High-standard farmland remote sensing recognition method, model training method and related device

Abstract

The invention provides a high-standard farmland remote sensing identification method, a model training method and a related device, and relates to the technical field of high-standard farmland identification, and the method comprises the steps: obtaining first multi-feature index data of each first farmland pixel of a first preset region, and the first multi-feature index data comprises the values of the corresponding first farmland pixel on a plurality of preset indexes; based on an entropy weight TOPSIS method, according to the normalized first multi-feature index data, obtaining a comprehensive evaluation index of each first farmland pixel for reflecting the conformity of the corresponding farmland pixel and the high-standard farmland; according to the comprehensive evaluation index, obtaining an initial classification result of each first farmland pixel; training a random forest model by taking the target data of the first farmland pixel passing the authenticity verification as a model training sample to obtain a high-standard farmland remote sensing recognition model; the target data comprises first multi-feature index data corresponding to the target first farmland pixel and an actual farmland type. According to the invention, refined remote sensing identification of high-standard farmland can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of high-standard farmland identification, and in particular to a high-standard farmland remote sensing identification method, a model training method and related devices. Background Art

[0002] The definition of high-standard farmland is "flat fields, concentrated and contiguous, well-equipped, water-saving and efficient, with supporting agricultural electricity, suitable for mechanized operations, fertile soil, ecological friendliness, strong disaster resistance, and arable land that can ensure stable and high yields in drought and flood, suitable for modern agricultural production and management methods."

[0003] In the existing high-standard farmland identification scheme, a simple model is used for high-standard farmland identification. However, since the relationship between the various indicators of farmland characteristics is not a linear relationship described by a simple model, but a complex relationship, the existing technology has the problem of difficulty in realizing the refined identification of high-standard farmland.

[0004] Based on this, how to achieve refined identification of high-standard farmland has become a technical problem that needs to be solved urgently. Summary of the invention

[0005] In view of this, in order to solve the above technical problems, the present invention provides a high-standard farmland remote sensing identification method, a model training method and related devices.

[0006] The present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a training method for a high-standard farmland remote sensing recognition model, comprising:

[0008] Acquire first multi-feature index data of each first farmland pixel in a first preset area; the first multi-feature index data includes values ​​of the first farmland pixel on a plurality of preset indicators;

[0009] Normalizing the first plurality of feature indicator data;

[0010] Based on the entropy weight TOPSIS method, a comprehensive evaluation index of each of the first farmland pixels is obtained according to the normalized first multi-feature index data; the comprehensive evaluation index is used to reflect the conformity of the corresponding farmland pixel with the high-standard farmland;

[0011] performing initial classification on each of the first farmland pixels according to the comprehensive evaluation index of each of the first farmland pixels to obtain an initial classification result of each of the first farmland pixels;

[0012] Verify the authenticity of the initial classification results of the first target farmland pixel;

[0013] The target data of the target first farmland pixel that has passed the authenticity verification is used as a model training sample to train a random forest model to obtain a high-standard farmland remote sensing identification model; the target data includes the first multi-feature indicator data corresponding to the target first farmland pixel and the actual farmland type.

[0014] Optionally, the obtaining of the first multi-feature indicator data of each first farmland pixel in the first preset area specifically includes:

[0015] Acquire high spatial resolution satellite remote sensing images, high spectral resolution satellite remote sensing images and digital elevation models of the first preset area;

[0016] The first multi-feature indicator data of each of the first farmland pixels are obtained according to the high spatial resolution satellite remote sensing image, the high spectral resolution satellite remote sensing image and the digital elevation model.

[0017] Optionally, the preset indicators include: field slope, field connectivity, field fragmentation, soil organic matter content, soil pH value, soil moisture index, soil degradation index, crop growth conditions, distance between the field and the nearest irrigation ditch, field ditch density, field road accessibility, distance from the field to residential areas, distance from the field to traffic arteries, forest network density, shelterbelt coverage and the distance from the field to its nearest transmission tower.

[0018] Optionally, performing initial classification on each of the first farmland pixels according to the comprehensive evaluation index of each of the first farmland pixels to obtain an initial classification result of each of the first farmland pixels specifically includes:

[0019] For each of the first farmland pixels, based on a preset correspondence between an index threshold range and a farmland type, the farmland type of the first farmland pixel is determined according to a comprehensive evaluation index of the first farmland pixel.

