Saline-alkali soil micro-ecological improvement method based on ditcher

Abstract

The invention relates to a saline-alkali soil micro-ecological improvement method based on a ditcher. The method comprises the following steps: controlling a ditcher to perform soil ditching so as to obtain a soil groove, acquiring a ditching soil state image of the soil groove through a camera, and then introducing an image processing and analysis algorithm based on artificial intelligence and deep learning at the rear end to analyze the ditching soil state image, therefore, the semantic representation of the soil state related to the groove in the image is captured and depicted, corresponding irrigation control is carried out, and an irrigation instruction aiming at the specific soil state is generated. Therefore, the irrigation amount can be automatically adjusted based on the actual ditching soil state, and the problem of salt accumulation caused by excessive irrigation is avoided, so that more intelligent and accurate water management is realized, the utilization efficiency of water resources is optimized, and ecological restoration and agricultural sustainable development of saline-alkali soil are promoted.

Description

Technical Field

[0001] The present application relates to the technical field of soil microecology improvement, and in particular, to a method for improving the soil microecology of saline-alkali land based on a trenching machine. Background Art

[0002] Saline-alkali land refers to land with a high salt content and alkaline pH value. This type of soil usually has low fertility and water retention capacity, which seriously affects the growth and yield of crops. Saline-alkali land is widely distributed around the world, especially in arid and semi-arid areas, and has become a major challenge for agricultural production. Therefore, improving the soil microecology of saline-alkali land is the key to improving the efficiency of such land use. Traditional methods for improving saline-alkali land mainly include physical, chemical and biological measures, such as deep plowing, application of organic fertilizers, and planting of salt-tolerant plants, but these methods often take a long time to be effective and may have certain negative impacts on the environment.

[0003] Therefore, an optimized solution for improving the soil microecology of saline-alkali land is desired. Summary of the invention

[0004] 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.

[0005] In a first aspect, the present application provides a method for improving soil microecology in saline-alkali land based on a trenching machine, the method comprising: Controlling a trenching machine to trench the soil to obtain a soil trench; Acquire a trench soil state image of the soil trench collected by a camera; Performing multi-scale feature extraction based on soil state on the trenching soil state image to obtain a trenching soil state shallow layer feature map and a trenching soil state deep layer feature map; Inputting the shallow characteristic map of the trenching soil state into a feature enhancement module based on feature gradient distribution to obtain a shallow significant characteristic map of the trenching soil state; Inputting the shallow significant feature map of the trenching soil state and the deep feature map of the trenching soil state into a feature significant joint perception network to obtain a shallow-deep joint significant perception feature map of the trenching soil state; Based on the shallow-deep joint significant perception feature map of the trenching soil state, an irrigation control instruction is generated, and the irrigation control instruction is used to represent a decoded value of the recommended irrigation amount.

[0006] Optionally, multi-scale feature extraction based on soil state is performed on the trenching soil state image to obtain a trenching soil state shallow feature map and a trenching soil state deep feature map, including: inputting the trenching soil state image into a soil state multi-scale feature extractor based on a hollow pyramid network to obtain the trenching soil state shallow feature map and the trenching soil state deep feature map.

[0007] Optionally, the shallow characteristic map of the trenching soil state is input into a feature enhancement module based on feature gradient distribution to obtain a shallow significant characteristic map of the trenching soil state, including: calculating the multi-directional gradient value distribution of each position in the shallow characteristic map of the trenching soil state, and determining the gradient amplitude value of each position in the shallow characteristic map of the trenching soil state based on the multi-directional gradient value distribution of each position to obtain a shallow characteristic gradient amplitude distribution feature map of the trenching soil state; calculating the gradient amplitude local description operator of each position in the shallow characteristic gradient amplitude distribution feature map of the trenching soil state to obtain a shallow characteristic gradient amplitude local significant distribution feature map of the trenching soil state; inputting the shallow characteristic gradient amplitude local significant distribution feature map of the trenching soil state into a gated masker based on a GELU function to obtain a shallow characteristic gradient amplitude local significant gated mask feature map of the trenching soil state; calculating the position point multiplication between the shallow characteristic gradient amplitude local significant gated mask feature map of the trenching soil state and the shallow characteristic map of the trenching soil state to obtain the shallow significant characteristic map of the trenching soil state.

[0008] Optionally, the multi-directional gradient value distribution of each position in the shallow characteristic map of the trenching soil state is calculated, and the gradient amplitude value of each position in the shallow characteristic map of the trenching soil state is determined based on the multi-directional gradient value distribution of each position to obtain the shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state, including: calculating the gradient amplitude value of each position characteristic value in the shallow characteristic map of the trenching soil state along the height direction to obtain the shallow characteristic height gradient distribution characteristic map of the trenching soil state; calculating the gradient amplitude value of each position characteristic value in the shallow characteristic map of the trenching soil state along the width direction to obtain the shallow characteristic width gradient distribution characteristic map of the trenching soil state; calculating the gradient amplitude value of each position characteristic value in the shallow characteristic map of the trenching soil state along the channel direction to obtain the shallow characteristic channel gradient distribution characteristic map of the trenching soil state; calculating the square root of the sum of the squares of the characteristic values ​​of the corresponding positions in the shallow characteristic height gradient distribution characteristic map of the trenching soil state, the shallow characteristic width gradient distribution characteristic map of the trenching soil state, and the shallow characteristic channel gradient distribution characteristic map of the trenching soil state to obtain the shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state.

[0009] Optionally, the local description operator of the gradient amplitude at each position in the shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state is calculated to obtain the local significant distribution characteristic map of the shallow characteristic gradient amplitude of the trenching soil state, including: determining the local neighborhood range of the gradient amplitude, and calculating the sequence of the shallow characteristic gradient amplitude distribution neighborhood characteristic map of the trenching soil state based on the local neighborhood range of the gradient amplitude and the shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state; counting the number of characteristic maps in the sequence of the shallow characteristic gradient amplitude distribution neighborhood characteristic map of the trenching soil state to obtain the statistical value of the number of shallow characteristic gradient amplitude distribution neighborhoods of the trenching soil state; calculating the shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state and the trenching soil state. The positional differences between each trenching soil state shallow characteristic gradient amplitude distribution neighborhood characteristic map in the sequence of soil state shallow characteristic gradient amplitude distribution neighborhood characteristic map are divided by the position point with the trenching soil state shallow characteristic gradient amplitude distribution characteristic map to obtain a sequence of trenching soil state shallow characteristic gradient relative amplitude distribution neighborhood characteristic map; the positional addition of each trenching soil state shallow characteristic gradient relative amplitude distribution neighborhood characteristic map in the sequence of trenching soil state shallow characteristic gradient relative amplitude distribution neighborhood characteristic map is divided by the position point with the shallow characteristic gradient amplitude distribution neighborhood number statistics value to obtain the trenching soil state shallow characteristic gradient amplitude local significant distribution characteristic map.

