A Cultivated Land Extraction Method and Device Based on Temporal Spectral Variation Characteristics

Abstract

The present invention relates to the field of artificial intelligence technology, and relates to a cultivated land extraction method and device based on temporal spectral change characteristics. The method includes obtaining first information, second information, and third information. The first information includes high-resolution images collected along the railway in time series, the second information includes medium-low resolution images collected along the railway in time series, and the third information includes cultivated land vector boundaries; preprocessing the first information, second information, and third information to obtain preprocessed sample information; constructing a deep learning data set based on the preprocessed sample information; training a preset cultivated land extraction deep network model using the deep learning data set to obtain a trained cultivated land extraction deep network model; extracting the cultivated land in the railway plan research area based on the trained cultivated land extraction deep network model to obtain an extraction result. The present invention introduces multi-temporal information to obtain the temporal spectral change law of vegetation to accurately distinguish easily confused ground objects.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and more particularly, to a method and device for extracting cultivated land based on temporal spectral change characteristics. Background Art

[0002] For the extraction of cultivated land in the railway plan research area, remote sensing technology is usually used for identification. Remote sensing technology can quickly identify the current cultivated land distribution by obtaining high-resolution image data on the ground. However, the current single-temporal image has limitations in the extraction of cultivated land, which is prone to confusion of ground objects (such as grasslands, forests, etc.), resulting in low identification accuracy. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and device for extracting cultivated land based on temporal spectral change characteristics.

[0004] To achieve the above purpose, the embodiments of the present application provide the following technical solutions:

[0005] On the one hand, the embodiments of the present application provide a method for extracting cultivated land based on temporal spectral change characteristics, the method comprising:

[0006] Obtaining first information, second information and third information, the first information including high-resolution images collected along the railway in time series, the second information including medium-low resolution images collected along the railway in time series, and the third information including cultivated land vector boundaries;

[0007] Preprocessing the first information, the second information and the third information to obtain preprocessed sample information, the preprocessing being used to correct the pixel error between images;

[0008] Constructing a deep learning data set based on the preprocessed sample information;

[0009] Training a preset cultivated land extraction deep network model using the deep learning data set to obtain a trained cultivated land extraction deep network model;

[0010] Extracting the cultivated land in the railway plan research area based on the trained cultivated land extraction deep network model to obtain an extraction result, the extraction result including cultivated land in at least one local area.

[0011] On the second hand, the embodiments of the present application provide a device for extracting cultivated land based on temporal spectral change characteristics, the device comprising:

[0012] An acquisition module for acquiring first information, second information, and third information, where the first information includes high-resolution images collected along the railway in sequence, the second information includes medium- and low-resolution images collected along the railway in sequence, and the third information includes cultivated land vector boundaries;

[0013] A first processing module for preprocessing the first information, the second information, and the third information to obtain preprocessed sample information, where the preprocessing is used to correct the pixel error between images;

[0014] A second processing module for constructing a deep learning dataset based on the preprocessed sample information;

[0015] A third processing module for training a preset cultivated land extraction deep network model using the deep learning dataset to obtain a trained cultivated land extraction deep network model;

[0016] A fourth processing module for extracting cultivated land in the railway plan study area based on the trained cultivated land extraction deep network model to obtain an extraction result, where the extraction result includes cultivated land in at least one local area.

[0017] In a third aspect, an embodiment of the present application provides a cultivated land extraction device based on temporal spectral change features, where the device includes a memory and a processor. The memory is used to store a computer program; the processor is used to implement the steps of the above-mentioned cultivated land extraction method based on temporal spectral change features when executing the computer program.

[0018] In a fourth aspect, an embodiment of the present application provides a readable storage medium with a computer program stored thereon, and the computer program, when executed by a processor, implements the steps of the above-mentioned cultivated land extraction method based on temporal spectral change features.

[0019] The beneficial effects of the present invention are as follows:

[0020] The present invention constructs a deep learning dataset through the first information, the second information, and the third information, and then uses the deep learning dataset to train the cultivated land extraction deep network model. The temporal features of medium- and low-resolution remote sensing images are introduced into the deep network model, enabling the deep network to stably locate and respond to cultivated land areas even in complex backgrounds, and accurately distinguish easily confused land covers such as cultivated land, grassland, forest land, and orchard land. The cultivated land boundary segmentation is fine and smooth, and the cultivated land semantic information is complete and continuous, with good robustness and reliability.