[0020] Optionally, the farmland types include: high-standard farmland and other farmland.

[0021] In a second aspect, the present invention provides a high-standard farmland remote sensing identification method, comprising:

[0022] Acquire second multi-feature index data of each second farmland pixel in a second preset area; the second multi-feature index data includes values ​​corresponding to the second farmland pixel on a plurality of preset indicators;

[0023] Based on the high-standard farmland remote sensing identification model, the farmland type of each of the second farmland pixels is obtained according to the second multi-feature indicator data; the high-standard farmland remote sensing identification model is trained using the training method of the high-standard farmland remote sensing identification model as described above, and the farmland types include high-standard farmland and other farmland.

[0024] In a third aspect, the present invention provides a training device for a high-standard farmland remote sensing identification model, comprising:

[0025] A first acquisition module is used to acquire first multi-feature index data of each first farmland pixel in a first preset area; the first multi-feature index data includes values ​​of the first farmland pixel on multiple preset indicators respectively;

[0026] A normalization module, used for normalizing the first plurality of feature indicator data;

[0027] A calculation module, for obtaining a comprehensive evaluation index of each of the first farmland pixels based on the normalized first multi-feature index data based on an entropy weight TOPSIS method; the comprehensive evaluation index is used to reflect the conformity of the corresponding farmland pixel with the high-standard farmland;

[0028] An initial classification module, used for performing initial classification on each of the first farmland pixels according to the comprehensive evaluation index of each of the first farmland pixels, and obtaining an initial classification result of each of the first farmland pixels;

[0029] A verification module is used to verify the authenticity of the initial classification result of the target first farmland pixel;

[0030] A training module is used to train a random forest model using the target data of the target first farmland pixel that has passed authenticity verification as a model training sample to obtain a high-standard farmland remote sensing identification model; the target data includes the first multi-feature indicator data corresponding to the target first farmland pixel and the actual farmland type.

[0031] In a fourth aspect, the present invention provides a high-standard farmland remote sensing identification device, comprising:

[0032] A second acquisition module is used to acquire second multi-feature index data of each second farmland pixel in a second preset area; the second multi-feature index data includes values ​​corresponding to the second farmland pixel on multiple preset indicators;

[0033] A classification module is used to obtain the farmland type of each of the second farmland pixels based on the high-standard farmland remote sensing identification model and the second multi-feature indicator data; the high-standard farmland remote sensing identification model is trained using the training method of the high-standard farmland remote sensing identification model as described above, and the farmland types include high-standard farmland and other farmland.

[0034] The present invention adopts the above technical scheme, a training method for a high-standard farmland remote sensing identification model, including: obtaining first multi-feature indicator data of each first farmland pixel in a first preset area; the first multi-feature indicator data includes the values ​​of the corresponding first farmland pixels on multiple preset indicators; the first multi-feature indicator data is normalized; based on the entropy weight TOPSIS method, the comprehensive evaluation index of each first farmland pixel is obtained according to the first multi-feature indicator data after normalization; the comprehensive evaluation index is used to reflect the conformity of the corresponding farmland pixel with the high-standard farmland; according to the comprehensive evaluation index of each first farmland pixel, each first farmland pixel is initially classified to obtain the initial classification result of each first farmland pixel; the initial classification result of the target first farmland pixel is verified for authenticity; the target data of the target first farmland pixel that has passed the authenticity verification is used as a model training sample to train a random forest model to obtain a high-standard farmland remote sensing identification model; the target data includes the first multi-feature indicator data corresponding to the target first farmland pixel and the actual farmland type.

