Method, device, storage medium and processor for determining a region boundary line

Abstract

Embodiments of the present application provide a method, device, processor and storage medium for determining a region boundary line. The method comprises: acquiring a first region image and a second region image, and performing matching calibration processing; inputting the processed first region image and the processed second region image into a region feature extraction network respectively to obtain a first region feature and a second region feature; determining a plurality of vegetation indexes of the first region image and the second region image respectively; stacking a panchromatic band on the basis of the first region image and the second region image respectively, and inputting into a crop growth change feature extraction network to obtain a third region feature; processing the first region feature, the second region feature and the third region feature to obtain a first enhanced feature and a second enhanced feature; inputting the first enhanced feature, the second enhanced feature and the third region feature into a feature fusion network to obtain a region mask image of the region; and determining a boundary line of the region through the region mask image.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smart agriculture, and in particular to a method and device for determining a region boundary line, a storage medium and a processor. BACKGROUND

[0002] Boundary extraction of an agricultural region is an important technical task in the field of agriculture, which aims to obtain field boundaries through remote sensing data such as unmanned aerial vehicles or satellites to support agricultural production and management. At present, the field data of the agricultural region is mainly obtained by relying on remote sensing interpretation combined with manual reporting. However, this method has problems such as difficulty in ensuring data reporting accuracy and low updating efficiency. With the continuous development of remote sensing technologies such as satellites and unmanned aerial vehicles, more and more high-resolution, multi-spectral remote sensing data can be obtained, which can provide a basis for field boundary extraction.

[0003] In the prior art, the method for extracting field boundaries mainly extracts color features and shape features of images, classifies pixels, and extracts field boundaries. However, in most areas in southern China, there are a large number of irregularly shaped paddy fields, small field boundaries, and villages distributed between fields. Complex vegetation coverage, crop types, and topography will affect the detection results, resulting in low extraction accuracy of field boundaries in such cases. SUMMARY

[0004] The purpose of the embodiments of the present application is to provide a method and device for determining a region boundary line, a storage medium and a processor.

[0005] To achieve the above-mentioned purpose, the first aspect of the present application provides a method for determining a region boundary line, comprising:

[0006] obtaining a first region image and a second region image of crops planted in a region in a first growth period and a second growth period, respectively;

[0007] performing matching and calibration processing on the first region image and the second region image;

[0008] inputting the processed first region image and the processed second region image into a region feature extraction network, respectively, to obtain first region features and second region features corresponding to the first region image and the second region image, respectively;

[0009] determining a plurality of vegetation indices of the first region image and the second region image, respectively;

[0010] stacking a panchromatic band based on the plurality of vegetation indices in the first region image and the second region image, respectively, and simultaneously inputting the first region image and the second region image into a crop growth change feature extraction network to obtain third region features;

[0011] The first region feature, the second region feature and the third region feature are processed to obtain a first enhanced feature and a second enhanced feature;

[0012] The first enhanced feature, the second enhanced feature and the third region feature are input into a feature fusion network to obtain a region mask image of the region;

[0013] The boundary line of the region is determined through the region mask image.

[0014] In the embodiments of the present application, processing the first region feature, the second region feature and the third region feature to obtain the first enhanced feature and the second enhanced feature includes: subtracting the third region feature from the first region feature to obtain the first enhanced feature of the crop growth in the region; and multiplying the third region feature by the second region feature to obtain the second enhanced feature of the crop growth in the region.

[0015] In the embodiments of the present application, inputting the first enhanced feature, the second enhanced feature and the third region feature into the feature fusion network includes: stacking the first enhanced feature, the second enhanced feature and the third region feature in the channel dimension and then inputting them into the feature fusion network.

[0016] In the embodiments of the present application, the loss function adopted by the feature fusion network is as formula (1):

[0017] Loss=Loss1+0.5*(Loss2+Loss3) (1)

[0018] Wherein, Loss1 is the loss function when the feature fusion network generates the region mask image, the expression is as formula (2), Loss2 is the loss function between the first enhanced feature and the second enhanced feature, Loss3 is the loss function between the first region feature and the second region feature, the expressions of Loss2 and Loss3 are as formula (3):

[0019]

[0020]

[0021] Wherein, y ij represents the true class of the i-th pixel of the input image for the feature fusion network, p ij is the prediction probability of the i-th pixel of the input image for the feature fusion network as class j, n is the total number of samples, k is the number of classes, y i represents the sign function, when the true class of the i-th pixel is j, the value of y i is 1, otherwise 0, p i represents the prediction probability of the i-th pixel belonging to class j.