[0021] Other features and advantages of the present invention will be described in the subsequent specification, and in part, will be obvious from the specification, or can be understood by implementing the embodiments of the present invention. The objectives and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in the written specification and the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0023] Figure 1 It is a schematic flowchart of the cultivated land extraction method based on the temporal spectral change characteristics described in the embodiments of the present invention.

[0024] Figure 2 It is a schematic structural diagram of the cultivated land extraction device based on the temporal spectral change characteristics described in the embodiments of the present invention.

[0025] Figure 3 It is a schematic structural diagram of the cultivated land extraction equipment based on the temporal spectral change characteristics described in the embodiments of the present invention.

[0026] Reference numerals in the figure: 800, cultivated land extraction equipment based on temporal spectral change characteristics; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component; 901, acquisition module; 902, first processing module; 903, second processing module; 904, third processing module; 905, fourth processing module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and shown in the accompanying drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the present invention to be protected, but only represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0028] It should be noted that similar reference numerals and letters denote similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. At the same time, in the description of the present invention, terms such as "first" and "second" are only used for distinguishing descriptions and cannot be construed as indicating or implying relative importance.

[0029] Embodiment 1:

[0030] This embodiment provides a cultivated land extraction method based on temporal spectral change features. It can be understood that in this embodiment, a scenario can be set up, for example: a scenario for extracting and identifying cultivated land near a newly built or existing railway.

[0031] Refer to Figure 1 , the figure shows that this method includes step S1, step S2, step S3, step S4 and step S5.

[0032] Step S1: Obtain the first information, the second information and the third information. The first information includes high-resolution images collected along the railway in time series, the second information includes medium-low resolution images collected along the railway in time series, and the third information includes the cultivated land vector boundary;

[0033] In this step, a preset resolution threshold is 0.3m×0.3m. Images better than the resolution threshold are used as high-resolution images, and images worse than the resolution threshold are used as medium-low resolution images. The first information can provide deep and refined ground object information for the model, and the second information is the time series of the normalized difference vegetation index and synthetic aperture radar images, providing the change trend of vegetation elements along the railway over time. By collecting the first information, the second information and the third information, the growth laws of easily confused land types such as crops, grasslands and forests are comprehensively considered.

[0034] Step S2: Preprocess the first information, the second information and the third information to obtain preprocessed sample information, and the preprocessing is used to correct the pixel error between images;

[0035] In step S2, it also includes step S21 and step S22, specifically:

[0036] Step S21: Project the first information, the second information and the third information into the same earth coordinate system to obtain the projected sample information;

[0037] Step S22: Use the feature point matching method to correct the three types of sample data included in the projected sample information to obtain the preprocessed sample information.

[0038] In this embodiment, since there are significant differences among the three types of heterogeneous data, namely the first information, the second information, and the third information, in order to ensure the effective operation of the model, it is necessary to preprocess the first information, the second information, and the third information to eliminate pixel errors, so that the position error is within 1 pixel.

[0039] Step S3: Construct a deep learning dataset based on the preprocessed sample information;

[0040] In step S3, it further includes steps S31, S32, S33, and S34, specifically as follows:

[0041] Step S31: Use a preset sliding window to segment the high-resolution image in the preprocessed sample information to obtain the segmented first image information;

[0042] In this step, due to the high fragmentation degree of cultivated land along the railway, the size of the preset sliding window is preset to a 256×256 grid, which can map 5 square kilometers of cultivated land, so as to effectively extract the cultivated land range. It should be noted that when using the sliding window to segment the high-resolution image in the preprocessed sample information, a 50% slice overlap is set to ensure the parameter sharing and connectivity of the convolutional neural network.

[0043] Step S32: Determine the geographic boundary information according to the segmented first image information;

[0044] Step S33: Use the geographic boundary information to block-process the medium- and low-resolution images in the preprocessed sample information to obtain the segmented second image information;

[0045] In this step, the boundary information can be identified in the high-resolution image, but the boundary information cannot be identified in the medium- and low-resolution images. Therefore, it is necessary to block-process the medium- and low-resolution images according to the boundary information identified in the high-resolution image.