[0010] Optionally, the shallow significant feature map of the ditching soil state and the deep feature map of the ditching soil state are input into a feature significant joint perception network to obtain a shallow-deep joint significant perception feature map of the ditching soil state, including: reshaping the shallow significant feature map of the ditching soil state and the deep feature map of the ditching soil state to obtain a shallow significant feature shape reshaping matrix of the ditching soil state and a deep feature shape reshaping matrix of the ditching soil state; inputting the shallow significant feature shape reshaping matrix of the ditching soil state and the deep feature shape reshaping matrix of the ditching soil state into a feature channel-by-channel interactive perception module to obtain a shallow-deep feature dependency matrix of the ditching soil state and a deep-shallow feature dependency matrix of the ditching soil state; inputting the shallow-deep feature dependency matrix of the ditching soil state and the deep-shallow feature dependency matrix of the ditching soil state into a random inactivation module to obtain a pruning optimized shallow-deep feature dependency matrix of the ditching soil state and a pruning optimized ditching soil state. The invention relates to a method for obtaining a shallow-shallow feature dependency matrix of a soil state; multiplying the shallow-deep feature dependency matrix of the pruning-optimized ditching soil state with the deep feature shape reshaping matrix of the ditching soil state to obtain a deep-shallow feature dependency matrix of the pruning-optimized ditching soil state; multiplying the deep-shallow feature dependency matrix of the pruning-optimized ditching soil state with the shallow significant feature shape reshaping matrix of the ditching soil state to obtain a shallow-deep feature dependency matrix of the dependency-optimized ditching soil state; reshaping the feature shape of the deep-shallow feature dependency matrix of the dependency-optimized ditching soil state and the shallow-deep feature dependency matrix of the dependency-optimized ditching soil state to obtain a shallow significant feature map of the optimized ditching soil state and a deep feature map of the optimized ditching soil state; calculating the weighted sum of the shallow significant feature map of the optimized ditching soil state and the deep feature map of the optimized ditching soil state to obtain the shallow-deep joint significant perception feature map of the ditching soil state.

[0011] Optionally, the trenching soil state shallow significant feature shape reshaping matrix and the trenching soil state deep feature shape reshaping matrix are input into the feature channel-by-channel interactive perception module to obtain the trenching soil state shallow-deep feature dependency matrix and the trenching soil state deep-shallow feature dependency matrix, including: performing matrix multiplication on the transposed matrices of the trenching soil state shallow significant feature shape reshaping matrix and the trenching soil state deep feature shape reshaping matrix, and then dividing the matrix by the scale of the trenching soil state shallow-deep feature dependency matrix by position to obtain the trenching soil state shallow-deep feature interaction matrix; The transposed matrix of the shallow significant feature shape reshaping matrix of the trenching soil state and the deep feature shape reshaping matrix of the trenching soil state are multiplied, and then the matrix scale is divided by the deep feature shape reshaping matrix of the trenching soil state by position to obtain the deep-shallow feature interaction matrix of the trenching soil state; Softmax is used to normalize the shallow-deep feature interaction matrix of the trenching soil state and the deep-shallow feature interaction matrix of the trenching soil state to obtain the shallow-deep feature dependency matrix of the trenching soil state and the deep-shallow feature dependency matrix of the trenching soil state.

[0012] Optionally, based on the shallow-deep joint significant perceptual feature map of the trenching soil state, an irrigation control instruction is generated, and the irrigation control instruction is used to represent the recommended irrigation amount decoded value, including: inputting the shallow-deep joint significant perceptual feature map of the trenching soil state into a decoder-based intelligent irrigation control module to obtain an irrigation control instruction, and the irrigation control instruction is used to represent the recommended irrigation amount decoded value.

[0013] The above technical solution is adopted to control the trenching machine to trench the soil to obtain the soil trench, and collect the trenching soil state image of the soil trench through the camera, and then introduce the image processing and analysis algorithm based on artificial intelligence and deep learning in the back end to analyze the trenching soil state image, so as to capture and depict the semantic representation of the soil state of the trench in the image, so as to perform corresponding irrigation control to generate irrigation instructions for specific soil states. In this way, the irrigation amount can be automatically adjusted based on the actual trenching soil state to avoid the problem of salt accumulation caused by excessive irrigation, thereby realizing more intelligent and precise water management, optimizing the utilization efficiency of water resources, and promoting the ecological restoration of saline-alkali land and sustainable agricultural development.

[0014] Other features and advantages of the present application will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] 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: Figure 1 It is a flow chart of a method for improving soil microecology in saline-alkali land based on a trenching machine according to an exemplary embodiment.

[0016] Figure 2 It is a schematic structural diagram of a trencher according to an exemplary embodiment.

[0017] Figure 3 It is a block diagram of a saline-alkali land soil microecological improvement system based on a trenching machine according to an exemplary embodiment.

[0018] Figure 4 It is a block diagram of an electronic device according to an exemplary embodiment.

[0019] Figure 5 It is a diagram of an application scenario of a method for improving soil microecology in saline-alkali land based on a trenching machine according to an exemplary embodiment. DETAILED DESCRIPTION

[0020] 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.

[0021] 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.

[0022] 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.

[0023] 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.

[0024] 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".

[0025] 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.

