A method and device for extracting sub-field boundaries from drone images

Abstract

The present invention provides a method and device for extracting sub-field block boundaries from drone images, the method comprising: performing multi-scale segmentation on a second drone image to generate at least one sub-image; classifying the sub-images based on vegetation elevation information of each sub-image, spatial features of each sub-image, and spectral features of each sub-image to create a vegetation image set and a bare soil image set; merging the sub-images in the vegetation image set and merging the sub-images in the bare soil image set to form a sub-farmland field block image; and determining the initial sub-farmland boundary associated with the sub-farmland field block image. The present invention obtains vegetation elevation information through a digital surface model of a multi-temporal drone image, and uses the vegetation elevation information as the main feature, and classifies the segmented sub-images in combination with spectral features and spatial features, thereby generating farmland sub-field block images, which can effectively realize the extraction of sub-field block boundaries.

Description

Technical Field

[0001] The present invention relates to the field of agricultural remote sensing technology, and in particular to a method and device for extracting sub-field boundaries from unmanned aerial vehicle images. Background Art

[0002] "Precision agriculture" is based on the spatial differences of farmland crops and environment, and uses various technologies to obtain basic information of different units of farmland, thereby achieving optimized farmland management. Basic farmland information is the most important basic data for implementing "precision agriculture". Among them, the acquisition of spatial distribution information of farmland boundaries can assist in the input of agricultural materials and agricultural machinery scheduling for precision farming. However, the traditional manual digitization method to extract field boundaries is not only time-consuming and laborious, but also requires interpreters to have rich professional knowledge. In recent years, with the continuous increase of high-spatial-resolution satellites, the research on field boundary extraction using high-resolution satellite images has continued to increase.

[0003] The use of high spatial resolution satellite images for field boundary extraction is mainly used for field boundary extraction and monitoring of crop distribution and area. In plain areas, there are ridges inside farmland, which are divided into sub-fields. Since the width of the ridges between adjacent plots is narrow, centimeter-level resolution images are required to extract sub-field boundaries. Centimeter-level satellite images are commercial data, which need to be purchased and used and are expensive. UAVs are low-cost, flexible and can obtain centimeter-level resolution images, and have been widely used in various studies.

[0004] At present, most studies use spectral information and texture features to extract farmland plot boundaries, which makes it difficult to effectively extract sub-plot boundaries of farmland. Summary of the invention

[0005] The present invention provides a method and device for extracting sub-field boundaries from unmanned aerial vehicle images, which are used to solve the problem that it is difficult to effectively extract sub-field boundaries of farmland in the prior art.

[0006] In a first aspect, the present invention provides a method for extracting sub-field block boundaries from drone images, comprising: acquiring vegetation elevation information of a target farmland area based on a digital surface model of a first drone image and a digital surface model of a second drone image; performing multi-scale segmentation on the second drone image to generate at least one sub-image; classifying all sub-images based on the vegetation elevation information of each sub-image, the spatial characteristics of each sub-image, and the spectral characteristics of each sub-image to create a vegetation image set and a bare soil image set; merging all sub-images in the vegetation image set and merging all sub-images in the bare soil image set to form a farmland sub-field block image; determining an initial farmland sub-field block boundary associated with the farmland sub-field block image; the first drone image is a digital orthophoto image of the target farmland area during a bare period, and the second drone image is a digital orthophoto image of the target farmland area during a vegetation covered period.

[0007] According to a method for extracting sub-field plot boundaries from drone images provided by the present invention, after determining the initial farmland sub-field plot boundaries related to the farmland sub-field plot images, the method further includes: binarizing the sub-farmland plot images to generate a binary image; performing Hough transform on the binary image to obtain all boundary lines in the binary image; extending each of the boundary lines, and extracting the intersection of all the extended boundary lines as sub-farmland boundaries.

[0008] According to a method for extracting sub-field boundaries from drone images provided by the present invention, a Hough transform is performed on the binary image to obtain all boundary lines in the binary image, including:

[0009] Convert the binary image from the xy plane space to the ab parameter space;

[0010] The straight lines in the ab parameter space correspond one-to-one to the points in the xy plane space;

[0011] Determine a plurality of endpoints in the xy plane space according to a plurality of straight lines passing through the point (a, b) in the ab parameter space;

[0012] The boundary line is generated by sequentially connecting a plurality of the endpoints.