[0046] Step S34: Import the cultivated land vector data into the segmented first image information and the segmented second image information to obtain a deep learning dataset.

[0047] In this step, the cultivated land vector data is rasterized and binarized. The cultivated land pixel value is set to 1, and the non-cultivated land pixel value is set to 2. The imported first image information and the geographic boundary information of the second image information are segmented to obtain cultivated land sample data corresponding to the first image information and the second image information, so as to construct a deep learning dataset. Each set of data in this dataset includes the first image information, the second image information, and the corresponding cultivated land sample information.

[0048] Step S4: Use the deep learning dataset to train a preset cultivated land extraction deep network model to obtain a trained cultivated land extraction deep network model;

[0049] In this step, a specific implementation is to divide the deep learning dataset into a training set, a validation set, and a test set according to a ratio of 8:1:1 to support the training, validation, and testing of the deep learning model.

[0050] Step S5: Based on the trained cultivated land extraction deep network model, extract the cultivated land in the railway plan study area to obtain an extraction result, where the extraction result includes cultivated land in at least one local area.

[0051] In this step, the results of multiple local areas are spatially merged to obtain the cultivated land distribution information of the final target area.

[0052] Step S5 also includes steps S51, S52, S53, and S54, specifically as follows:

[0053] Step S51: Downsample the first information of the cultivated land to be extracted to obtain first feature information;

[0054] In this step, the downsampling includes four stages, which are connected by a max pooling layer between each stage, and the first information is downsampled four times.

[0055] Step S52: Upsample the second information of the cultivated land to be extracted to obtain second feature information;

[0056] In this step, the second information of the cultivated land to be extracted is upsampled so that its size is the same as that of the first information of the cultivated land to be extracted after four times of downsampling.

[0057] Step S53: Stack the first feature information and the second feature information to obtain stacked feature information;

[0058] Step S53 also includes steps S531, S532, S533, S534, and S535, specifically as follows:

[0059] Step S531: Obtain four dilated convolutional layers, and the dilation rates of the four dilated convolutional layers are 1, 2, 4, and 8 respectively;

[0060] Step S532: Use the four dilated convolutional layers to perform dilated convolution on the first feature information respectively to obtain four third feature information;

[0061] Step S533: Stack the four third feature information to obtain fourth feature information;

[0062] Step S534: Convolve the fourth feature information to obtain the first feature information;

[0063] Step S535: Stack the first feature information and the second feature information to obtain the stacked feature information.

[0064] In this embodiment, dilated convolutions with dilation rates of 1, 2, 4, and 8 are used in parallel to balance the computational efficiency and model performance, increase the receptive field, and capture multi-scale information. Subsequently, 1×1 convolutions are used to fuse feature maps of different scales, and then multi-temporal features are introduced through medium- and low-resolution temporal image slices and superimposed with the multi-temporal features to generate the stacked feature information. The present invention introduces the temporal change features of vegetation, effectively solving the problem of accurately distinguishing easily confused land types (such as grasslands and cultivated lands in the growth peak period) in remote sensing images.

[0065] Step S54: Upsample the stacked feature information to obtain the extraction result.

[0066] In step S54, it further includes step S541, step S542, and step S543, specifically as follows:

[0067] Step S541: Obtain four upsampling modules, and each upsampling module includes an attention gating unit and a convolutional layer;

[0068] Step S542: Send the stacked feature information to the four upsampling modules for repeated upsampling to obtain the prediction probability map;

[0069] Step S543: Determine the extraction result according to the prediction probability map.

[0070] In this step, the prediction probability map is binarized. With a threshold of 0.5, the pixels greater than 0.5 are cultivated lands, and the pixels less than or equal to 0.5 are non-cultivated lands, thereby obtaining the final predicted raster map, that is, the extraction result.

[0071] In this embodiment, the upsampling also includes four stages. The stacked feature information is upsampled four times, and finally, the number of channels of the feature map is adjusted to 1 through one layer of convolution, and a logical activation function is applied to generate the prediction probability map. Between the corresponding encoder and decoder stages, skip connections are used to retain multi-level feature information, and at the same time, attention gating units are introduced to screen important feature information and enhance the expressive ability of the decoding process.