[0026] In recent years, with the development of intelligent agricultural technology, more efficient and environmentally friendly methods for improving saline-alkali land have emerged. In particular, methods combining modern information technology and mechanical automation technology can achieve precise management and dynamic adjustment, thereby improving the intelligent level of soil microecological improvement in saline-alkali land. As an efficient soil tillage equipment, the application potential of trenching machines in saline-alkali land improvement has been gradually explored. Through trenching, the permeability and water distribution of the soil can be effectively improved, creating good conditions for subsequent irrigation and fertilization. However, in the irrigation process after soil trenching, soil moisture management is particularly important. Excessive irrigation will cause the groundwater level to rise, thereby aggravating soil salinization.

[0027] Based on this, in the technical solution of the present application, a method for improving the microecology of saline-alkali soil based on a trenching machine is proposed, which can reduce the accumulation of salt in the soil by precisely controlling irrigation, and improve the soil structure and microecological environment. Specifically, the technical concept of the present application is to control the trenching machine to trench the soil to obtain a soil trench, and collect the trenching soil state image of the soil trench through a camera, and then introduce an image processing and analysis algorithm based on artificial intelligence and deep learning in the back end to analyze the trenching soil state image, so as to capture and depict the semantic representation of the soil state of the trench in the image, so as to perform corresponding irrigation control to generate irrigation instructions for specific soil states. In this way, the irrigation amount can be automatically adjusted based on the actual trenching soil state to avoid the problem of salt accumulation caused by excessive irrigation, thereby achieving more intelligent and precise water management, optimizing the utilization efficiency of water resources, and promoting the ecological restoration of saline-alkali land and sustainable agricultural development.

[0028] The specific implementation of the present application is described in detail below with reference to the accompanying drawings.

[0029] Figure 1 is a flow chart of a method for improving soil microecology in saline-alkali land based on a trenching machine according to an exemplary embodiment, such as Figure 1As shown, the method includes: Step S101, controlling a trenching machine to trench the soil to obtain a soil trench; Step S102, obtaining a trenching soil state image of the soil trench collected by a camera; Step S103, performing multi-scale feature extraction based on soil state on the trenching soil state image to obtain a trenching soil state shallow layer feature map and a trenching soil state deep layer feature map; Step S104, inputting the shallow characteristic map of the trenching soil state into a feature enhancement module based on feature gradient distribution to obtain a shallow significant characteristic map of the trenching soil state; Step S105, inputting the shallow significant feature map of the trenching soil state and the deep feature map of the trenching soil state into a feature significant joint perception network to obtain a shallow-deep joint significant perception feature map of the trenching soil state; Step S106: Generate an irrigation control instruction based on the shallow-deep joint significant perception feature map of the trenching soil state, wherein the irrigation control instruction is used to represent a decoded value of the recommended irrigation amount.

[0030] Specifically, in the technical solution of the present application, first, the trenching machine is controlled to trench the soil to obtain a soil trench, and the trenching soil state image of the soil trench collected by the camera is obtained. Figure 2 shown.

[0031] Then, considering that the trench soil state image contains different detailed features and semantic information about the trench soil state, such as the size, structure, color and humidity of soil particles, and different features correspond to different scale ranges and hierarchical semantics in the soil state image, which can reflect the moisture and salinity of the soil. Based on this, in the technical solution of the present application, the trench soil state image is further input into the soil state multi-scale feature extractor based on the hollow pyramid network to obtain the shallow feature map of the trench soil state and the deep feature map of the trench soil state. Through the processing of the soil state multi-scale feature extractor based on the hollow pyramid network, not only the shallow features such as the edge, texture and color of the soil state in the trench soil state image can be captured, so the shallow feature map of the trench soil state obtained can help identify the texture, cracks and color distribution information on the soil surface. It can also capture higher-level soil state semantic information, such as the overall pattern of soil structure, changes in soil composition and changes in soil moisture, which are very important for understanding the state of the soil (such as moisture, compactness, etc.). Therefore, the soil state multi-scale feature extractor based on the hollow pyramid network can effectively extract the shallow and deep features of the trench soil state, thereby enhancing the ability to understand the soil state and the perception of the multi-dimensional state.

[0032] In one embodiment of the present application, multi-scale feature extraction based on soil state is performed on the trenching soil state image to obtain a trenching soil state shallow feature map and a trenching soil state deep feature map, including: inputting the trenching soil state image into a soil state multi-scale feature extractor based on a hollow pyramid network to obtain the trenching soil state shallow feature map and the trenching soil state deep feature map.

[0033] Then, since the shallow feature map of the trenching soil state contains shallow feature information about the trenching soil state, some of the features such as edges, textures, color distribution and details are of great significance for the identification and feature expression of the trenching soil state, while some are irrelevant interference features and redundant information. Therefore, in order to enhance the key information expression ability in the shallow feature map of the trenching soil state, such as cracks, wet areas or uneven color changes in the soil, so as to help more accurately understand and perceive the trenching soil state for irrigation control, in the technical solution of the present application, the shallow feature map of the trenching soil state is further input into the feature enhancement module based on the feature gradient distribution to obtain the shallow significant feature map of the trenching soil state. It should be understood that the feature enhancement module based on the feature gradient distribution is a technology for enhancing the expression ability of the feature map in the deep learning model. Its main function is to highlight the important information in the shallow feature map of the trenching soil state through the feature gradient amplitude measurement and the gated mask mechanism, thereby improving the performance of the model and the expression ability of the trenching soil state features. Specifically, firstly, by performing multi-directional gradient calculation on the shallow feature map of the trenching soil state, it is possible to detect which regions in the feature map have the most obvious shallow features such as edges and details, that is, which regions contain more trenching soil state related information. Then, based on the results of the gradient calculation, it is analyzed which features are most critical for representing specific objects or scenes. In this scheme, it is used to analyze which directions of gradient features are more important for the characterization of trenching soil states, such as cracks, wet areas, or uneven color changes in the soil. The importance of each feature is evaluated based on the gradient information. Larger gradient values ​​represent more important features because they represent significant changes in the soil image. Then, through the gating mechanism, certain regions in the feature map are enhanced or suppressed to ensure that key features related to subsequent tasks are retained, while non-key features are weakened. Finally, through the weighted point multiplication operation, the enhanced shallow features of the trenching soil state are combined with the original features to obtain the optimized shallow significant feature map of the trenching soil state. In this way, different trenching soil states can be better identified and distinguished, the sensitivity and generalization ability of the model to the input data can be improved, and the subsequent feature fusion and decision-making process can be more accurate.