[0013] According to a method for extracting sub-field blocks from drone images provided by the present invention, multi-scale segmentation is performed on the second drone image to generate at least one sub-image, including: based on a Canny edge detection operator, the second drone image is segmented using pre-set segmentation parameters to generate at least one sub-image; the segmentation parameters include: segmentation scale, shape factor, and compactness factor.

[0014] According to a method for extracting sub-field blocks from drone images provided by the present invention, the vegetation elevation information of the target farmland area is obtained based on a digital surface model of a first drone image and a digital surface model of a second drone image, including: determining the vegetation elevation information based on a difference between the digital surface model of the first drone image and the digital surface model of the second drone image; the digital surface model of the first drone image includes bare soil elevation information and ridge elevation information; the digital surface model of the second drone image includes vegetation coverage area elevation information and ridge elevation information.

[0015] According to a method for extracting sub-field blocks from drone images provided by the present invention, before obtaining vegetation elevation information of a target farmland area based on a digital surface model of a first drone image and a digital surface model of a second drone image, the method further includes: respectively obtaining first initial image data and second initial image data of the target farmland area; respectively resampling the first initial image data and the second initial image data using a bilinear interpolation method; obtaining a digital surface model of the first drone image based on the resampled first initial image data; and obtaining a digital surface model of the second drone image based on the resampled second initial image data.

[0016] According to a method for extracting sub-field boundaries from drone images provided by the present invention, after extending the boundary straight line according to the slope of the boundary straight line and extracting intersecting straight lines to obtain the sub-farmland boundary, it also includes: converting the sub-farmland boundary into a vector file, performing remote sensing mapping, and displaying it on a platform.

[0017] In a second aspect, the present invention further provides a sub-field block boundary extraction device from drone images, comprising: a first module for acquiring vegetation elevation information of a target farmland area based on a digital surface model of a first drone image and a digital surface model of a second drone image;

[0018] A second module is used to perform multi-scale segmentation on the second drone image to generate at least one sub-image;

[0019] The third module is used to classify all sub-images based on the vegetation elevation information of each sub-image, the spatial characteristics of each sub-image and the spectral characteristics of each sub-image, and create a vegetation image set and a bare soil image set;

[0020] The fourth module is used to merge all the sub-images in the vegetation image set and merge all the sub-images in the bare soil image set to form a sub-agricultural field image;

[0021] A fifth module is used to determine initial farmland sub-field block boundaries associated with the farmland sub-field block images;

[0022] The first drone image is a digital orthophoto image of the target farmland area during the bare period, and the second drone image is a digital orthophoto image of the target farmland area during the vegetation covered period.

[0023] In a third aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of any of the above-mentioned methods for extracting sub-field boundaries from drone images are implemented.

[0024] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-described methods for extracting sub-field block boundaries from drone images.

[0025] The method and device for extracting sub-field plot boundaries from drone images provided by the present invention obtain vegetation elevation information through a digital surface model of multi-temporal drone images, and use the vegetation elevation information as the main feature, and classify the segmented sub-images in combination with spectral features and spatial features, thereby generating sub-agricultural field images, which can effectively realize the extraction of sub-field plot boundaries. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are 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.

[0027] Figure 1 It is a flow chart of the method for extracting sub-field block boundaries from drone images provided by the present invention;

[0028] Figure 2 It is a structural schematic diagram of a sub-field block boundary extraction device for drone images provided by the present invention;

[0029] Figure 3 A schematic structural diagram of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] It should be noted that in the description of the embodiments of the present invention, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "include one..." do not exclude the existence of other identical elements in the process, method, article or device including the elements. The orientation or position relationship indicated by the terms "upper", "lower" and the like is based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. Unless otherwise clearly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or it can be a connection between two elements. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0032] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0033] Combine the following Figure 1-Figure 3 The invention describes a method and device for extracting sub-field boundaries from drone images provided by an embodiment of the invention.