[0022] In the embodiments of the present application, the boundary line of the region is determined by the region mask image, including: extracting edge data of the region mask image to obtain a vectorized boundary line of the region.

[0023] In the embodiments of the present application, the first region feature and / or the second region feature at least include color, texture, shape, and the third region feature represents growth change information of the crop.

[0024] In the embodiments of the present application, the plurality of vegetation indexes include NDVI and NDWI.

[0025] The second aspect of the present application provides a processor configured to execute the above-mentioned method for determining a region boundary line.

[0026] The third aspect of the present application provides an apparatus for determining a region boundary line, including the above-mentioned processor.

[0027] The fourth aspect of the present application provides a machine readable storage medium, the machine readable storage medium has instructions stored thereon, the instructions, when executed by a processor, cause the processor to be configured to execute the above-mentioned method for determining a region boundary line.

[0028] Through the above technical solution, the crop growth feature, the field feature in the seedling stage and the field feature in the harvesting stage are fused and operated, so that the internal features of the field are strengthened, the boundary line of the field is more obvious, and the extraction accuracy of the field boundary is improved.

[0029] Other features and advantages of the embodiments of the present application will be described in detail in the following specific implementation part. BRIEF DESCRIPTION OF DRAWINGS

[0030] The accompanying drawings are included to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used together with the following specific implementation to explain the embodiments of the present application, but do not constitute a limitation on the embodiments of the present application. In the drawings:

[0031] Figure 1 The flowchart of the method for determining a region boundary line according to the embodiments of the present application is schematically shown;

[0032] Figure 2 The flowchart of the field boundary extraction algorithm based on multi-spectral image according to the embodiments of the present application is schematically shown;

[0033] Figure 3 The structure diagram of the feature fusion network according to the embodiments of the present application is schematically shown;

[0034] Figure 4Fig. 1 schematically shows an internal structure diagram of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0035] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. It should be understood that the specific embodiments described herein are merely used to explain and illustrate the embodiments of the present application, and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present application.

[0036] Figure 1 Fig. 2 schematically shows a flowchart of a method for determining a region boundary line according to an embodiment of the present application. The method comprises the following steps:

[0037] Step 101: Obtain a first region image and a second region image of crops planted in a region in a first growth period and a second growth period respectively.

[0038] Step 102: Perform matching calibration processing on the first region image and the second region image.

[0039] Step 103: Input the processed first region image and the processed second region image into a region feature extraction network respectively, to obtain a first region feature and a second region feature corresponding to the first region image and the second region image respectively.

[0040] Step 104: Determine a plurality of vegetation indices of the first region image and the second region image respectively.

[0041] Step 105: Based on the plurality of vegetation indices, stack a panchromatic band on the basis of the first region image and the second region image respectively, and simultaneously input the first region image and the second region image into a crop growth change feature extraction network, to obtain a third region feature.

[0042] Step 106: Process the first region feature, the second region feature and the third region feature, to obtain a first enhanced feature and a second enhanced feature.

[0043] Step 107: Input the first enhanced feature, the second enhanced feature and the third region feature into a feature fusion network, to obtain a region mask image of the region.

[0044] Step 108: Determine a boundary line of the region through the region mask image.

[0045] The region in the embodiment can refer to an agricultural region, for example, a farmland, a paddy field, a field plot, and the like. The growth process of the crop includes multiple growth periods. In the present scheme, the boundary line of the region can be determined based on the region image of the region when the crop is in different growth periods. Specifically, the first growth period can refer to the seedling stage of the crop, and the second growth period can refer to the harvesting stage of the crop. When the crop in the agricultural region is in the first growth period, a first region image of the region is acquired. When the crop in the agricultural region is in the second growth period, a second region image of the region is acquired. The first region image and the second region image are both multispectral images. Multispectral refers to spectral data composed of multiple wave bands in the visible light range and the near-infrared range. Compared with a single-band black-and-white grayscale image, a multispectral image can provide spectral reflectivity, absorption rate, and the like of an object or a scene under different wave bands, and thus is of great significance for studying environmental parameters such as the land cover type, the vegetation growth state, the soil humidity, and the temperature of the target region.