[0034] In one embodiment of the present application, the trenching soil state shallow characteristic map is input into a feature enhancement module based on feature gradient distribution to obtain a trenching soil state shallow significant characteristic map, including: calculating the multi-directional gradient value distribution of each position in the trenching soil state shallow characteristic map, and determining the gradient amplitude value of each position in the trenching soil state shallow characteristic map based on the multi-directional gradient value distribution of each position to obtain the trenching soil state shallow characteristic gradient amplitude distribution feature map; calculating the gradient amplitude local description operator of each position in the trenching soil state shallow characteristic gradient amplitude distribution feature map to obtain the trenching soil state shallow characteristic gradient amplitude local significant distribution feature map; inputting the trenching soil state shallow characteristic gradient amplitude local significant distribution feature map into a gated masker based on a GELU function to obtain a trenching soil state shallow characteristic gradient amplitude local significant gated mask feature map; calculating the position point multiplication between the trenching soil state shallow characteristic gradient amplitude local significant gated mask feature map and the trenching soil state shallow characteristic map to obtain the trenching soil state shallow significant characteristic map.

[0035] Furthermore, in one embodiment of the present application, the multi-directional gradient value distribution of each position in the shallow characteristic map of the trenching soil state is calculated, and the gradient amplitude value of each position in the shallow characteristic map of the trenching soil state is determined based on the multi-directional gradient value distribution of each position to obtain the shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state, including: calculating the gradient amplitude value of each position in the shallow characteristic map of the trenching soil state along the height direction to obtain the shallow characteristic height gradient distribution characteristic map of the trenching soil state; calculating the gradient amplitude value of each position in the shallow characteristic map of the trenching soil state along the width ... The gradient amplitude value in the ditching soil state shallow characteristic width gradient distribution characteristic map is calculated; the gradient amplitude value of each position characteristic value in the ditching soil state shallow characteristic map is calculated along the channel direction to obtain the ditching soil state shallow characteristic channel gradient distribution characteristic map; the square root of the sum of the squares of the characteristic values ​​of the corresponding positions in the ditching soil state shallow characteristic height gradient distribution characteristic map, the ditching soil state shallow characteristic width gradient distribution characteristic map and the ditching soil state shallow characteristic channel gradient distribution characteristic map are calculated to obtain the ditching soil state shallow characteristic gradient amplitude distribution characteristic map.

[0036] Furthermore, in one embodiment of the present application, the local description operator of the gradient amplitude of each position in the shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state is calculated to obtain the local significant distribution characteristic map of the shallow characteristic gradient amplitude of the trenching soil state, including: determining the local neighborhood range of the gradient amplitude, and calculating the sequence of the shallow characteristic gradient amplitude distribution neighborhood characteristic map of the trenching soil state based on the local neighborhood range of the gradient amplitude and the shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state; counting the number of characteristic maps in the sequence of the shallow characteristic gradient amplitude distribution neighborhood characteristic map of the trenching soil state to obtain the statistical value of the number of shallow characteristic gradient amplitude distribution neighborhoods of the trenching soil state; calculating the shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state. The positional differences between the characteristic maps and the sequence of the shallow characteristic gradient amplitude distribution neighborhood characteristic maps of the trenching soil state are divided by the shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state according to the position points to obtain a sequence of the shallow characteristic gradient relative amplitude distribution neighborhood characteristic maps of the trenching soil state; the positional addition of the shallow characteristic gradient relative amplitude distribution neighborhood characteristic maps of the trenching soil state in the sequence of the shallow characteristic gradient relative amplitude distribution neighborhood characteristic maps of the trenching soil state is divided by the number statistics of the shallow characteristic gradient amplitude distribution neighborhood of the trenching soil state according to the position points to obtain the local significant distribution characteristic map of the shallow characteristic gradient amplitude of the trenching soil state.

[0037] Specifically, the shallow characteristic map of the trenching soil state is input into the feature enhancement module based on the feature gradient distribution and processed with the following gradient significant feature enhancement formula to obtain the shallow significant feature map of the trenching soil state; Among them, the gradient significant feature enhancement formula is:

[0038] in, , , , , and They are respectively , , , , and The eigenvalues ​​of the positions, The first The gradient amplitude value at each position, is the first The gradient amplitude value at each position, is the first The gradient amplitude value at each position, The shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state is The eigenvalues ​​at the positions, The shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state is The eigenvalues ​​at the positions, The range is Neighborhood of express The number of elements in , The first one represents the local significant distribution characteristic map of shallow characteristic gradient amplitude of trench soil state The eigenvalues ​​at the positions, represents the Gaussian error linear gating function, The local significant gated mask feature map of shallow characteristic gradient amplitude of trenching soil state is shown in the first The eigenvalues ​​at the positions, is the local significant gated mask feature map of the shallow characteristic gradient amplitude of the trenching soil state, is the shallow characteristic map of the trenching soil state, It means point multiplication by position. This is a diagram of the shallow significant characteristics of the trenching soil state.

[0039] Furthermore, it is considered that the shallow significant feature map of the trenching soil state and the deep feature map of the trenching soil state respectively contain feature information of different levels and aspects about the trenching soil state. The shallow significant feature map of the trenching soil state emphasizes the shallow features such as texture, edge and color distribution that are more obvious in the trenching soil state, while the deep feature map focuses more on the deep features and semantic information that are more critical for identifying the trenching soil state. Therefore, in order to effectively integrate the shallow features and deep features of the trenching soil state, so as to better perceive and understand the trenching soil state, so as to provide a more complete and detailed soil state representation to provide a basis for irrigation control, in the technical solution of the present application, the shallow significant feature map of the trenching soil state and the deep feature map of the trenching soil state are further input into the feature significant joint perception network to obtain the shallow-deep joint significant perception feature map of the trenching soil state.