[0034] Figure 1 FIG. 1 is a flow chart of the method for extracting sub-field boundaries from drone images provided by the present invention. Figure 1 As shown, including but not limited to the following steps:

[0035] Step 101: Based on the digital surface model of the first drone image and the digital surface model of the second drone image, obtain vegetation elevation information of the target farmland area.

[0036] Among them, the first drone image is a digital orthophoto of the target farmland area during the bare period, and the second drone image is a digital orthophoto of the target farmland area during the vegetation coverage period.

[0037] The following is an explanation of the process of obtaining the first drone image and the second drone image:

[0038] The RGB image data of the target farmland area can be obtained by drone at the end of October and mid-April of the following year. At the end of October (bare period), the land is mainly bare, and in mid-April (vegetation covered period), most of the farmland plots have been covered by green vegetation as the seasons change. This period is suitable for extracting sub-farmland boundaries.

[0039] In order to ensure that the RGB image data of the UAV is obtained within a consistent range, the same route planning is fully utilized in the two route planning. During the route planning process, the same UAV take-off and landing points are set, and the shooting altitude is set to 200m. To ensure the image quality, the heading overlap and lateral overlap rates are set to 80%.

[0040] After the drone acquires the RGB image data, it can be exported to check whether its integrity and clarity meet the requirements of subsequent experiments. After checking the quality of the RGB image data, the acquired photos can be imported into the Pix4Dmapper software to set the image input and output.

[0041] Since the UAV positioning information adopts real-time difference and the coordinate system of the RGB image obtained by solution is CGCS2000, the input coordinate system needs to be unified with the actual coordinate system when setting the input coordinate system. The output coordinate system also selects CGCS2000, and then selects "3D Map" in the processing option template to finally generate a digital orthophoto map (Digital Orthophoto Map, DOM) and a digital surface model (Digital Surface Model, DSM). The spatial resolution of the above DOM and DSM maps can be set to 0.03m.

[0042] The above method can be used to obtain the DOM map of the target farmland area at the end of October and use it as the first UAV image, and the DOM map of the target farmland area and use it as the digital surface model of the first UAV image.

[0043] Based on the method in the above embodiment, a DOM map of the target farmland area in mid-April of the following year can also be obtained and used as the second drone image, and a DOM map of the target farmland area can be used as the digital surface model of the second drone image.

[0044] Optionally, by analyzing the difference in elevation information included in the digital surface model of the first drone image and the digital surface model of the second drone image, the vegetation elevation information of the target farmland area can be obtained.

[0045] Taking the target farmland area's wheat vegetation as an example, the plant height during the wheat filling period (vegetation coverage period) is quite different from the field ridge height, so it is easier to obtain characteristic vegetation elevation information. Therefore, processing the drone image during the vegetation coverage period, i.e., the second drone image, is more conducive to extracting the farmland sub-field boundary. The following steps are mainly the process of processing the second drone image.

[0046] Step 102: Perform multi-scale segmentation on the second drone image to generate at least one sub-image.

[0047] In order to better classify the types of objects in the second drone image, it is necessary to first segment the objects in the second drone image into different sub-images. For example, the above segmentation process can be achieved by using an edge detection operator.

[0048] Step 103: Classify all sub-images based on the vegetation elevation information of each sub-image, the spatial characteristics of each sub-image, and the spectral characteristics of each sub-image, and create a vegetation image set and a bare soil image set.

[0049] The vegetation elevation information of each sub-image mainly reflects the plant height characteristics of wheat, which is helpful to distinguish the wheat planting area from the bare soil area in the target farmland area.

[0050] Spatial features mainly include shape index and aspect ratio, etc. When the sub-images in the second drone image are regular rectangles or long strips, the differences between sub-images of different categories can be reflected by shape index and aspect ratio.

[0051] Vegetation images and bare soil images have different spectral characteristics. The red band mean of the second UAV image can be selected as the spectral feature to participate in the classification of the sub-images.

[0052] The present invention can use eCognition software, by taking vegetation elevation information as the main feature of threshold classification, and combining spatial features and spectral features to perform threshold classification on all sub-images, so that all sub-images can be divided into two categories of sub-images: vegetation images and bare soil images, and a vegetation image set and a bare soil image set can be created.