[0046] The first region image and the second region image can then be subjected to matching calibration processing, so as to reduce the light and image differences caused by different times of shooting, and improve the consistency and comparability of the features. The specific matching calibration processing method can adopt a conventional processing method, which will not be described herein. After the calibration processing, the processed first region image and the processed second region image can be obtained. Next, the processed first region image and the processed second region image can be input into the region feature extraction network, so that the first region feature and the second region feature corresponding to the first region image and the second region image, respectively, can be obtained through the region feature extraction network. The first region feature and the second region feature represent the features of the agricultural region when the crop is in different growth periods, respectively.

[0047] Specifically, the region feature extraction network can adopt a Unet network, which consists of two parts of an encoder and a decoder. The down-sampling module belongs to the encoder part, which is responsible for gradually reducing the input image and extracting the high-level features of the image. The combination of convolutional layers and pooling layers can be used to realize down-sampling. The convolutional layer is used to extract the features of the image, and the pooling layer is used to reduce the size of the feature map. Through successive convolution and pooling operations, the network can capture more and more global and abstract features. The up-sampling module belongs to the decoder part, which is responsible for gradually restoring the feature map extracted by the encoder and generating a segmentation result with the same size as the input image. The up-sampling module can use the deconvolutional layer (or transposed convolution) to gradually increase the size of the feature map. In each up-sampling step, the decoder merges the feature map of the previous layer with the feature map of the corresponding layer of the encoder to fuse low-level and high-level features. Further, after obtaining the first region image and the second region image, a plurality of vegetation indices of the first region image and the second region image can be determined respectively. The plurality of vegetation indices include NDVI and NDWI. NDVI (Normalized Difference Vegetation Index), i.e., normalized vegetation index, can separate vegetation from water and soil, and can objectively reflect the change of vegetation coverage, which is the best indicator of vegetation growth state and vegetation coverage. NDWI (Normalized Difference Water Index), i.e., normalized water index, uses specific bands of remote sensing images for normalized difference processing to highlight water information in the image. Then, the plurality of vegetation indices calculated can be stacked with the panchromatic band based on the first region image and the second region image respectively. Specifically, the calculated vegetation indices can be denoted as M1 and M2, the panchromatic band can be denoted as Mp, and the three matrices can be stacked into a high-dimensional matrix using the concat(function) method.

[0048] The first region image and the second region image stacked with the panchromatic band are used as input data of the crop growth change feature extraction network, and output data of the crop growth change feature extraction network, i.e., the third region feature, is obtained. Since the first region image is an image collected when the crop is in the seedling stage, and the second region image is an image collected when the crop is in the harvesting stage, the third region feature obtained by simultaneously inputting the first region image and the second region image into the crop growth change feature extraction network can reflect the change feature of the region from the seedling stage to the harvesting stage. Then, the first region feature, the second region feature and the third region feature can be processed to obtain the first enhanced feature and the second enhanced feature.

[0049] In the above scheme, by optimizing the four-band input into NDVI, NDWI and panchromatic input, the characteristics of vegetation and water bodies can be highlighted, thereby reducing the difficulty of model convergence.

[0050] Specifically, such as Figure 2 The diagram illustrates the flowchart of a field boundary extraction algorithm based on multispectral imagery. First, the growth stage of the crops planted within the field is determined. When the crops are in their first growth stage, an image of the first region of the field is acquired; when the crops are in their second growth stage, an image of the second region of the field is acquired. Then, based on a region feature extraction network (…),… Figure 2 The field feature extraction network extracts image features from the two regions respectively, obtaining field features at the seedling stage and the harvest stage. Alternatively, the first and second region images can be input into the crop growth change feature extraction network to obtain features from the third region. Figure 2 (The crop growth change characteristics in the data). Then, the first region features, the second region features, and the third region features can be processed to obtain the first enhanced features and the second enhanced features. The first enhanced features, the second enhanced features, and the third region features are then input into the feature fusion network to obtain the region mask image of the field, so as to further determine the boundary line of the field.