[0035] Based on this, by adopting the entropy weight TOPSIS method and the random forest model, the present invention can reasonably interpret the mutual influence and connection between the preset indicators, thereby enabling the present invention to achieve refined identification of high-standard farmland. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0037] Figure 1 It is a flow chart of a training method for a high-standard farmland remote sensing identification model provided by an embodiment of the present invention;

[0038] Figure 2 It is a flow chart of a high-standard farmland remote sensing identification method provided by an embodiment of the present invention;

[0039] Figure 3 It is a structural schematic diagram of a training device for a high-standard farmland remote sensing identification model provided by an embodiment of the present invention;

[0040] Figure 4 It is a structural schematic diagram of a high-standard farmland remote sensing identification device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0041] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0042] Example

[0043] Figure 1 FIG. 1 is a flow chart of a training method for a high-standard farmland remote sensing recognition model provided by an embodiment of the present invention. Figure 1 As shown, the training method of this high-standard farmland remote sensing recognition model includes:

[0044] Step 101: obtaining first multi-feature index data of each first farmland pixel in a first preset area; the first multi-feature index data includes values ​​of the corresponding first farmland pixels on multiple preset indicators.

[0045] Specifically, the present invention is based on the principles of pertinence, practicality, timeliness and verifiability, and focuses on the objective conditions that high-standard farmland should have in six categories, namely, field, soil, water, road, forest and electricity, to determine a multi-feature index system for remote sensing identification of high-standard farmland, with a total of 16 preset indicators, namely: field slope X1, field contiguousness X2, field fragmentation X3, soil organic matter content X4, soil pH value X5, soil moisture index X6, soil degradation index X7, crop growth status X8, distance between the field and the nearest irrigable ditch X9, field ditch density X10, field slope X11, field contiguousness X2, field fragmentation X3, soil organic matter content X4, soil pH value X5, soil moisture index X6, soil degradation index X7, crop growth status X8, distance between the field and the nearest irrigable ditch X9, field ditch density X11, field slope X12, field slope X13, field slope X14, field slope X15, field slope X16, field slope X17, field slope X18, field slope X19, field slope X20, field slope X21 10 , Field road accessibility X 11 , the distance from the field to the settlement X 12 , Distance from field to traffic artery X 13 , forest network density X 14 , Protection forest coverage rate X 15 The distance from the hotan block to the nearest transmission tower is X 16 Among them, the soil degradation index X7 is specifically the ratio vegetation index, and the crop growth condition X8 is specifically the normalized vegetation index.

[0046] The criterion layers and multi-feature representations corresponding to each preset indicator are shown in Table 1 below:

[0047] Table 1

[0048]

[0049] In addition, among the 16 preset indicators, the data source of field slope X1 is DEM (Digital Elevation Model). Except for field slope X1, the data sources of other preset indicators are high spatial resolution satellite remote sensing images and high spectral resolution satellite remote sensing images.

[0050] Based on this, in the embodiment of the present invention, obtaining the first multi-feature index data of each first farmland pixel in the first preset area may specifically include:

[0051] A high spatial resolution satellite remote sensing image, a high spectral resolution satellite remote sensing image and a digital elevation model of a first preset area are obtained, and then first multi-feature indicator data of each first farmland pixel are obtained based on the high spatial resolution satellite remote sensing image, the high spectral resolution satellite remote sensing image and the digital elevation model.

[0052] Specifically, the object-oriented classification method of ENVI multi-scale segmentation was used to extract remote sensing interpretation information (i.e., multi-feature representation in Table 1) of fields, roads, ditches, settlements, woodlands, and transmission towers from high spatial resolution satellite remote sensing images (from Gaofen-1 or Gaofen-2 satellites) and high spectral resolution satellite remote sensing images (from Gaofen-5 satellites). At the same time, the land use data was superimposed as a reference and combined with the panchromatic and multispectral image fusion data of Gaofen-2 satellite (resolution of 0.8m) to verify the accuracy of the classification results. Using ArcGIS slope analysis, kernel density analysis, fishing net analysis, proximity analysis and other functions, the field slope X1, field contiguousness X2, the distance between the field and the nearest irrigable ditch X9, and the field ditch density X1 were obtained. 10 , Field road accessibility X 11 , the distance from the field to the settlement X 12 , Distance from field to traffic artery X 13 , forest network density X 14 , Protection forest coverage rate X 15 The distance from the hotan block to the nearest transmission tower is X 16; Through raster and vector processing, we obtain identification indicators that reflect the site conditions of cultivated land (fields and soil), farming convenience, transportation convenience, high-standard farmland protection and power transmission convenience, namely: field fragmentation X3, soil moisture index X6, soil degradation index X7, crop growth status X8; Based on the first inversion model, according to the double-high remote sensing data (i.e., high spatial resolution satellite remote sensing images and high spectral resolution satellite remote sensing images), we obtain the soil organic matter content X4, where the first inversion model is to use the spectral characteristics of the remote sensing image (such as reflectivity) and the soil organic matter content (ground measured data of soil organic matter content) to establish an inversion model. Based on the second inversion model, according to the double-high remote sensing data (i.e., high spatial resolution satellite remote sensing images and high spectral resolution satellite remote sensing images), we obtain the soil pH value X5, where the second inversion model is a pre-established regression model between spectral reflectivity and pH value (soil pH value measured data).