[0040] In one embodiment of the present application, the shallow significant feature map of the trenching soil state and the deep feature map of the trenching soil state are input into a feature significant joint perception network to obtain a shallow-deep joint significant perception feature map of the trenching soil state, including: reshaping the shallow significant feature map of the trenching soil state and the deep feature map of the trenching soil state to obtain a shallow significant feature shape reshaping matrix of the trenching soil state and a deep feature shape reshaping matrix of the trenching soil state; inputting the shallow significant feature shape reshaping matrix of the trenching soil state and the deep feature shape reshaping matrix of the trenching soil state into a feature channel-by-channel interactive perception module to obtain a shallow-deep feature dependency matrix of the trenching soil state and a deep-shallow feature dependency matrix of the trenching soil state; inputting the shallow-deep feature dependency matrix of the trenching soil state and the deep-shallow feature dependency matrix of the trenching soil state into a random dropout module to obtain a pruning optimized shallow-deep feature dependency matrix of the trenching soil state and a pruning optimized trenching soil state. A ditch soil state deep-shallow feature dependency matrix; matrix multiplying the pruning optimized ditch soil state shallow-deep feature dependency matrix with the ditch soil state deep feature shape reshaping matrix to obtain a dependency optimized ditch soil state deep-shallow feature dependency matrix; multiplying the pruning optimized ditch soil state deep-shallow feature dependency matrix with the ditch soil state shallow significant feature shape reshaping matrix to obtain a dependency optimized ditch soil state shallow-deep feature dependency matrix; feature shape reshaping of the dependency optimized ditch soil state deep-shallow feature dependency matrix and the dependency optimized ditch soil state shallow-deep feature dependency matrix to obtain an optimized ditch soil state shallow significant feature map and an optimized ditch soil state deep feature map; calculating the weighted sum of the optimized ditch soil state shallow significant feature map and the optimized ditch soil state deep feature map to obtain the ditch soil state shallow-deep joint significant perception feature map.

[0041] In particular, the feature saliency joint perception network is a feature fusion technology that enhances the interaction between the shallow salient feature map of the trenching soil state and the deep feature map of the trenching soil state through multiple steps, and finally obtains a shallow-deep joint salient perception feature map of the trenching soil state that combines the advantages of both. Specifically, first, the feature matrix along the channel dimension in the input shallow salient feature map of the trenching soil state and the deep feature map of the trenching soil state is expanded into a feature vector, which is convenient for the subsequent channel-by-channel interactive perception operation. Then, the dependency relationship between the two groups of feature vectors of the trenching soil state shallow feature and the deep feature is calculated to obtain the trenching soil state shallow-deep feature and trenching soil state deep-shallow feature dependency matrix. Then, the dependency matrix is ​​sparsely processed by pruning operation to enhance the generalization ability of the model. Then, the pruned dependency matrix is ​​used to weight the original features to strengthen the features most relevant to the trenching soil state perception and recognition tasks, while suppressing unimportant features. Finally, the optimized feature matrix is ​​reshaped into a feature map form, and the weighted sum is calculated to obtain the shallow-deep joint perception feature map of the trenching soil state. In this way, the intuitive shallow information in the shallow salient feature map of the trenching soil state and the deep semantic information in the deep feature map can be fully utilized to form a more comprehensive and rich feature representation. In addition, the salient joint perception module based on the attention mechanism can strengthen the key information in the shallow salient feature map of the trenching soil state and the deep feature map of the trenching soil state through the dependency matrix, while removing the redundant parts, further improving the feature quality, and improving the model's sensitivity and generalization ability to the input data, so that the model is more accurate and efficient when performing trenching soil state detection and subsequent irrigation control tasks.

[0042] In one embodiment of the present application, the trenching soil state shallow layer significant feature shape reshaping matrix and the trenching soil state deep layer feature shape reshaping matrix are input into the feature channel-by-channel interactive perception module to obtain the trenching soil state shallow-deep layer feature dependency matrix and the trenching soil state deep-shallow layer feature dependency matrix, including: performing matrix multiplication on the transposed matrix of the trenching soil state shallow layer significant feature shape reshaping matrix and the trenching soil state deep layer feature shape reshaping matrix, and then dividing the matrix by the scale of the trenching soil state shallow-deep layer feature dependency matrix to obtain the trenching soil state shallow-deep layer feature interaction matrix; performing matrix multiplication on the transposed matrix of the trenching soil state shallow layer significant feature shape reshaping matrix and the trenching soil state deep layer feature shape reshaping matrix, and then dividing the matrix by the scale of the trenching soil state deep layer feature shape reshaping matrix to obtain the trenching soil state deep-shallow layer feature interaction matrix; using The trenching soil state shallow-deep layer feature interaction matrix and the trenching soil state deep-shallow layer feature interaction matrix are respectively normalized to obtain the trenching soil state shallow-deep layer feature dependency matrix and the trenching soil state deep-shallow layer feature dependency matrix.

[0043] Specifically, the shallow significant feature map of the trenching soil state and the deep feature map of the trenching soil state are input into the feature significant joint perception network for processing using the following significant feature joint perception formula to obtain the shallow-deep joint significant perception feature map of the trenching soil state; The joint perception formula of the salient features is:

[0044] in, This is the shallow significant characteristic diagram of the trenching soil state, is the deep characteristic map of the trenching soil state, represents the feature shape reshaping process, and They are respectively the shallow significant characteristic shape reshaping matrix of the trenched soil state and the deep characteristic shape reshaping matrix of the trenched soil state. The scale of the shallow significant characteristic shape remodeling matrix of the trenching soil state and the deep characteristic shape remodeling matrix of the trenching soil state, for function, is matrix multiplication, and They are respectively the shallow-deep characteristic dependency matrix of trenching soil state and the deep-shallow characteristic dependency matrix of trenching soil state, represents pruning optimization processing, and They are respectively the shallow-deep characteristic dependency matrix of soil state for pruning optimization and ditching and the deep-shallow characteristic dependency matrix of soil state for pruning optimization and ditching, and They are respectively the dependency relationship optimization trenching soil state shallow-deep characteristic dependency relationship matrix and the dependency relationship optimization trenching soil state deep-shallow characteristic dependency relationship matrix. and They are respectively the shallow significant feature map of optimized trenching soil state and the deep feature map of optimized trenching soil state. and are all weighted hyperparameters, It is the shallow-deep joint significant perception feature map of the trenching soil state.

