Method for extracting fuzzy boundary of hot area and hilly agricultural plot based on remote sensing image

Abstract

The invention relates to the technical field of land parcel boundary recognition and extraction, and particularly discloses a hot area and hilly agricultural land parcel fuzzy boundary extraction method based on remote sensing images, which comprises the following steps: S1, acquiring remote sensing image data, and constructing a land parcel boundary fuzzy index by using image processing, deep learning and a statistical model to obtain a remote sensing image data set; grading and zooming the image according to the index to obtain an image with obvious boundary features; s2, constructing a multi-branch cascade deep neural network model by using spectrum, texture and spatial information of the remote sensing image to perform land block boundary recognition; and S3, performing post-processing on an identification result by using an algorithm of maximum suppression binaryzation, connectivity detection and smoothness detection to obtain a clear boundary of the tropical region hilly cultivated land. According to the method for extracting the fuzzy boundary of the hot-area hilly agricultural land parcel based on the remote sensing image, intelligent extraction of the fuzzy boundary of the hilly agricultural land parcel can be realized.

Description

Technical Field

[0001] The invention relates to the technical field of land parcel boundary recognition and extraction, and in particular to a method for extracting fuzzy boundaries of hot zone hilly agricultural land parcels based on remote sensing images. Background Art

[0002] China is a major agricultural producer, and agriculture occupies a pivotal position in our national economy. With the vigorous development of agricultural informatization, promoting the construction of smart agriculture and precision agriculture is of great significance to improving my country's agricultural competitiveness. The vector boundary of a plot is an important feature necessary to determine the location, shape and spatial range of an agricultural unit, and is also the basis for achieving precision and large-scale agricultural production. For tropical crop planting areas in southern China, most of the crops in this area are planted in mountainous and hilly areas with fragmented cultivated land and complex topography (such as about 40% in Hainan and about 70% in Guangxi), which leads to unclear plot boundaries in remote sensing images of tropical areas, and also makes it difficult to identify various crop types in the images, affecting the relevant departments to grasp local agricultural information in a timely manner, and making it difficult to carry out smart management of plots, agricultural yield prediction, agricultural risk assessment, etc. Therefore, realizing accurate and rapid automatic extraction of fuzzy plot boundaries in complex environments such as hills and mountains is a key issue to be solved in the development of tropical agriculture.

[0003] However, the traditional manual digitization method is not only time-consuming, labor-intensive and costly to extract plot boundaries, but also requires the interpreter to have rich professional knowledge. With the rapid development of image processing and computer vision, remote sensing image segmentation technology and deep learning technology can quickly and accurately extract plot information. The plot boundary extraction method based on image segmentation mainly divides the image according to the grayscale, color, surface texture and geometric structure of the image. The extraction of plot boundaries using deep learning technology is a hot topic of research at home and abroad, and deep learning has further improved the extraction accuracy of plot boundaries. However, most of the existing research on plot boundary extraction is aimed at single crop plots in plain areas, while there is less research on the extraction of fuzzy plot boundaries in complex environments such as hills and mountains, especially the lack of research on tropical crop planting areas in southern my country. Summary of the invention

[0004] The present invention aims to solve at least one of the above-mentioned technical problems and provide a method for extracting fuzzy boundaries of hilly agricultural plots in hot areas based on remote sensing images, which can realize intelligent extraction of fuzzy boundaries of hilly agricultural plots.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is: a method for extracting fuzzy boundaries of hilly agricultural plots in hot areas based on remote sensing images, comprising the following steps:

[0006] Step S1, obtaining remote sensing image data, using image processing, deep learning, and statistical models to construct a fuzzy index for the land parcel boundary, and grading and scaling the image according to the index to obtain an image with obvious boundary features;

[0007] Step S2, using the spectrum, texture, and spatial information of remote sensing images, construct a multi-branch cascade deep neural network model to identify land parcel boundaries;

[0008] Step S3: Post-process the recognition results using the maximum suppression binarization, connectivity detection, and smoothness detection algorithms to obtain clear boundaries of hilly cultivated land in tropical areas.