[0053] It should be noted that the calculation methods of various preset indicators are all existing technologies.

[0054] Compared with the prior art, soil property indicators (soil organic matter content X4 and soil pH value X5) are usually obtained by spatial interpolation of soil census data, which have the problems of poor representativeness, long age, and insufficient reflection of spatial heterogeneity. In the present invention, based on the dual-high remote sensing data that can simultaneously obtain information on the spatial dimension and spectral dimension of the ground object, the relationship between optical properties and soil properties is established, and the inversion of soil organic matter and soil pH value parameters at the field scale is realized, so that the present invention can improve the accuracy and stability of soil organic matter content X4 and soil pH value X5.

[0055] Step 102: normalize the first multi-feature indicator data.

[0056] Specifically, in order to eliminate the influence of dimension, order of magnitude and attribute orientation on the optimal indicator set, standardization processing is used to uniformly map the first multi-feature indicator data to the range of 0 to 1, so as to realize the quantitative processing of the multi-feature indicator system for remote sensing identification of high-standard farmland.

[0057] Step 103: Based on the entropy weight TOPSIS method, a comprehensive evaluation index of each first farmland pixel is obtained according to the normalized first multi-feature index data; the comprehensive evaluation index is used to reflect the conformity of the corresponding farmland pixel with the high-standard farmland.

[0058] The specific implementation process of step 103 is as follows:

[0059] (1) Use the normalized first multi-feature indicator data to construct a decision matrix.

[0060] Here, the construction formula of the decision matrix is ​​as follows:

[0061]

[0062] Wherein, n is the number of samples (i.e., n=the number of the first farmland pixels); m is the number of preset indicators. For example, in the above example, the number of preset indicators is 16, so m=16; x ij It represents the value of the i-th first farmland pixel on the j-th preset index, i = 0, 1, 2, ..., n; j = 0, 1, 2, ..., m.

[0063] (2) Based on the entropy weight method, the weights of the preset indicators are calculated according to the decision matrix X.

[0064] (2.1) Calculate the ratio P ij .

[0065] Specifically, P ij The calculation formula is as follows:

[0066]

[0067] Among them, P ij It represents the relative weight of the i-th first farmland pixel on the j-th preset index.

[0068] (2.2) Calculate information entropy e j :

[0069] specific, e j The calculation formula is as follows:

[0070]

[0071] Among them, when P ij = 0, it is agreed that P ij ln(P ij )=0.

[0072] (2.3) Calculate the weight w j .

[0073] Specifically, w j The calculation formula is as follows:

[0074]

[0075] (3) Calculate the weighted normalization matrix.

[0076] Specifically, the weight w of each preset indicator is j Applied to the decision matrix X, we get the weighted normalization matrix V, using the following calculation formula:

[0077] V=wj·X......(5)

[0078] (4) Calculate the positive ideal solution V + and negative ideal solution V - .

[0079] Specifically, the positive ideal solution V + The calculation formula of (representing the optimal solution) is as follows:

[0080]

[0081] Negative ideal solution V - The calculation formula for (representing the worst solution) is as follows:

[0082]

[0083] (5) Calculate the distance between the first farmland pixel (i.e., sample) and the positive and negative ideal solutions.

[0084] Specifically, the first farmland pixel of the i-th field is consistent with the positive ideal solution V + The distance between The calculation formula is as follows:

[0085]

[0086] The i-th first farmland pixel and the negative ideal solution V - The distance between The calculation formula is as follows:

[0087]

[0088] (6) Use the TOPSIS method to calculate the comprehensive evaluation index of each first farmland pixel.