[0072] Embodiment 2:

[0073] As Figure 2As shown in the figure, this embodiment provides a cultivated land extraction device based on the temporal spectral change characteristics. The device includes an acquisition module 901, a first processing module 902, a second processing module 903, a third processing module 904, and a fourth processing module 905, specifically as follows:

[0074] The acquisition module 901 is configured to acquire first information, second information, and third information. The first information includes high-resolution images collected along the railway line in time series, the second information includes medium- and low-resolution images collected along the railway line in time series, and the third information includes cultivated land vector boundaries.

[0075] The first processing module 902 is configured to preprocess the first information, the second information, and the third information to obtain preprocessed sample information. The preprocessing is used to correct the pixel error between images.

[0076] The second processing module 903 is configured to construct a deep learning data set based on the preprocessed sample information.

[0077] The third processing module 904 is configured to train a preset cultivated land extraction deep network model using the deep learning data set to obtain a trained cultivated land extraction deep network model.

[0078] The fourth processing module 905 is configured to extract cultivated land in the railway plan research area based on the trained cultivated land extraction deep network model to obtain an extraction result, where the extraction result includes cultivated land in at least one local area.

[0079] In a specific implementation manner of the present disclosure, the first processing module further includes a first processing unit and a second processing unit, specifically as follows:

[0080] The first processing unit is configured to project the first information, the second information, and the third information onto the same earth coordinate system to obtain projected sample information.

[0081] The second processing unit is configured to correct the three types of sample data included in the projected sample information using the feature point matching method to obtain preprocessed sample information.

[0082] In a specific implementation manner of the present disclosure, the second processing module further includes a third processing unit, a fourth processing unit, a fifth processing unit, and a sixth processing unit, specifically as follows:

[0083] The third processing unit is configured to segment the high-resolution image in the preprocessed sample information using a preset sliding window to obtain segmented first image information.

[0084] A fourth processing unit, configured to determine geographical boundary information according to the segmented first image information;

[0085] A fifth processing unit, configured to perform block processing on the medium and low resolution images in the preprocessed sample information by using the geographical boundary information, so as to obtain segmented second image information;

[0086] A sixth processing unit, configured to import the cultivated land vector data into the segmented first image information and the segmented second image information, so as to obtain a deep learning data set.

[0087] In a specific embodiment of the present disclosure, the fourth processing module further includes a seventh processing unit, an eighth processing unit, a ninth processing unit, and a tenth processing unit, specifically:

[0088] The seventh processing unit is configured to perform downsampling on the first information of the cultivated land to be extracted, so as to obtain first feature information;

[0089] The eighth processing unit is configured to perform upsampling on the second information of the cultivated land to be extracted, so as to obtain second feature information;

[0090] The ninth processing unit is configured to stack the first feature information and the second feature information, so as to obtain stacked feature information;

[0091] The tenth processing unit is configured to perform upsampling on the stacked feature information, so as to obtain an extraction result.

[0092] In a specific embodiment of the present disclosure, the ninth processing unit further includes a first acquisition unit, an eleventh processing unit, a twelfth processing unit, a thirteenth processing unit, and a fourteenth processing unit, specifically:

[0093] The first acquisition unit is configured to acquire four dilated convolutional layers, and the dilation rates of the four dilated convolutional layers are 1, 2, 4, and 8 respectively;

[0094] The eleventh processing unit is configured to perform dilated convolution on the first feature information by using the four dilated convolutional layers respectively, so as to obtain four third feature information;

[0095] The twelfth processing unit is configured to stack the four third feature information, so as to obtain fourth feature information;

[0096] The thirteenth processing unit is configured to perform convolution on the fourth feature information, so as to obtain first feature information;

[0097] The fourteenth processing unit is configured to stack the first feature information and the second feature information, so as to obtain stacked feature information.

[0098] In a specific embodiment of the present disclosure, the tenth processing unit further includes a second acquisition unit, a fifteenth processing unit, and a sixteenth processing unit, specifically as follows:

[0099] The second acquisition unit is configured to acquire four upsampling modules, and the upsampling module includes an attention gating unit and a convolutional layer;

[0100] The fifteenth processing unit is configured to send the stacked feature information to the four upsampling modules for repeated upsampling to obtain a predicted probability map;

[0101] The sixteenth processing unit is configured to determine an extraction result according to the predicted probability map.

