vector and the ambient temperature time series associated implicit feature vector into a significant fusion network based on feature principal component query matching to obtain an ambient temperature-soil state multimodal essential fusion representation vector; The spraying control module is used to input the multi-modal essential fusion representation vector of the ambient temperature-soil state into a spraying intelligent controller based on a decoder to obtain a control instruction, wherein the control instruction is used to represent a decoded value of a recommended spraying amount.

2. The intelligent agricultural equipment for improving soil according to claim 1, characterized in that: The soil surface local area state semantic selection optimization aggregation module includes: A soil surface local state feature selection and optimization unit, used for inputting the sequence of soil surface local area state semantic feature vectors into a feature sequence selection and optimization module based on a characteristic energy distribution spectrum to obtain a selected sequence of soil surface local area state semantic feature vectors; The soil surface state global semantic feature expression unit is used to cascade the selected sequence of soil surface local area state semantic feature vectors to obtain the soil surface state global semantic enhanced feature vector.

3. The intelligent agricultural equipment for improving soil according to claim 2, characterized in that: The soil surface local state feature selection optimization unit comprises: A characteristic energy distribution spectrum vector calculation subunit is used to calculate the characteristic energy distribution spectrum vector of each soil surface local area state semantic feature vector in the sequence of soil surface local area state semantic feature vectors to obtain a sequence of soil surface local area state semantic feature energy distribution spectrum vectors; A position mean vector calculation subunit is used to calculate the position mean vector of the sequence of the soil surface local area state semantic feature energy distribution spectrum vector as the soil surface local area state semantic energy spectrum distribution center vector; An energy distribution spectrum span factor calculation subunit is used to calculate the energy distribution spectrum span factor between the soil surface local area state semantic energy spectrum distribution center vector and each soil surface local area state semantic feature energy distribution spectrum vector in the sequence of the soil surface local area state semantic feature energy distribution spectrum vector to obtain a sequence of soil surface local area state semantic energy distribution spectrum span factors; A span factor and preset threshold comparison subunit, used to determine the sequence of the selected soil surface local area state semantic feature vectors based on the comparison between each soil surface local area state semantic energy distribution spectrum span factor in the sequence of soil surface local area state semantic energy distribution spectrum span factors and the preset threshold; Wherein, the energy distribution spectrum span factor calculation subunit includes: A secondary subunit for calculating positional differences is used to calculate the positional differences between the soil surface local area state semantic feature energy distribution spectrum vector and the soil surface local area state semantic energy spectrum distribution center vector to obtain the soil surface local area state semantic differential feature vector; The covariance matrix calculation secondary subunit is used to calculate the covariance matrix between the central vector of the soil surface local area state semantic energy spectrum distribution and the soil surface local area state semantic feature energy distribution spectrum vector, and calculate the inverse of the covariance matrix to obtain the soil surface local area state semantic inverse covariance matrix; The state semantic distribution value calculation secondary subunit is used to calculate the transposed vector of the state semantic differential feature vector of the local area of ​​the soil surface and the product of the state semantic inverse covariance matrix of the local area of ​​the soil surface and the state semantic differential feature vector of the local area of ​​the soil surface to obtain the state semantic distribution value of the local area of ​​the soil surface; The semantic energy distribution spectrum span factor calculation secondary subunit is used to calculate the square root of the semantic distribution value of the local area state of the soil surface to obtain the semantic energy distribution spectrum span factor of the local area state of the soil surface.

4. The intelligent agricultural equipment for improving soil according to claim 3, characterized in that: The characteristic energy distribution spectrum vector calculation subunit is used for: A feature distribution energy collaborative representation vector calculation secondary subunit is used to respectively calculate the feature distribution energy collaborative representation vector between each soil surface local area state semantic feature vector and each other soil surface local area state semantic feature vector in the sequence of soil surface local area state semantic feature vectors to obtain a sequence of multiple soil surface local area state semantic feature distribution energy collaborative representation vectors; The secondary subunit for calculating the feature distribution energy synergy factor vector is used to calculate the feature distribution energy synergy factor vector of each sequence of the soil surface local area state semantic feature distribution energy synergy representation vector in the sequence of the multiple soil surface local area state semantic feature distribution energy synergy representation vectors to obtain the sequence of the soil surface local area state semantic feature energy distribution spectrum vectors.