[0089] Specifically, the comprehensive evaluation index C of the i-th first farmland pixel is i The calculation formula is as follows:

[0090]

[0091] Step 104: performing initial classification on each first farmland pixel according to the comprehensive evaluation index of each first farmland pixel to obtain an initial classification result of each first farmland pixel.

[0092] In an embodiment of the present invention, the value range of the comprehensive evaluation index is divided in advance to obtain multiple index threshold ranges, and the farmland types corresponding to different index threshold ranges are set. For example, when the comprehensive evaluation index of the first farmland pixel is greater than a, and the comprehensive evaluation index is less than or equal to b, the farmland type of the first farmland pixel is determined to be high-standard farmland, otherwise (that is, the comprehensive evaluation index is less than or equal to a, or the comprehensive evaluation index is greater than b), the farmland type of the first farmland pixel is determined to be other farmland, so that the corresponding relationship between the index threshold range and the farmland type is obtained. Subsequently, after the comprehensive evaluation index of the first farmland pixel is actually obtained, the farmland type corresponding to the index threshold range to which the comprehensive evaluation index belongs is determined as the farmland type of the first farmland pixel.

[0093] Based on this, each first farmland pixel is initially classified according to the comprehensive evaluation index of each first farmland pixel, and the initial classification result of each first farmland pixel is obtained, which may specifically include:

[0094] For each first farmland pixel, based on the corresponding relationship between the preset index threshold range and the farmland type, the farmland type of the first farmland pixel is determined according to the comprehensive evaluation index of the first farmland pixel.

[0095] In the embodiment of the present invention, the farmland types may include: high-standard farmland and other farmland.

[0096] Step 105: verify the authenticity of the initial classification result of the target first farmland pixel.

[0097] Here, the authenticity of the initial classification result of the target first farmland pixel can be verified based on the actual survey result of the target first farmland pixel.

[0098] Step 106: Using the target data of the target first farmland pixel that has passed the authenticity verification as a model training sample, a random forest model is trained to obtain a high-standard farmland remote sensing identification model; the target data includes the first multi-feature indicator data corresponding to the target first farmland pixel and the actual farmland type.

[0099] Specifically, 80% of the target first farmland pixels that have passed authenticity verification can be selected, and the target data of these 80% of the target first farmland pixels can be used for model training, and 20% of the target first farmland pixels can be selected, and the target data of these 20% of the target first farmland pixels can be used for model verification.

[0100] The first multi-feature index data is used as the independent variable, the farmland type is used as the dependent variable, and the random forest machine learning algorithm is used to construct a high-standard farmland remote sensing recognition model. The final recognition result is determined by voting of K decision trees, and the accuracy is determined by the average out-of-bag data (OOB) error of each regression tree. The number of decision trees K and the depth of the decision tree N are optimized based on the principle of minimizing the OOB error, and finally the high-standard farmland remote sensing recognition model of this application is obtained.

[0101] The present invention adopts the above technical solution, and by adopting the entropy weight TOPSIS method and the random forest model, the present invention can reasonably interpret the mutual influence and connection between the preset indicators, thereby enabling the present invention to achieve refined identification of high-standard farmland. In addition, based on the dual-high remote sensing data that can simultaneously obtain information on the spatial dimension and spectral dimension of the ground object, the relationship between optical properties and soil properties is established, and the inversion of soil organic matter and soil pH parameters at the field scale is achieved, so that the present invention can improve the accuracy and stability of soil organic matter content X4 and soil pH value X5.

[0102] Based on a general inventive concept, the present invention also provides a high-standard farmland remote sensing identification method.

[0103] Figure 2 It is a flow chart of a high-standard farmland remote sensing identification method provided by an embodiment of the present invention.

[0104] like Figure 2 As shown, the high-standard farmland remote sensing identification method includes:

[0105] Step 201: obtaining second multi-feature index data of each second farmland pixel in a second preset area; the second multi-feature index data includes values ​​of the corresponding second farmland pixels on multiple preset indicators.

[0106] Step 202: Based on the high-standard farmland remote sensing identification model, according to the second multi-feature indicator data, the farmland type of each second farmland pixel is obtained; the high-standard farmland remote sensing identification model is trained using the training method of the high-standard farmland remote sensing identification model described in the above embodiment, and the farmland types include: high-standard farmland and other farmland.