[0053] Step 104: Merge all sub-images in the vegetation image set and merge all sub-images in the bare soil image set to form a sub-agricultural field image.

[0054] After threshold classification, the sub-images may be missed or misclassified. The classification results can be refined using the rectangular fitting factor to reduce the misclassification and missed sub-images. The refined sub-images are then post-processed. The post-processing mainly involves merging sub-images of the same type and removing the influence of some small spots.

[0055] In the present invention, all sub-images in the vegetation image set are merged, and all sub-images in the bare soil image set are merged. Accordingly, the image generated after the vegetation image is merged is the vegetation area image, and the image generated after the bare soil image is merged is the background area image. The wheat area image and the background area image are combined to finally form a farmland sub-field image. The above-mentioned image merging process can also be completed with the help of eCognition software.

[0056] Step 105: Determine initial farmland sub-field boundaries associated with the sub-farmland field image.

[0057] For example, the farmland sub-field image is exported as a shapefile in the eCognition software and opened in the ArcGIS software. The shp layer is superimposed on the second drone image to view and analyze the extraction results of the initial sub-farmland boundary.

[0058] The method for extracting sub-field blocks from drone images provided by the present invention obtains vegetation elevation information through a digital surface model of multi-temporal drone images, takes the vegetation elevation information as the main feature, and classifies the segmented sub-images in combination with spectral features and spatial features, thereby generating farmland sub-field block images, which can effectively realize the extraction of sub-field block boundaries.

[0059] Based on the above embodiments, as an optional embodiment, the present invention provides a method for extracting sub-field block boundaries from drone images. After determining the initial farmland sub-field block boundaries related to the sub-farmland block images, the method further includes: binarizing the farmland sub-field block images to generate a binary image; performing Hough transform on the binary image to obtain all boundary lines in the binary image; extending each of the boundary lines, and extracting the intersection of all extended boundary lines as farmland sub-field block boundaries.

[0060] In order to further extract the boundaries of farmland sub-field blocks from farmland sub-field blocks, the present invention converts the vector shapefile file of the farmland sub-field block image into a raster tif file, and then binarizes the sub-farmland block image in tif format in Matlab to generate a binarized image.

[0061] The binary image only contains information of one channel and has a single pixel value. By comparing the brightness difference between adjacent pixels, the place where the brightness changes suddenly is the edge pixel. Connecting these edge pixels together forms an image with straight edge, which is more conducive to the subsequent extraction of boundary lines.

[0062] Based on the above embodiment, as an optional embodiment, the present invention performs Hough transform on the binary image to obtain a boundary line, including: converting the binary image from the xy plane space to the ab parameter space; the straight lines in the ab parameter space correspond one-to-one to the points in the xy plane space; according to multiple straight lines passing through the point (a, b) in the ab parameter space, determining multiple endpoints in the xy plane space; and sequentially connecting the multiple endpoints to generate the boundary line.

[0063] The present invention converts the binary image from the xy plane space to the ab parameter space, and the conversion formula is:

[0064] b=-ax+y

[0065] Among them, x and y are the coordinates in the xy plane space, a is the slope in the xy plane space, and b is the intercept in the xy plane space.

[0066] The following is a brief description of the principle and process of boundary line detection based on Hough transform:

[0067] Suppose the binary image is at a point (x i ,y i ), then after converting the xy plane space into the ab parameter space, the point (x i ,y i ) is converted into a straight line b = -ax in the ab parameter space i +y i .

[0068] Similarly, for another point (x i ,y i ), the corresponding straight line in the ab parameter space is b=-ax j +y j , and the above two lines intersect at point (a, b). That is, the points in the xy plane space correspond one to one with the lines in the ab parameter space.

[0069] Based on the above principle, if multiple straight lines (at least two) intersect at one point (a, b) in the ab parameter space, then multiple endpoints can be determined in the xy plane. By connecting the multiple endpoints in the xy plane, the boundary line in the xy plane space to be detected can be determined.