[0045] Then, the shallow-deep joint significant perception feature map of the ditching soil state is input into the decoder-based intelligent irrigation control module to obtain an irrigation control instruction, which is used to represent the recommended irrigation amount decoding value. In other words, the joint significant perception semantics between the shallow and deep features of the ditching soil state is used for decoding regression, so that irrigation instructions for specific soil states are generated through decoding by the intelligent irrigation control module to perform irrigation control. In this way, the irrigation amount can be automatically adjusted based on the actual ditching soil state to avoid the problem of salt accumulation caused by excessive irrigation, thereby achieving more intelligent and accurate water management, optimizing the utilization efficiency of water resources, and promoting the ecological restoration of saline-alkali land and sustainable agricultural development.

[0046] In one embodiment of the present application, an irrigation control instruction is generated based on the shallow-deep joint significant perceptual feature map of the trenching soil state, and the irrigation control instruction is used to represent the recommended irrigation amount decoded value, including: inputting the shallow-deep joint significant perceptual feature map of the trenching soil state into a decoder-based intelligent irrigation control module to obtain an irrigation control instruction, and the irrigation control instruction is used to represent the recommended irrigation amount decoded value.

[0047] In a preferred example, considering that the shallow salient feature map of the trenched soil state and the deep feature map of the trenched soil state respectively represent the shallow image features and deep image features determined by convolutional coding at different depths of the trenched soil state image, when performing attention-driven feature joint saliency perception on them, the difference in attention weights caused by the difference in feature depth and receptive field expressed in the spatial domain of image semantic features will introduce image semantic joint perception aggregation offset, making the joint perception interaction distribution diversity of the shallow-deep joint salient perception feature map of the trenched soil state across the depth of image semantic features. Therefore, when the shallow-deep joint salient perception feature map of the trenched soil state is decoded by a decoder, it will affect the accuracy of the decoding result.

[0048] Therefore, considering that the weight matrix of the decoder acts on the shallow-deep joint salient perceptual feature vector of the trenched soil state after the shallow-deep joint salient perceptual feature map of the trenched soil state is expanded, the distribution diversity of the shallow-deep joint salient perceptual feature map of the trenched soil state causes the uncertainty of the weight stimulus parameters of the weight matrix of the decoder, resulting in the lack of posterior probability density inference of the shallow-deep joint salient perceptual feature map of the trenched soil state through the action of the weight matrix, which will affect the accuracy of the decoding results.

[0049] Preferably, inputting the shallow-deep joint significant perceptual feature map of the trenching soil state into a decoder-based intelligent irrigation control module to obtain an irrigation control instruction comprises: Determine the characteristic mean and characteristic variance of the shallow-deep joint significant perception characteristic map of the trenching soil state, and divide the characteristic mean by the characteristic variance to obtain the shallow-deep joint significant perception probability statistic of the trenching soil state:

[0050] in, and represent the feature mean and feature variance of the shallow-deep joint significant perceptual feature map of the trenching soil state, respectively, It represents the statistical value of the joint significant perception probability of shallow-deep soil state in trenching.

[0051] The shallow-deep joint significant perception feature map of the ditching soil state is point-multiplied with the inverse of the maximum eigenvalue in the shallow-deep joint significant perception feature map of the ditching soil state to obtain the shallow-deep joint significant perception probability constraint map of the ditching soil state:

[0052] in, The shallow-deep joint significant perceptual feature map representing the trenching soil state, represents the maximum eigenvalue in the shallow-deep joint significant perceptual feature map of the trenching soil state, It means point multiplication by position. Represents the shallow-deep joint saliency perception probability constraint map of trenching soil state.

[0053] After adding the shallow-deep joint significant perception probability constraint graph of the ditching soil state and the shallow-deep joint significant perception probability statistical value of the ditching soil state, the logarithmic value with base 2 is calculated to obtain the shallow-deep joint significant perception information interaction graph of the ditching soil state:

[0054] in, Represents the shallow-deep joint saliency perception probability constraint map of trenching soil state, It represents the statistical value of the shallow-deep joint significant perception probability of the trenching soil state. Indicates adding by position point. A diagram showing the interaction of shallow-deep joint salient perceptual information of trenching soil status.

[0055] After point-subtracting the shallow-deep joint significant perception probability constraint map of the ditching soil state with the shallow-deep joint significant perception probability statistics of the ditching soil state, the inverse of each eigenvalue is calculated to obtain the shallow-deep joint significant perception sequence constraint map of the ditching soil state:

[0056] in, Represents the shallow-deep joint saliency perception probability constraint map of trenching soil state, It represents the statistical value of the shallow-deep joint significant perception probability of the trenching soil state. It means to subtract by position point. The shallow-deep joint significant perception sequence constraint diagram of trenching soil state is shown.

[0057] The shallow-deep joint significant perception information interaction graph of the ditching soil state is interpolated with the shallow-deep joint significant perception sequence constraint graph of the ditching soil state to obtain an optimized shallow-deep joint significant perception feature graph of the ditching soil state; The optimized shallow-deep joint significant perception feature map of trenching soil state is input into a decoder-based intelligent irrigation control module to obtain irrigation control instructions.

[0058] Accordingly, the probability statistical characteristics of the shallow-deep joint significant perception feature map of the ditching soil state are used to simulate the mesoscale interaction structure under probability constraints between the feature value scale and the feature map scale of the shallow-deep joint significant perception feature map of the ditching soil state, so as to construct a bidirectional latent variable motif based on the mesoscale short sequence relative to the probability statistical value to perform mesoscale bidirectional migration, and perform posterior recovery based on the sequence constraint on the interactive information, thereby improving the convergence effect in the probability density domain and improving the accuracy of the irrigation control instructions obtained by the decoder-based intelligent irrigation control module of the shallow-deep joint significant perception feature map of the ditching soil state. In this way, the irrigation amount can be adjusted more accurately based on the actual ditching soil state, avoiding the problem of salt accumulation caused by excessive irrigation, thereby achieving more intelligent and precise water management, optimizing the utilization efficiency of water resources, and promoting the ecological restoration of saline-alkali land and sustainable agricultural development.