[0051] In one embodiment, processing the first region feature F1, the second region feature F2, and the third region feature F3 to obtain the first enhanced feature and the second enhanced feature includes: subtracting the third region feature F3 from the first region feature F1 to obtain the first enhanced feature Fel of crop growth within the region; and multiplying the third region feature F3 by the second region feature to obtain the second enhanced feature Fe2 of crop growth within the region. The first enhanced feature Fe1, the second enhanced feature Fe2, and the third region feature F3 can then be input into a feature fusion network, which outputs a region mask image of the area. The feature fusion network consists of multiple operators, classifiers, convolutional layers, and deconvolutional layers. Figure 3 The diagram illustrates the structure of the feature fusion network. The first enhanced feature Fe1 is obtained by subtracting the crop growth change feature F3 from the field feature F1 at the seedling stage. The second enhanced feature Fe2 is obtained by multiplying the field feature F2 at the harvest stage with the crop growth change feature F3. The first enhanced feature, the second enhanced feature, and the third region feature are then input into the feature fusion network.

[0052] The first area feature and / or the second area feature at least includes color, texture, shape and other features of the area, and the third area feature represents growth change information of the crops in the area. Specifically, the first enhancement feature, the second enhancement feature and the third area feature can be stacked in the channel dimension and then input into the feature fusion network. In this way, multiple features can be fused together to extract higher-dimensional feature information.

[0053] After the three features are input into the feature fusion network, a loss function Loss2 is used as the loss function between the first enhancement feature Fe1 and the second enhancement feature Fe2, a loss function Loss1 is used as the loss function when the feature fusion network generates the area mask image, and a loss function Loss3 is used as the loss function between the first area feature and the second area feature. Specifically, the loss function used by the feature fusion network is as formula (1):

[0054] Loss = Loss1 + 0.5 * (Loss2 + Loss3) (1)

[0055] wherein Loss1 is the loss function when the feature fusion network generates the area mask image, the expression is as formula (2), Loss2 is the loss function between the first enhancement feature and the second enhancement feature, Loss3 is the loss function between the first area feature and the second area feature, and the expressions of Loss2 and Loss3 are as formula (3):

[0056]

[0057]

[0058] wherein y ij represents the true class of the i-th pixel of the input image for the feature fusion network, p ij is the predicted probability of the i-th pixel of the input image for the feature fusion network being class j, n is the total number of samples, k is the number of classes, y i represents the sign function, and when the true class of the i-th pixel is j, y i is 1, otherwise 0, and p iThis represents the predicted probability that the i-th pixel belongs to category j. In this scheme, the value of k can be 2, indicating that each pixel may have two categories: foreground and background. Correspondingly, the value of j can also be two, namely 1 and 0, or 1 and 2. The following uses the values ​​of j as 1 and 2 as examples. When the value of j is 1, it indicates that the category is foreground; when the value of j is 2, it indicates that the category is background. Conversely, when the value of j is 1, it indicates that the category is background; when the value of j is 2, it indicates that the category is foreground. It can be seen that this scheme combines the loss function used to calculate the field mask result with the loss function used to enhance the feature, which can suppress the overfitting phenomenon of the field mask. Among them, the region mask image is a binary image. When the pixel value is 0, it represents the background (background); when the pixel value is 1, it represents the field (foreground). Thus, the boundary lines of the fields can be further distinguished.

[0059] After obtaining the region mask image of the agricultural area, the boundary lines of the region can be determined based on this region mask image. Specifically, the findContours function in OpenCV can be used to extract the edge vector data of the mask image to obtain the vectorized boundary lines of the region.

[0060] In this solution, the region can refer to a field. The above technical solution integrates crop growth characteristics, field characteristics during the seedling stage, and field characteristics during the harvest stage to enhance the internal characteristics of the field, making the boundary between the field and the field more obvious and improving the extraction accuracy of the field boundary.