[0070] The more straight lines that pass through the point (a, b) in the ab parameter space, the longer the boundary line in the plane xy space. By identifying the number of straight lines passing through the point (a, b) in the ab parameter space, the length of the intercepted straight line can be adjusted, thereby further refining the boundary line, eliminating small edges that do not meet the conditions, and completing the detection of boundary lines based on Hough transform.

[0071] In addition, for the fuzzy areas where the ridges at the ends of the fields are worn away due to the turning of agricultural machinery, the boundary lines of the fuzzy areas may not be detected when the boundary lines are detected by the Hough transform, resulting in the missing of boundary lines and the extracted sub-farmland boundaries may not be closed.

[0072] The extension line of the incomplete boundary line caused by the fuzzy zone is usually nearly perpendicular to the boundary line of the farmland plot. The sub-farmland boundary can be fitted based on the above characteristics.

[0073] Optionally, each boundary line is extended according to the slope of each extracted boundary line, and the intersection of all extended boundary lines is extracted as the sub-farmland boundary.

[0074] The following is a brief introduction to the principle of how to extract the intersection of all boundary lines. The boundary lines are converted into polar coordinates using the following formula and described using distance-angle (ρ, θ).

[0075] ρ=xcosθ+ysinθ

[0076] Where x and y are the coordinates in the plane, θ is the angle, and ρ is the length of the boundary line.

[0077] A Gaussian kernel is defined at the starting point of the boundary line to be extended, and the changes of the Gaussian kernel around each point are checked in turn along the points on the boundary line. If there is a large mutation in the horizontal / vertical direction, and the change angle θ ≥ 45°, it indicates that the boundary line intersects with another boundary line, and the boundary line after the point corresponding to θ at this time will be truncated. Based on the method in the above embodiment, the above truncation operation is performed on all the boundary lines to be extended, and finally the sub-agricultural field boundary can be generated.

[0078] Optionally, since the unused portion of the ridge is usually larger than the width of the agricultural machinery, in order to ensure the fitting effect of the sub-field boundary, the length of the extension line of the boundary straight line is set to 8 meters in the present invention.

[0079] The present invention further performs boundary detection on the initial sub-farmland boundary obtained by Hough transform, and fits the sub-farmland boundary missing in the fuzzy zone, thereby generating a more complete and accurate sub-field block boundary.

[0080] Based on the above embodiments, as an optional embodiment, the method for extracting sub-field blocks from drone images provided by the present invention, before obtaining the vegetation elevation information of the target farmland area based on the digital surface model of the first drone image and the digital surface model of the second drone image, also includes: respectively obtaining the first initial image data and the second initial image data of the target farmland area; respectively resampling the first initial image data and the second initial image data using a bilinear interpolation method; obtaining the digital surface model of the first drone image based on the resampled first initial image data; and obtaining the digital surface model of the second drone image based on the resampled second initial image data.

[0081] According to research, sub-field boundaries can be effectively extracted when the spatial resolution is 10 cm. The present invention uses bilinear interpolation to resample the acquired first initial image data and second initial image data respectively, so that the spatial resolution of the resampled image data is 10 cm.

[0082] The resampled first initial image data and the second initial image data are processed by Pix4Dmapper software to generate a digital surface model of the first drone image and a digital surface model of the second drone image.

[0083] It should be noted that the first initial image data is the RGB image data of the target farmland area at the end of October, and the second initial image data is the RGB image data of the target farmland area in mid-April of the following year.

[0084] The present invention adjusts the spatial resolution of the initial image data by resampling the initial image data acquired by the unmanned aerial vehicle, which is more conducive to the extraction of the boundaries of farmland sub-fields.

[0085] Based on the above embodiments, as an optional embodiment, the present invention provides a method for extracting sub-field blocks from drone images, wherein the vegetation elevation information of the target farmland area is obtained based on the digital surface model of the first drone image and the digital surface model of the second drone image, including: determining the vegetation elevation information based on the difference between the digital surface model of the first drone image and the digital surface model of the second drone image; the digital surface model of the first drone image includes bare soil elevation information and ridge elevation information; the digital surface model of the second drone image includes vegetation coverage area elevation information and ridge elevation information.