[0107] Based on a general inventive concept, the present invention also provides a training device for a high-standard farmland remote sensing identification model.

[0108] Figure 3 Schematic diagram of the structure of a training device for a high-standard farmland remote sensing recognition model provided by an embodiment of the present invention. Figure 3 As shown, the training device of the high-standard farmland remote sensing identification model includes:

[0109] The first acquisition module 31 is used to acquire first multi-feature index data of each first farmland pixel in a first preset area; the first multi-feature index data includes values ​​of the first farmland pixel on multiple preset indicators respectively;

[0110] The normalization module 32 is used to perform normalization processing on the first multi-feature indicator data.

[0111] The calculation module 33 is used to obtain the comprehensive evaluation index of each first farmland pixel based on the first multi-feature index data after normalization based on the entropy weight TOPSIS method; the comprehensive evaluation index is used to reflect the conformity of the corresponding farmland pixel with the high-standard farmland.

[0112] The initial classification module 34 is used to perform initial classification on each first farmland pixel according to the comprehensive evaluation index of each first farmland pixel to obtain the initial classification result of each first farmland pixel.

[0113] The verification module 35 is used to verify the authenticity of the initial classification result of the target first farmland pixel.

[0114] The training module 36 is used to train the random forest model using the target data of the target first farmland pixel that has passed the authenticity verification as the model training sample to obtain a high-standard farmland remote sensing identification model; the target data includes the first multi-feature indicator data corresponding to the target first farmland pixel and the actual farmland type.

[0115] Optionally, the first acquisition module 31 is specifically configured to:

[0116] (1) Obtaining high spatial resolution satellite remote sensing images, high spectral resolution satellite remote sensing images and digital elevation models of the first preset area.

[0117] (2) Obtaining first multi-feature indicator data of each of the first farmland pixels based on the high spatial resolution satellite remote sensing image, the high spectral resolution satellite remote sensing image and the digital elevation model.

[0118] Optionally, the preset indicators include: field slope, field connectivity, field fragmentation, soil organic matter content, soil pH value, soil moisture index, soil degradation index, crop growth conditions, distance between the field and the nearest irrigation ditch, field ditch density, field road accessibility, distance from the field to residential areas, distance from the field to traffic arteries, forest network density, shelterbelt coverage and the distance from the field to its nearest transmission tower.

[0119] Optionally, the initial classification module 34 is specifically used for:

[0120] For each of the first farmland pixels, based on a preset correspondence between an index threshold range and a farmland type, the farmland type of the first farmland pixel is determined according to a comprehensive evaluation index of the first farmland pixel.

[0121] Optionally, the farmland types include: high-standard farmland and other farmland.

[0122] Based on a general inventive concept, the present invention also provides a high-standard farmland remote sensing identification device.

[0123] Figure 4 It is a structural schematic diagram of a high-standard farmland remote sensing identification device provided by an embodiment of the present invention.

[0124] like Figure 4 As shown, the high-standard farmland remote sensing identification device includes:

[0125] The second acquisition module 41 is used to acquire second multi-feature index data of each second farmland pixel in the second preset area; the second multi-feature index data includes values ​​of the corresponding second farmland pixels on multiple preset indicators.

[0126] The classification module 42 is used to obtain the farmland type of each second farmland pixel based on the high-standard farmland remote sensing identification model and the second multi-feature indicator data; the high-standard farmland remote sensing identification model is trained using the training method of the high-standard farmland remote sensing identification model described in the above embodiment, and the farmland types include: high-standard farmland and other farmland.

[0127] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0128] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0129] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present invention belong.

[0130] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0131] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0132] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0133] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0134] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0135] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A training method for a high-standard farmland remote sensing recognition model, characterized in that: include: Acquire first multi-feature index data of each first farmland pixel in a first preset area; The first multi-feature index data includes values ​​corresponding to the first farmland pixel on multiple preset indicators; Normalizing the first plurality of feature indicator data; Based on the entropy weight TOPSIS method, according to the normalized first multi-feature index data, a comprehensive evaluation index of each of the first farmland pixels is obtained; The comprehensive evaluation index is used to reflect the conformity of the corresponding farmland pixel with the high-standard farmland; performing initial classification on each of the first farmland pixels according to the comprehensive evaluation index of each of the first farmland pixels to obtain an initial classification result of each of the first farmland pixels; Verify the authenticity of the initial classification results of the first target farmland pixel; The target data of the target first farmland pixel that has passed the authenticity verification is used as a model training sample to train a random forest model to obtain a high-standard farmland remote sensing identification model; the target data includes the first multi-feature indicator data corresponding to the target first farmland pixel and the actual farmland type.