[0059] In summary, the above scheme is adopted to control the trenching machine to trench the soil to obtain the soil trench, and collect the trenching soil state image of the soil trench through the camera, and then introduce the image processing and analysis algorithm based on artificial intelligence and deep learning in the back end to analyze the trenching soil state image, so as to capture and characterize the semantic representation of the soil state of the trench in the image, so as to perform corresponding irrigation control to generate irrigation instructions for specific soil states. In this way, the irrigation amount can be automatically adjusted based on the actual trenching soil state to avoid the problem of salt accumulation caused by excessive irrigation, thereby achieving more intelligent and precise water management, optimizing the utilization efficiency of water resources, and promoting the ecological restoration of saline-alkali land and sustainable agricultural development.

[0060] Figure 3FIG. 1 is a block diagram of a saline-alkali soil microecological improvement system based on a trenching machine according to an exemplary embodiment. Figure 3 As shown, the system 200 includes: A soil trenching module 201 is used to control a trenching machine to perform soil trenching to obtain a soil trench; A trenching soil state image acquisition module 202 is used to acquire a trenching soil state image of the soil trench acquired by a camera; A multi-scale feature extraction module 203 is used to perform multi-scale feature extraction based on soil state on the trenching soil state image to obtain a trenching soil state shallow layer feature map and a trenching soil state deep layer feature map; A gradient mask feature enhancement module 204 is used to input the shallow feature map of the trenching soil state into a feature enhancement module based on feature gradient distribution to obtain a shallow significant feature map of the trenching soil state; A feature joint saliency perception module 205 is used to input the shallow saliency feature map of the trenching soil state and the deep saliency feature map of the trenching soil state into a feature saliency joint perception network to obtain a shallow-deep joint saliency perception feature map of the trenching soil state; The irrigation control instruction generation module 206 is used to generate an irrigation control instruction based on the shallow-deep joint significant perception feature map of the trenching soil state, and the irrigation control instruction is used to represent the recommended irrigation amount decoding value.

[0061] 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.

[0062] like Figure 4 As 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.

[0063] 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.

[0064] 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.

[0065] 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.

[0066] 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.

[0067] The computer-readable medium may be included in the electronic device, or may exist independently without being installed in the electronic device.

[0068] 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).

[0069] 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.

[0070] 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".

[0071] 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.

[0072] 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.

[0073] Figure 5 1 is an application scenario diagram of a method for improving the microecology of saline-alkali soil based on a trenching machine according to an exemplary embodiment. Figure 5 As shown, in this application scenario, first, the trenching machine is controlled to trench the soil to obtain a soil trench (for example, Figure 5 C1) and obtaining the trench soil state image of the soil trench collected by the camera (for example, as shown in Figure 5 Then, the obtained soil trench and trench soil state images are input to a server (for example, Figure 5 In S) shown in the figure, the server is capable of processing the soil trench and the trenching soil status image based on the saline-alkali soil microecological improvement algorithm of the trenching machine to generate an irrigation control instruction, and the irrigation control instruction is used to represent the recommended irrigation amount decoded value.

[0074] 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.

[0075] 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 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.

[0076] 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. A method for improving soil microecology in saline-alkali land based on a trenching machine, characterized in that: include: Controlling a trenching machine to trench the soil to obtain a soil trench; Acquire a trench soil state image of the soil trench collected by a camera; Performing multi-scale feature extraction based on soil state on the trenching soil state image to obtain a trenching soil state shallow layer feature map and a trenching soil state deep layer feature map; Inputting the shallow characteristic map of the trenching soil state into a feature enhancement module based on feature gradient distribution to obtain a shallow significant characteristic map of the trenching soil state; Inputting the shallow significant feature map of the trenching soil state and the deep feature map of the trenching soil state into a feature significant joint perception network to obtain a shallow-deep joint significant perception feature map of the trenching soil state; Based on the shallow-deep joint significant perception feature map of the trenching soil state, an irrigation control instruction is generated, and the irrigation control instruction is used to represent a decoded value of the recommended irrigation amount.

2. The method for improving soil microecology in saline-alkali land based on a trenching machine according to claim 1, characterized in that: The trenching soil state image is subjected to multi-scale feature extraction based on soil state to obtain a trenching soil state shallow feature map and a trenching soil state deep feature map, including: inputting the trenching soil state image into a soil state multi-scale feature extractor based on a hollow pyramid network to obtain the trenching soil state shallow feature map and the trenching soil state deep feature map.

3. The method for improving soil microecology in saline-alkali land based on a trenching machine according to claim 2, characterized in that: The shallow characteristic map of the trenching soil state is input into a feature enhancement module based on feature gradient distribution to obtain a shallow significant characteristic map of the trenching soil state, including: Calculating the multi-directional gradient value distribution of each position in the shallow characteristic map of the trenching soil state, and determining the gradient amplitude value of each position in the shallow characteristic map of the trenching soil state based on the multi-directional gradient value distribution of each position to obtain a shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state; Calculating the local description operator of the gradient amplitude at each position in the shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state to obtain a local significant distribution characteristic map of the shallow characteristic gradient amplitude of the trenching soil state; Inputting the shallow characteristic gradient amplitude local significant distribution characteristic map of the trenching soil state into a gated masker based on the GELU function to obtain the shallow characteristic gradient amplitude local significant gated mask characteristic map of the trenching soil state; The shallow characteristic gradient amplitude local significant gated mask feature map of the trenched soil state is calculated and multiplied by the position points between the shallow characteristic map of the trenched soil state to obtain the shallow significant feature map of the trenched soil state.

4. The method for improving soil microecology in saline-alkali land based on a trenching machine according to claim 3, characterized in that: Calculating the multi-directional gradient value distribution of each position in the shallow characteristic map of the trenching soil state, and determining the gradient amplitude value of each position in the shallow characteristic map of the trenching soil state based on the multi-directional gradient value distribution of each position to obtain the shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state, including: Calculate the gradient amplitude value of each position characteristic value in the shallow characteristic map of the trenching soil state along the height direction to obtain a shallow characteristic height gradient distribution characteristic map of the trenching soil state; Calculate the gradient amplitude value of each position characteristic value in the trenching soil state shallow characteristic map along the width direction to obtain the trenching soil state shallow characteristic width gradient distribution characteristic map; Calculating the gradient amplitude value of each position characteristic value in the shallow characteristic map of the trenching soil state along the channel direction to obtain a gradient distribution characteristic map of the shallow characteristic channel of the trenching soil state; The square root of the sum of squares of the characteristic values ​​at corresponding positions in the shallow characteristic height gradient distribution characteristic map of the trenching soil state, the shallow characteristic width gradient distribution characteristic map of the trenching soil state and the shallow characteristic channel gradient distribution characteristic map of the trenching soil state is calculated to obtain the shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state.