[0086] The digital surface model of the first drone image and the digital surface model of the second drone image are obtained respectively. The digital surface model of the first drone image (denoted as DSM 1 ) Since it was obtained during the exposure period in October, DSM 1It mainly includes bare soil elevation information and field ridge elevation information, and DSM 1 As basic elevation information.

[0087] The digital surface model (DSM) of the second drone image 2 ) Since it was obtained in April when vegetation was covered, DSM 2 It mainly includes the elevation information of vegetation coverage areas and ridge elevation information.

[0088] The DSM is obtained by comparing the digital surface models of drone images from the above two periods. 0 The above process can be expressed mathematically as follows:

[0089] DSM 0 =DSM 2 -DSM 1

[0090] Then the obtained DSM 0 The ridge elevation information in DSM is close to 0. 0 The elevation information of the medium vegetation coverage area is greater than 0.

[0091] The present invention effectively extracts vegetation elevation information of vegetation-covered areas from drone images through digital surface models of drone images of two periods.

[0092] Based on the above embodiment, as an optional embodiment, the present invention performs multi-scale segmentation on the second UAV image to generate at least one sub-image, including: based on the Canny edge detection operator, the second UAV image is segmented using pre-set segmentation parameters to generate at least one sub-image; the segmentation parameters include: segmentation scale, shape factor and compactness factor.

[0093] The present invention can perform Canny operator edge detection on the R band of the second UAV image, and then use the image after Canny operator edge detection and the second UAV image to participate in image segmentation. The main segmentation parameters set during segmentation include segmentation scale, shape factor, compactness factor and weight factor.

[0094] When the spatial resolution is higher, the segmentation scale needs to be set larger, and when the spatial resolution is smaller, the segmentation scale needs to be set smaller. When determining the parameters of the segmentation scale, experiments were conducted from 10 to 100 with an interval of 10. After many attempts, the initial segmentation scale was initially selected as 30, the shape factor was 0.6, and the compactness was 0.5, and the ROC curve was obtained.

[0095] According to the peak information in the ROC curve and the ground features in the second drone image, the optimal segmentation scale for extracting the initial farmland sub-field boundaries is finally determined to be 70, and the shape factor and compactness factor are 0.6 and 0.5 respectively. The segmentation process of the second drone image can be completed in the eCognition software.

[0096] The present invention determines segmentation parameters suitable for extracting the boundaries of initial farmland sub-fields by optimizing the segmentation parameters, and realizes more accurate cutting of the second UAV image based on the optimized segmentation parameters.

[0097] Based on the above embodiments, as an optional embodiment, the sub-field boundary extraction method of drone images provided by the present invention, after extending the boundary straight line according to the slope of the boundary straight line and extracting the intersecting straight lines to obtain the sub-farmland boundary, also includes: converting the sub-farmland boundary into a vector file, performing remote sensing mapping, and displaying it on the platform.

[0098] The invention converts the sub-farmland boundary into a vector file through ArcGIS software and loads it into ArcGIS software, and the loading result is presented as a polygon with attributes. Then, the sub-field boundary is expressed by remote sensing mapping, and finally displayed on a plane according to a certain ratio and rules.

[0099] The present invention expresses the sub-field boundary through remote sensing mapping and displays it on a platform, which is more conducive to workers viewing the extraction results of the sub-field boundary.

[0100] Figure 2 Schematic diagram of the structure of the sub-field boundary extraction device of the drone image provided by the present invention. Figure 2 As shown, the device includes: a first module 201, a second module 202, a third module 203, a fourth module 204, and a fifth module 205.

[0101] The first module 201 is used to obtain vegetation elevation information of a target farmland area based on a digital surface model of a first drone image and a digital surface model of a second drone image;

[0102] The second module 202 is used to perform multi-scale segmentation on the second drone image to generate at least one sub-image;

[0103] The third module 203 is used to classify all sub-images based on the vegetation elevation information of each sub-image, the spatial characteristics of each sub-image and the spectral characteristics of each sub-image, and create a vegetation image set and a bare soil image set;

[0104] The fourth module 204 is used to merge all the sub-images in the vegetation image set and merge all the sub-images in the bare soil image set to form a farmland sub-field image.