5. The intelligent agricultural equipment for improving soil according to claim 4, characterized in that: The feature distribution energy collaborative representation vector calculation secondary subunit is used to: Extracting a predetermined soil surface local area state semantic feature vector from a sequence of characteristic energy distribution spectrum vectors of the soil surface local area state semantic feature vector; The predetermined soil surface local area state semantic feature vector and each other soil surface local area state semantic feature vector are respectively subjected to element-by-element maximum value extraction to obtain a sequence of soil surface local area state semantic feature distribution energy collaborative representation vectors corresponding to the predetermined soil surface local area state semantic feature vector.

6. The intelligent agricultural equipment for improving soil according to claim 5, characterized in that: The characteristic distribution energy synergy factor vector calculation secondary subunit is used for: Respectively extracting the maximum value of each soil surface local area state semantic feature distribution energy collaborative representation vector in the sequence of soil surface local area state semantic feature distribution energy collaborative representation vector to obtain a sequence of soil surface local area state semantic feature maximum values; The mean and variance of the energy cooperative representation vector of each local area state semantic feature distribution of the soil surface are calculated respectively to obtain a sequence of mean values ​​of the local area state semantic features of the soil surface and a sequence of variances of the local area state semantic features of the soil surface; Adding the sequence of the variance of the local area state semantic feature of the soil surface and the preset hyperparameter by position to obtain a sequence of the first energy distribution coefficient of the local area state semantic feature of the soil surface; Calculating the square of the difference between the maximum value of the local area state semantic feature of the soil surface and the mean value of the local area state semantic feature of the soil surface in each group corresponding to the sequence of the maximum value of the local area state semantic feature of the soil surface and the sequence of the mean value of the local area state semantic feature of the soil surface to obtain a sequence of local area state semantic difference values ​​of the soil surface; Multiplying the sequence of the variance of the semantic feature of the local area state of the soil surface by a constant 2 by position to obtain a sequence of the variance of the semantic feature of the local area state of the soil surface modulated; After adding each group of corresponding soil surface local area state semantic difference values ​​and modulated soil surface local area state semantic feature variances in the sequence of soil surface local area state semantic difference values ​​and the sequence of modulated soil surface local area state semantic feature variances, the sum is added to the preset hyperparameter to obtain a sequence of second energy distribution coefficients of soil surface local area state semantic features; The sequence of the first energy distribution coefficients of the semantic features of the local area state of the soil surface is divided by the sequence of the second energy distribution coefficients of the semantic features of the local area state of the soil surface by position to obtain the energy distribution spectrum vector of the semantic features of the local area state of the soil surface composed of multiple feature distribution energy synergy factors.

7. The intelligent agricultural equipment for improving soil according to claim 6, characterized in that: The ambient temperature-soil state multi-modal essential fusion module includes: A standardization processing unit, used for performing standardization processing on the soil surface state global semantic enhancement feature vector and the ambient temperature time series associated implicit feature vector to obtain a standardized soil surface state global semantic enhancement feature vector and a standardized ambient temperature time series associated implicit feature vector; A sample covariance matrix calculation unit is used to calculate the sample covariance matrix of the standardized soil surface state global semantic enhancement feature vector and the standardized ambient temperature time series associated implicit feature vector to obtain a soil surface state sample covariance matrix and an ambient temperature time series sample covariance matrix; A feature vector extraction calculation unit is used to perform feature vector extraction based on matrix decomposition on the soil surface state sample covariance matrix and the ambient temperature time series sample covariance matrix to obtain a set of soil surface state principal component feature vectors and a set of ambient temperature time series principal component feature vectors; A maximum approximate query matching unit, used for inputting the set of the soil surface state principal component feature vectors and the set of the ambient temperature time series principal component feature vectors into a maximum approximate query matching network to obtain a set of best matching pairs of the soil surface state principal component feature vectors and the ambient temperature time series principal component feature vectors; A semantic fine-grained gating union unit is used to input each best matching pair of the soil surface state principal component feature vector and the ambient temperature time series principal component feature vector in the set of best matching pairs of the soil surface state principal component feature vector and the ambient temperature time series principal component feature vector into a semantic fine-grained gating union module to obtain a set of ambient temperature-soil state principal component fusion feature vectors; A vector cascading unit, used for cascading each ambient temperature-soil state principal component fusion feature vector in the set of the ambient temperature-soil state principal component fusion feature vector to obtain the ambient temperature-soil state multimodal essential fusion representation vector; The semantic fine-grained gating joint unit is used to: The best matching pairs of the principal component eigenvectors of the soil surface state and the principal component eigenvectors of the ambient temperature time series are respectively subjected to position difference, position dot multiplication and position summation processing to obtain a set of ambient temperature-soil state principal component difference eigenvectors, a set of ambient temperature-soil state principal component dot product eigenvectors and a set of ambient temperature-soil state principal component summation eigenvectors; The set of the ambient temperature-soil state principal component difference eigenvectors, the set of the ambient temperature-soil state principal component dot product eigenvectors and the set of the ambient temperature-soil state principal component sum eigenvectors are cascaded by position to obtain a set of ambient temperature-soil state principal component multidimensional fusion expression vectors; Each of the multidimensional fusion expression vectors of the environmental temperature-soil state principal component in the set of the multidimensional fusion expression vectors of the environmental temperature-soil state principal component is processed by a one-dimensional convolution layer and then by a maximum pooling layer to obtain a set of the environmental temperature-soil state principal component fusion feature vectors.