2. The training method of the high-standard farmland remote sensing identification model according to claim 1 is characterized in that: The step of obtaining the first multi-feature index data of each first farmland pixel in the first preset area specifically includes: Acquire high spatial resolution satellite remote sensing images, high spectral resolution satellite remote sensing images and digital elevation models of the first preset area; The first multi-feature indicator data of each of the first farmland pixels are obtained according to the high spatial resolution satellite remote sensing image, the high spectral resolution satellite remote sensing image and the digital elevation model.

3. The training method of the high-standard farmland remote sensing identification model according to claim 1 or 2, characterized in that: The preset indicators include: field slope, field continuity, field fragmentation, soil organic matter content, soil pH value, soil moisture index, soil degradation index, crop growth conditions, distance between the field and the nearest irrigation ditch, field ditch density, field road accessibility, distance from the field to residential areas, distance from the field to traffic arteries, forest network density, shelterbelt coverage rate and distance from the field to the nearest transmission tower.

4. The training method of the high-standard farmland remote sensing identification model according to claim 1 is characterized in that: The initial classification of each of the first farmland pixels according to the comprehensive evaluation index of each of the first farmland pixels to obtain the initial classification result of each of the first farmland pixels specifically includes: For each of the first farmland pixels, based on a preset correspondence between an index threshold range and a farmland type, the farmland type of the first farmland pixel is determined according to a comprehensive evaluation index of the first farmland pixel.

5. The training method of the high-standard farmland remote sensing identification model according to claim 4 is characterized in that: The types of farmland include: high-standard farmland and other farmland.

6. A high-standard farmland remote sensing identification method, characterized in that: include: Acquire second multi-feature index data of each second farmland pixel in a second preset area; the second multi-feature index data includes values ​​corresponding to the second farmland pixel on a plurality of preset indicators; Based on the high-standard farmland remote sensing identification model, the farmland type of each of the second farmland pixels is obtained according to the second multi-feature indicator data; the high-standard farmland remote sensing identification model is trained using the training method of the high-standard farmland remote sensing identification model as described in any one of claims 1 to 5, and the farmland types include high-standard farmland and other farmland.

7. A training device for a high-standard farmland remote sensing identification model, characterized in that: include: A first acquisition module, used to acquire first multi-feature index data of each first farmland pixel in a first preset area; The first multi-feature index data includes values ​​corresponding to the first farmland pixel on multiple preset indicators; A normalization module, used for normalizing the first plurality of feature indicator data; A calculation module, used for obtaining a comprehensive evaluation index of each of the first farmland pixels based on the first multi-feature index data after normalization based on an entropy weight TOPSIS method; The comprehensive evaluation index is used to reflect the conformity of the corresponding farmland pixel with the high-standard farmland; An initial classification module, used for performing initial classification on each of the first farmland pixels according to the comprehensive evaluation index of each of the first farmland pixels, and obtaining an initial classification result of each of the first farmland pixels; A verification module is used to verify the authenticity of the initial classification result of the target first farmland pixel; A training module is used to train a random forest model using the target data of the target first farmland pixel that has passed authenticity verification as a model training sample to obtain a high-standard farmland remote sensing identification model; the target data includes the first multi-feature indicator data corresponding to the target first farmland pixel and the actual farmland type.

8. A high-standard farmland remote sensing identification device, characterized in that: include: A second acquisition module, used to acquire second multi-feature index data of each second farmland pixel in a second preset area; The second multi-feature index data includes values ​​corresponding to the second farmland pixels on multiple preset indicators; A classification module is used to obtain the farmland type of each of the second farmland pixels based on the high-standard farmland remote sensing identification model and the second multi-feature indicator data; the high-standard farmland remote sensing identification model is trained using the training method of the high-standard farmland remote sensing identification model as described in any one of claims 1 to 5, and the farmland types include high-standard farmland and other farmland.

Images