[0105] The fifth module 205 is used to determine the initial farmland sub-field block boundary related to the farmland sub-field block image;

[0106] The first drone image is a digital orthophoto image of the target farmland area during the bare period, and the second drone image is a digital orthophoto image of the target farmland area during the vegetation covered period.

[0107] The device for extracting sub-field blocks from drone images provided by the present invention obtains vegetation elevation information through a digital surface model of multi-temporal drone images, takes the vegetation elevation information as the main feature, and classifies the segmented sub-images in combination with spectral features and spatial features, thereby generating farmland sub-field block images, which can effectively realize the extraction of sub-field block boundaries.

[0108] It should be noted that the drone image sub-field block boundary extraction device provided in the embodiment of the present invention can execute the drone image sub-field block boundary extraction method described in any of the above embodiments during specific operation, which will not be elaborated in this embodiment.

[0109] Figure 3 is a schematic diagram of the structure of the electronic device provided by the present invention, such as Figure 3As shown, the electronic device may include: a processor (processor) 310, a communication interface (Communications Interface) 320, a memory (memory) 330 and a communication bus 340, wherein the processor 310, the communication interface 320, and the memory 330 communicate with each other through the communication bus 340. The processor 310 can call the logic instructions in the memory 330 to execute the drone image sub-field block boundary extraction method, which includes: obtaining vegetation elevation information of the target farmland area based on the digital surface model of the first drone image and the digital surface model of the second drone image; performing multi-scale segmentation on the second drone image to generate at least one sub-image; classifying all sub-images based on the vegetation elevation information of each sub-image, the spatial characteristics of each sub-image and the spectral characteristics of each sub-image to create a vegetation image set and a bare soil image set; merging all sub-images in the vegetation image set and merging all sub-images in the bare soil image set to form a farmland sub-field block image; determining the initial farmland sub-field block boundary related to the farmland sub-field block image; the first drone image is a digital orthophoto image of the target farmland area in the bare period, and the second drone image is a digital orthophoto image of the target farmland area in the vegetation covered period.

[0110] In addition, the logic instructions in the above-mentioned memory 330 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0111] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the drone image sub-field block boundary extraction method provided by the above-mentioned methods, the method comprising: obtaining vegetation elevation information of the target farmland area based on the digital surface model of the first drone image and the digital surface model of the second drone image; performing multi-scale segmentation on the second drone image to generate at least one sub-image; classifying all sub-images based on the vegetation elevation information of each sub-image, the spatial characteristics of each sub-image and the spectral characteristics of each sub-image, and creating a vegetation image set and a bare soil image set; merging all sub-images in the vegetation image set and merging all sub-images in the bare soil image set to jointly form a farmland sub-field block image; determining the initial farmland sub-field block boundary related to the farmland sub-field block image; the first drone image is a digital orthophoto image of the target farmland area during the bare period, and the second drone image is a digital orthophoto image of the target farmland area during the vegetation covered period.

[0112] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the drone image sub-field block boundary extraction method provided in the above-mentioned embodiments, the method comprising: obtaining vegetation elevation information of the target farmland area based on the digital surface model of the first drone image and the digital surface model of the second drone image; performing multi-scale segmentation on the second drone image to generate at least one sub-image; classifying all sub-images based on the vegetation elevation information of each sub-image, the spatial characteristics of each sub-image and the spectral characteristics of each sub-image, and creating a vegetation image set and a bare soil image set; merging all sub-images in the vegetation image set and merging all sub-images in the bare soil image set to jointly constitute a farmland sub-field block image; determining the initial farmland sub-field block boundary related to the farmland sub-field block image; the first drone image is a digital orthophoto image of the target farmland area during the bare period, and the second drone image is a digital orthophoto image of the target farmland area during the vegetation covered period.