8. The intelligent agricultural equipment for improving soil according to claim 7, characterized in that: The standardization processing unit is used for: Calculating the mean and standard deviation of the global semantic enhancement feature vector of the soil surface state to obtain the mean of the soil surface state and the standard deviation of the soil surface state; Calculating the positional difference between the global semantic enhancement feature vector of the soil surface state and the mean value of the soil surface state, and then dividing the positional difference by the standard deviation of the soil surface state to obtain the standardized global semantic enhancement feature vector of the soil surface state; Calculating the mean and standard deviation of the ambient temperature time series associated implicit feature vector to obtain the ambient temperature time series mean and the ambient temperature time series standard deviation; The positional difference between the ambient temperature time series associated implicit feature vector and the ambient temperature time series mean is calculated and then divided by the ambient temperature time series standard deviation to obtain the standardized ambient temperature time series associated implicit feature vector.

9. The intelligent agricultural equipment for improving soil according to claim 8, characterized in that: The maximum approximate query matching unit is used to: Taking the soil surface state principal component feature vector at a predetermined position in the set of soil surface state principal component feature vectors to obtain a predetermined soil surface state principal component feature vector; Calculating the inner products of the predetermined soil surface state principal component eigenvector and each ambient temperature time series principal component eigenvector in the set of ambient temperature time series principal component eigenvectors to obtain a sequence of ambient temperature-soil state inner product values; Calculating a norm of the principal component characteristic vector of the predetermined soil surface state to obtain a norm of the principal component characteristic of the predetermined soil surface state; Calculating the one-norm of each ambient temperature time series principal component feature vector in the set of the ambient temperature time series principal component feature vectors to obtain a sequence of one-norm features of the ambient temperature time series principal component; Multiplying the sequence of the predetermined soil surface state principal component characteristic one norm and the ambient temperature time series principal component characteristic one norm by position points to obtain a sequence of ambient temperature-soil state semantic interaction values; After calculating the point-by-point division between the sequence of the ambient temperature-soil state inner product values ​​and the sequence of the ambient temperature-soil state semantic interaction values, the ambient temperature time series principal component eigenvector corresponding to the maximum eigenvalue in the obtained eigenvalue sequence and the predetermined soil surface state principal component eigenvector form a set of optimal matching pairs.

10. A smart agricultural method for improving soil, comprising: Use a soil turning mechanism to loosen the soil to form loose soil and large clumps of soil; The loose soil and the bulk soil are scooped up by a shoveling mechanism, and the improver is sprayed into the loose soil and the bulk soil by an improver spraying mechanism; the loose soil and the bulk soil are transported to a crushing mechanism by a transporting mechanism; and the bulk soil is crushed by a crushing mechanism to obtain improved soil; The invention is characterized in that the improver is sprayed into the loose soil and the bulk soil by an improver spraying mechanism, comprising: Get the time queue of the ambient temperature collected by the temperature sensor; Acquire a soil surface state image collected by a camera; Inputting the soil surface state image into a soil surface state feature extractor based on the ViT model to obtain a sequence of semantic feature vectors of local area states of the soil surface; Performing a feature sequence selection optimization aggregation process based on a feature energy distribution spectrum on the sequence of the semantic feature vectors of the local area state of the soil surface to obtain a global semantic enhancement feature vector of the soil surface state; Inputting the time queue of the ambient temperature into an ambient temperature sequence encoder based on an RNN network to obtain an ambient temperature time series associated implicit feature vector; Inputting the soil surface state global semantic enhancement feature vector and the ambient temperature time series associated implicit feature vector into a significant fusion network based on feature principal component query matching to obtain an ambient temperature-soil state multimodal essential fusion representation vector; The multi-modal essential fusion representation vector of the ambient temperature and soil state is input into a spraying intelligent controller based on a decoder to obtain a control instruction, wherein the control instruction is used to represent a decoded value of a recommended spraying amount.

Images