[0102] It should be noted that regarding the device in the above embodiments, the specific manners in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated herein.

[0103] Embodiment 3:

[0104] Corresponding to the above method embodiment, a cultivated land extraction device based on temporal spectral change features is further provided in this embodiment. The cultivated land extraction device based on temporal spectral change features described below can be correspondingly referred to the cultivated land extraction method based on temporal spectral change features described above.

[0105] Figure 3 It is a block diagram of a cultivated land extraction device 800 based on temporal spectral change features shown according to an exemplary embodiment. As Figure 3 shown, the cultivated land extraction device 800 based on temporal spectral change features may include: a processor 801, a memory 802. The cultivated land extraction device 800 based on temporal spectral change features may further include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0106] Among them, the processor 801 is used to control the overall operation of the cultivated land extraction device 800 based on the temporal spectral change characteristics to complete all or part of the steps in the above-mentioned cultivated land extraction method based on the temporal spectral change characteristics. The memory 802 is used to store various types of data to support the operation of the cultivated land extraction device 800 based on the temporal spectral change characteristics. These data may include, for example, instructions for any application or method operating on the cultivated land extraction device 800 based on the temporal spectral change characteristics, as well as application-related data, such as contact data, sent and received messages, pictures, audio, video, and so on. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic memory, flash memory, a magnetic disk, or an optical disc. The multimedia component 803 may include a screen and an audio component. Among them, the screen can be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone, and the microphone is used to receive external audio signals. The received audio signals can be further stored in the memory 802 or sent through the communication component 805. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules can be a keyboard, a mouse, buttons, etc. These buttons can be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the cultivated land extraction device 800 based on the temporal spectral change characteristics and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination of one or more of them. Therefore, the corresponding communication component 805 may include: a Wi-Fi module, a Bluetooth module, and an NFC module.

[0107] In an exemplary embodiment, the cultivated land extraction device 800 based on temporal spectral change features can be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components, and is used to execute the above-mentioned cultivated land extraction method based on temporal spectral change features.

[0108] In another exemplary embodiment, a computer-readable storage medium including program instructions is further provided. When the program instructions are executed by a processor, the steps of the above-mentioned cultivated land extraction method based on temporal spectral change features are implemented. For example, the computer-readable storage medium can be the above-mentioned memory 802 including program instructions, and the above-mentioned program instructions can be executed by the processor 801 of the cultivated land extraction device 800 based on temporal spectral change features to complete the above-mentioned cultivated land extraction method based on temporal spectral change features.

[0109] Embodiment 4:

[0110] Corresponding to the above method embodiment, a readable storage medium is further provided in this embodiment. A readable storage medium described below can be correspondingly referred to with a cultivated land extraction method based on temporal spectral change features described above.

[0111] A readable storage medium stores a computer program. When the computer program is executed by a processor, the steps of the cultivated land extraction method based on temporal spectral change features in the above method embodiment are implemented.

[0112] The readable storage medium can specifically be various readable storage media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disc that can store program codes.

[0113] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0114] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A cultivated land extraction method based on the characteristics of temporal spectral changes, characterized in that, Including: Obtain the first information, the second information, and the third information. The first information includes high-resolution images collected along the railway in chronological order, the second information includes medium- and low-resolution images collected along the railway in chronological order, and the third information includes cultivated land vector boundaries; Preprocess the first information, the second information, and the third information to obtain preprocessed sample information. The preprocessing is used to correct the pixel error between images; Construct a deep learning dataset based on the preprocessed sample information; Use the deep learning dataset to train a preset cultivated land extraction deep network model to obtain a trained cultivated land extraction deep network model; Extract the cultivated land in the railway plan study area based on the trained cultivated land extraction deep network model to obtain an extraction result. The extraction result includes cultivated land in at least one local area, including: Downsample the first information of the cultivated land to be extracted to obtain first feature information; Upsample the second information of the cultivated land to be extracted to obtain second feature information; Stack the first feature information and the second feature information to obtain stacked feature information, including: Obtain four dilated convolutional layers, and the dilation rates of the four dilated convolutional layers are 1, 2, 4, and 8 respectively; Use the four dilated convolutional layers to perform dilated convolution on the first feature information respectively to obtain four third feature information; Superimpose the four third feature information to obtain fourth feature information; Convolve the fourth feature information to obtain first feature information; Stack the first feature information and the second feature information to obtain stacked feature information; Upsample the stacked feature information to obtain an extraction result.