[0113] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0114] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for extracting sub-field boundaries from drone images, characterized in that: include: Obtaining vegetation elevation information of a target farmland area based on a digital surface model of the first drone image and a digital surface model of the second drone image; Performing multi-scale segmentation on the second drone image to generate at least one sub-image; Classifying all sub-images based on vegetation elevation information of each sub-image, spatial features of each sub-image, and spectral features of each sub-image, and creating a vegetation image set and a bare soil image set; Merging all sub-images in the vegetation image set and merging all sub-images in the bare soil image set to form a farmland sub-field image; Determining initial farmland sub-plot boundaries associated with the farmland sub-plot image; The first drone image is a digital orthophoto image of the target farmland area during the bare period, and the second drone image is a digital orthophoto image of the target farmland area during the vegetation covered period.

2. The method for extracting sub-field boundaries from drone images according to claim 1, characterized in that: After determining the initial farmland sub-field block boundaries related to the farmland sub-field block images, the method further includes: Binarizing the farmland sub-field image to generate a binary image; Performing Hough transform on the binary image to obtain all boundary lines in the binary image; Each of the boundary lines is extended, and the intersection of all the extended boundary lines is extracted as the sub-farmland boundary.

3. The method for extracting sub-field boundaries from drone images according to claim 2, characterized in that: Performing Hough transform on the binary image to obtain all boundary lines in the binary image, including: Convert the binary image from the xy plane space to the ab parameter space; The straight lines in the ab parameter space correspond one-to-one to the points in the xy plane space; Determine a plurality of endpoints in the xy plane space according to a plurality of straight lines passing through the point (a, b) in the ab parameter space; The boundary line is generated by sequentially connecting a plurality of the endpoints.

4. The method for extracting sub-field boundaries from drone images according to claim 1, characterized in that: Performing multi-scale segmentation on the second drone image to generate at least one sub-image includes: Based on the Canny edge detection operator, segment the second UAV image using preset segmentation parameters to generate at least one sub-image; The segmentation parameters include: segmentation scale, shape factor and compactness factor.

5. The method for extracting sub-field boundaries from drone images according to claim 1, characterized in that: The obtaining of vegetation elevation information of the target farmland area based on the digital surface model of the first drone image and the digital surface model of the second drone image includes: Determining the vegetation elevation information based on a difference between the digital surface model of the first drone image and the digital surface model of the second drone image; The digital surface model of the first drone image includes bare soil elevation information and field ridge elevation information; The digital surface model of the second UAV image includes vegetation coverage area elevation information and field ridge elevation information.

6. The method for extracting sub-field boundaries from drone images according to claim 1, characterized in that: Before obtaining vegetation elevation information of the target farmland area based on the digital surface model of the first drone image and the digital surface model of the second drone image, the method further includes: Acquire first initial image data and second initial image data of the target farmland area respectively; Resampling the first initial image data and the second initial image data respectively by using a bilinear interpolation method; Acquire a digital surface model of the first UAV image based on the resampled first initial image data; A digital surface model of the second UAV image is obtained based on the resampled second initial image data.

7. The method for extracting sub-field boundaries from drone images according to claim 2, characterized in that: After extending the boundary straight line according to the slope of the boundary straight line and extracting intersecting straight lines to obtain the farmland sub-field boundary, the method further includes: The farmland sub-field boundaries are converted into vector files, remote sensing mapping is performed, and displayed on the platform.

8. A device for extracting sub-field boundaries from drone images, characterized in that: include: The first module is used to obtain vegetation elevation information of a target farmland area based on a digital surface model of the first drone image and a digital surface model of the second drone image; A second module is used to perform multi-scale segmentation on the second drone image to generate at least one sub-image; The third module is used to classify all sub-images based on the vegetation elevation information of each sub-image, the spatial characteristics of each sub-image and the spectral characteristics of each sub-image, and create a vegetation image set and a bare soil image set; The fourth module is used to merge all the sub-images in the vegetation image set and merge all the sub-images in the bare soil image set to form a farmland sub-field image; A fifth module is used to determine the initial sub-farmland boundary associated with the farmland sub-field image; The first drone image is a digital orthophoto image of the target farmland area during the bare period, and the second drone image is a digital orthophoto image of the target farmland area during the vegetation covered period.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for extracting sub-field blocks from drone images as described in any one of claims 1 to 7 are implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for extracting sub-field blocks from drone images as described in any one of claims 1 to 7 are implemented.

Images