5. The method for improving soil microecology in saline-alkali land based on a trenching machine according to claim 4, characterized in that: Calculating the local description operator of the gradient amplitude at each position in the shallow characteristic gradient amplitude distribution characteristic map of the trenching soil state to obtain the local significant distribution characteristic map of the shallow characteristic gradient amplitude of the trenching soil state, including: Determine a local neighborhood range of the gradient amplitude, and calculate a sequence of neighborhood characteristic maps of shallow characteristic gradient amplitude distribution of trenching soil state based on the local neighborhood range of the gradient amplitude and the shallow characteristic gradient amplitude distribution characteristic map of trenching soil state; Counting the number of characteristic graphs in the sequence of the neighborhood characteristic graphs of the shallow characteristic gradient amplitude distribution of the ditching soil state to obtain the neighborhood number statistics of the shallow characteristic gradient amplitude distribution of the ditching soil state; Calculate the position difference between each shallow layer characteristic gradient amplitude distribution neighborhood characteristic graph of the trenching soil state in the sequence of the shallow layer characteristic gradient amplitude distribution characteristic graph of the trenching soil state and the shallow layer characteristic gradient amplitude distribution neighborhood characteristic graph of the trenching soil state, and divide the difference by position point with the shallow layer characteristic gradient amplitude distribution characteristic graph of the trenching soil state to obtain a sequence of shallow layer characteristic gradient relative amplitude distribution neighborhood characteristic graph of the trenching soil state; The relative amplitude distribution neighborhood characteristic map of the shallow characteristic gradient of the trenching soil state in the sequence of the relative amplitude distribution neighborhood characteristic map of the shallow characteristic gradient of the trenching soil state is added at each position point, and then the result is divided by the statistical value of the neighborhood number of the shallow characteristic gradient amplitude distribution of the trenching soil state at each position point to obtain the local significant distribution characteristic map of the shallow characteristic gradient amplitude of the trenching soil state.

6. The method for improving soil microecology in saline-alkali land based on a trenching machine according to claim 5, characterized in that: Inputting the shallow significant feature map of the trenching soil state and the deep feature map of the trenching soil state into a feature significant joint perception network to obtain a shallow-deep joint significant perception feature map of the trenching soil state, including: Performing feature shape reshaping on the shallow significant feature map of the trenching soil state and the deep feature map of the trenching soil state to obtain a trenching soil state shallow significant feature shape reshaping matrix and a trenching soil state deep feature shape reshaping matrix; Input the trenching soil state shallow layer significant feature shape reshaping matrix and the trenching soil state deep layer feature shape reshaping matrix into the feature channel-by-channel interactive perception module to obtain the trenching soil state shallow layer-deep layer feature dependency matrix and the trenching soil state deep layer-shallow layer feature dependency matrix; Input the trenching soil state shallow-deep layer feature dependency matrix and the trenching soil state deep-shallow layer feature dependency matrix into a random dropout module to obtain a pruning-optimized trenching soil state shallow-deep layer feature dependency matrix and a pruning-optimized trenching soil state deep-shallow layer feature dependency matrix; The pruning optimization trenching soil state shallow-deep layer characteristic dependency relationship matrix is ​​matrix-multiplied with the trenching soil state deep layer characteristic shape reshaping matrix to obtain the dependency optimization trenching soil state deep-shallow layer characteristic dependency relationship matrix; The pruning optimization trenching soil state deep-shallow layer characteristic dependency matrix is ​​multiplied by the trenching soil state shallow layer significant characteristic shape reshaping matrix to obtain the dependency optimization trenching soil state shallow-deep layer characteristic dependency matrix; Reshaping the deep-shallow characteristic dependency matrix of the dependency relationship optimized trenching soil state and the shallow-deep characteristic dependency matrix of the dependency relationship optimized trenching soil state to obtain a shallow significant characteristic map of the optimized trenching soil state and a deep characteristic map of the optimized trenching soil state; The weighted sum of the optimized shallow salient feature map of the trenching soil state and the optimized deep feature map of the trenching soil state is calculated to obtain the shallow-deep joint salient perceptual feature map of the trenching soil state.

7. The method for improving soil microecology in saline-alkali land based on a trenching machine according to claim 6, characterized in that: The trenching soil state shallow layer significant feature shape reshaping matrix and the trenching soil state deep layer feature shape reshaping matrix are input into the feature channel-by-channel interactive perception module to obtain the trenching soil state shallow layer-deep layer feature dependency matrix and the trenching soil state deep layer-shallow layer feature dependency matrix, including: The transposed matrix of the shallow significant characteristic shape reshaping matrix of the trenching soil state and the deep characteristic shape reshaping matrix of the trenching soil state are multiplied, and then divided by the scale of the trenching soil state shallow-deep characteristic dependency matrix by position to obtain the trenching soil state shallow-deep characteristic interaction matrix; The transposed matrix of the shallow significant characteristic shape remodeling matrix of the trenching soil state and the deep characteristic shape remodeling matrix of the trenching soil state are matrix-multiplied, and then the matrix is ​​divided by the scale of the deep characteristic shape remodeling matrix of the trenching soil state by position to obtain the deep-shallow characteristic interaction matrix of the trenching soil state; The shallow-deep feature interaction matrix of the trenching soil state and the deep-shallow feature interaction matrix of the trenching soil state are normalized by Softmax to obtain the shallow-deep feature dependency matrix of the trenching soil state and the deep-shallow feature dependency matrix of the trenching soil state.

8. The method for improving soil microecology in saline-alkali land based on a trenching machine according to claim 7, characterized in that: Based on the shallow-deep joint significant perception feature map of the trenching soil state, an irrigation control instruction is generated, and the irrigation control instruction is used to represent the recommended irrigation amount decoding value, including: inputting the shallow-deep joint significant perception feature map of the trenching soil state into a decoder-based intelligent irrigation control module to obtain an irrigation control instruction, and the irrigation control instruction is used to represent the recommended irrigation amount decoding value.

Images