2. The cultivated land extraction method based on the temporal spectral change characteristics according to claim 1, wherein Preprocess the first information, the second information, and the third information to obtain preprocessed sample information, including: Project the first information, the second information, and the third information onto the same earth coordinate system to obtain projected sample information; Use the feature point matching method to correct the three types of sample data included in the projected sample information to obtain preprocessed sample information.

3. The cultivated land extraction method based on the temporal spectral change characteristics according to claim 1, wherein Construct a deep learning dataset based on the preprocessed sample information, including: Use a preset sliding window to segment the high-resolution images in the preprocessed sample information to obtain segmented first image information; Determine geographical boundary information according to the segmented first image information; ​ ​ 4. The cultivated land extraction method based on temporal spectral variation features according to claim 1, wherein ​ ​ ​ ​ 5. An arable land extraction device based on the characteristics of temporal spectral variation, characterized in that, ​ An acquisition module for acquiring first information, second information, and third information, where the first information includes high-resolution images collected along the railway in chronological order, the second information includes medium- and low-resolution images collected along the railway in chronological order, and the third information includes arable land vector boundaries; A first processing module for preprocessing the first information, the second information, and the third information to obtain preprocessed sample information, where the preprocessing is used to correct the pixel error between images; A second processing module for constructing a deep learning data set based on the preprocessed sample information; A third processing module for training a preset arable land extraction deep network model using the deep learning data set to obtain a trained arable land extraction deep network model; A fourth processing module for extracting arable land in the railway plan study area based on the trained arable land extraction deep network model to obtain an extraction result, where the extraction result includes arable land in at least one local area; Among them, the fourth processing module includes: A seventh processing unit for downsampling the first information of the arable land to be extracted to obtain first feature information; An eighth processing unit for upsampling the second information of the arable land to be extracted to obtain second feature information; A ninth processing unit for stacking the first feature information and the second feature information to obtain stacked feature information; A tenth processing unit for upsampling the stacked feature information to obtain an extraction result; Among them, the ninth processing unit includes: A first acquisition unit for acquiring four dilated convolutional layers, and the dilation rates of the four dilated convolutional layers are 1, 2, 4, and 8 respectively; An eleventh processing unit for performing dilated convolution on the first feature information using the four dilated convolutional layers respectively to obtain four third feature information; A twelfth processing unit for superimposing the four third feature information to obtain fourth feature information; A thirteenth processing unit for performing convolution on the fourth feature information to obtain first feature information; A fourteenth processing unit for stacking the first feature information and the second feature information to obtain stacked feature information.

6. The cultivated land extraction device based on the temporal spectral change characteristics according to claim 5, characterized in that, The first processing module includes: A first processing unit for projecting the first information, the second information, and the third information onto the same earth coordinate system to obtain projected sample information; A second processing unit for correcting the three types of sample data included in the projected sample information using the feature point matching method to obtain preprocessed sample information.

7. The cultivated land extraction device based on the temporal spectral change characteristics according to claim 5, characterized in that, The second processing module includes: A third processing unit for splitting the high-resolution image in the preprocessed sample information using a preset sliding window to obtain split first image information; A fourth processing unit for determining geographical boundary information based on the split first image information; A fifth processing unit for performing block processing on the medium- and low-resolution images in the preprocessed sample information using the geographical boundary information to obtain split second image information; A sixth processing unit, configured to import the cultivated land vector data into the segmented first image information and the segmented second image information to obtain a deep learning data set.

8. The cultivated land extraction method based on the temporal spectral variation characteristics according to claim 5, wherein The tenth processing unit includes: A second acquisition unit, configured to acquire four upsampling modules, where each upsampling module includes an attention gating unit and a convolutional layer; A fifteenth processing unit, configured to send the stacked feature information to the four upsampling modules for repeated upsampling to obtain a predicted probability map; A sixteenth processing unit, configured to determine an extraction result according to the predicted probability map.

Images