[0061] Figure 1 This is a flowchart illustrating a method for determining region boundaries in one embodiment. It should be understood that, although... Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise explicitly stated herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0062] This application provides a storage medium storing a program that, when executed by a processor, implements the method described above for determining region boundaries.

[0063] The embodiment of the present application provides a processor used for running a program, wherein the program is used for executing the method for determining the region boundary line.

[0064] In one embodiment, a device for determining a region boundary line is provided, comprising the processor, and the program of the processor is used for executing the method for determining the region boundary line.

[0065] The processor comprises a core, and the core is used for calling a corresponding program unit from the memory. The core can be one or more, and a method for determining a region boundary line is realized by adjusting core parameters.

[0066] The memory can comprise a non-permanent memory in a computer readable medium, a random access memory (RAM) and / or a non-volatile memory such as a read-only memory (ROM) or a flash memory (flash RAM), and the memory comprises at least one memory chip.

[0067] In one embodiment, a computer device is provided, which can be a server, and an internal structure diagram of the computer device can be as shown in Figure 4 The computer device comprises a processor A01, a network interface A02, a memory (not shown in the figure) and a database (not shown in the figure) connected through a system bus. The processor A01 of the computer device is used for providing computing and control capabilities. The memory of the computer device comprises an internal memory A03 and a non-volatile storage medium A04. The non-volatile storage medium A04 stores an operating system B01, a computer program B02 and a database (not shown in the figure). The internal memory A03 provides an environment for the operating system B01 and the computer program B02 in the non-volatile storage medium A04. The network interface A02 of the computer device is used for communicating with an external terminal through a network connection. The computer program B02 is executed by the processor A01 and realizes a method for determining a region boundary line.

[0068] Those skilled in the art can understand that Figure 4 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can comprise more or fewer components than those shown in the figure, or some components can be combined, or have a different component arrangement.

[0069] An embodiment of the present application provides a device, the device comprising a processor, a memory, and a program stored in the memory and capable of running on the processor, when the processor executes the program, the following steps are implemented: obtaining a first regional image and a second regional image of crops planted in a region in a first growth period and a second growth period respectively; performing matching calibration processing on the first regional image and the second regional image; inputting the processed first regional image and the processed second regional image into a regional feature extraction network respectively, to obtain a first regional feature and a second regional feature corresponding to the first regional image and the second regional image respectively; determining a plurality of vegetation indexes of the first regional image and the second regional image respectively; stacking a panchromatic band on the basis of the first regional image and the second regional image based on the plurality of vegetation indexes respectively, and simultaneously inputting the first regional image and the second regional image into a crop growth change feature extraction network, to obtain a third regional feature; processing the first regional feature, the second regional feature, and the third regional feature, to obtain a first enhanced feature and a second enhanced feature; inputting the first enhanced feature, the second enhanced feature, and the third regional feature into a feature fusion network, to obtain a regional mask image of the region; and determining a boundary line of the region through the regional mask image.

[0070] In one embodiment, processing the first regional feature, the second regional feature, and the third regional feature to obtain the first enhanced feature and the second enhanced feature comprises: subtracting the third regional feature from the first regional feature, to obtain a first enhanced feature of crop growth in the region; and multiplying the third regional feature by the second regional feature, to obtain a second enhanced feature of crop growth in the region.

[0071] In one embodiment, inputting the first enhanced feature, the second enhanced feature, and the third regional feature into the feature fusion network comprises: inputting the first enhanced feature, the second enhanced feature, and the third regional feature into the feature fusion network after stacking in a channel dimension.

[0072] In one embodiment, determining the boundary line of the region through the regional mask image comprises: extracting edge data of the regional mask image, to obtain a vectorized boundary line of the region.

[0073] In one embodiment, the first regional feature and / or the second regional feature at least comprises color, texture, shape, and the third regional feature represents growth change information of the crops.

[0074] In one embodiment, the plurality of vegetation indexes comprises NDVI and NDWI.

[0075] The present application also provides a computer program product, when executed on a data processing device, is adapted to execute a program initialized with the method steps for determining a boundary line of a region.

[0076] Those skilled in the art will appreciate that embodiments of the application can be readily used as a method, a system or a computer program product. Accordingly, the application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the application can take the form of a computer program product on one or more computer readable storage media (including, but not limited to, disk memory, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0077] The application is described in reference to the flowchart illustrations and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing device or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams.

[0078] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams.

[0079] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. ​ one or more functions specified in the flowchart illustrations and / or block diagrams.

[0080] In one typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0081] The memory can include non-persistent memory, Random Access Memory (RAM), and / or non-volatile memory, such as Read Only Memory (ROM) or flash memory, among others. The memory is an example of computer readable media.

[0082] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0083] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusions, so that a process, method, article or apparatus that includes a list of elements does not only include those elements, but also includes other elements not explicitly listed, or further includes elements inherent in such a process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0084] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the scope of claims of the present application.

Claims

1. A method for determining the boundary line of a region, characterized in that, The method includes: Acquire images of the first and second regions of the planted crops within the area when they are in their first and second growth stages, respectively; Perform matching calibration processing on the first region image and the second region image; The processed first region image and second region image are respectively input into the region feature extraction network to obtain the first region feature and second region feature corresponding to the first region image and the second region image, respectively. Multiple vegetation indices are determined for the first region image and the second region image, respectively; Based on the multiple vegetation indices, panchromatic bands are stacked on the first and second region images respectively, and the first and second region images are simultaneously input into the crop growth change feature extraction network to obtain the third region features. The features of the first region, the second region, and the third region are processed to obtain the first enhancement feature and the second enhancement feature. The first enhancement feature, the second enhancement feature, and the third region feature are input into a feature fusion network to obtain a region mask image of the region. The boundary line of the region is determined using the region mask image; The process of processing the first region features, the second region features, and the third region features to obtain the first enhancement feature and the second enhancement feature includes: Subtract the third region feature from the first region feature to obtain the first enhanced feature of crop growth in the region; The third region feature is multiplied by the second region feature to obtain a second enhanced feature of crop growth within the region.

2. The method for determining a region boundary line according to claim 1, characterized in that, Inputting the first enhanced feature, the second enhanced feature, and the third region feature into the feature fusion network includes: The first enhanced feature, the second enhanced feature, and the third region feature are stacked according to channel dimension and then input into the feature fusion network.

3. The method for determining a region boundary line according to claim 1, characterized in that, The loss function used by the feature fusion network is as shown in formula (1): Loss = +0.5*( + )(1) in, The loss function for generating region mask images by the feature fusion network is expressed as in formula (2). Let the loss function be the difference between the first enhancement feature and the second enhancement feature. Let be the loss function between the features of the first region and the features of the second region. and The expressions are all as shown in formula (3): (2) （3） in, This represents the true category of the i-th pixel in the input image for which the feature fusion network is located. The feature fusion network predicts the probability that the i-th pixel of the input image belongs to class j, where n is the total number of samples and k is the number of classes. Represented as a sign function, when the true class of the i-th pixel is j, The value is 1, and the value is 0 otherwise. This represents the predicted probability that the i-th pixel belongs to category j.

4. The method for determining a region boundary line according to claim 1, characterized in that, Determining the boundary line of the region using the region mask image includes: The edge data of the region mask image is extracted to obtain the vectorized boundary line of the region.

5. The method for determining a region boundary line according to claim 1, characterized in that, The first region image and the second region image are multispectral images, and the features of the first region and / or the second region include at least color, texture, and shape. The third region features characterize the growth and change information of the crop.

6. The method for determining a region boundary line according to claim 1, characterized in that, The vegetation indices include NDVI and NDWI.

7. A processor, characterized in that, It is configured to perform the method for determining the boundary line of a region as described in any one of claims 1 to 6.

8. An apparatus for determining the boundary line of a region, characterized in that, Includes the processor according to claim 7.

9. A machine-readable storage medium storing instructions thereon, characterized in that, When executed by a processor, this instruction causes the processor to be configured to perform the method for determining a region boundary line according to any one of claims 1 to 6.

Images