[0009] Preferably, in step S1, the specific steps of constructing a fuzzy index of a land parcel boundary and grading and scaling the image according to the fuzzy index include:

[0010] S11. Cut the remote sensing image into image blocks of uniform size, calculate the matrix variance of the Laplace operator of all image blocks, and divide the image into two parts with clear boundaries and fuzzy boundaries;

[0011] S12. Use the image with clear boundary to train the RCF boundary recognition model, then infer the boundary fuzzy image based on the obtained model to obtain the boundary grayscale map, and analyze the grayscale value distribution of the boundary map of all images;

[0012] S13. Combining the Laplace variance and grayscale distribution, a fuzziness index is constructed using a statistical model. The calculation formula of the fuzziness index is as follows:

[0013] B=α×B G +(1-α)×B L (1)

[0014] Where B is the fuzzy index of the plot boundary; α is used to weight the grayscale distribution index and Laplace variance. In actual operation, α is taken as 0.7; B L is the Laplace blur index; B G is the grayscale fuzzy index, where B L The calculation formula is as follows:

[0015]

[0016] In the formula, B L is the Laplace blur index; σ 2 is the variance of the Laplace operator matrix of the image block; v max and v min are the maximum and minimum variance values ​​of the image set, respectively.

[0017] σ 2 The calculation formula is as follows:

[0018]

[0019] In the formula, σ 2 is the variance; M and N are the length and width of image I respectively; L is the matrix of image I and

[0020] The logarithmic function can compress large variance values ​​and make the changes smooth. Adding 1 is to avoid negative infinity or zero in logarithmic calculation. L The value range is between [0,1]. The closer the value is to 1, the clearer the image is; conversely, the closer the value is to 0, the blurrier the image is.

[0021] B G The calculation formula is as follows:

[0022]

[0023] In the formula, B G is the grayscale blur index; P left , P mid and P right They represent the peaks of the left, middle and right grayscale values ​​(0 to 255) in the grayscale histogram of the image respectively; C is a constant used to control the grayscale blur index to fall between [0,1], and is set to 9 based on experience in practice;

[0024] B G The closer the value is to 1, the clearer the image is; conversely, G The closer the value is to 0, the blurrier the image.

[0025] S14. According to the fuzzy index and the feedback of the model recognition, all images are graded, and the images with blurred land boundary are adaptively scaled according to the grade.

[0026] Preferably, the multi-branch cascade deep neural network model in step 2 performs land parcel boundary recognition, including the following steps:

[0027] S21. Based on the semantic segmentation network, Mamba is introduced to divide the image into subsets of cultivated land, forest land, grassland, garden land, water area, construction land and other land types. The division of other non-agricultural plots is conducive to improving the extraction accuracy of agricultural plot boundaries. For different subsets, multiple weights are trained using the edge extraction network, and the corresponding weights are adaptively selected for reasoning during reasoning;

[0028] S22. The edge extraction network adopts a dual-task branch structure and uses the boundary information and attribute information of the plot at the same time. The boundary information is used to supervise the edge extraction task, and the attribute information is used to supervise the semantic segmentation task.

[0029] S23. Use the trained fuzzy boundary extraction model library to determine the boundaries of plots of various types and extract the boundaries of hilly agricultural plots from them.

[0030] Preferably, the post-processing in step S3 specifically includes the following steps:

[0031] S31. Use the NMS maximum suppression algorithm to perform binarization, simplify the grayscale distribution, and refine the lines for subsequent processing;

[0032] S32. Perform connectivity detection based on the Region Growing algorithm, scan from the line breakpoints, and consider the directionality of fitting nearby pixels to achieve breakpoint continuity;

[0033] S33. The Douglas-Peucker algorithm is used to simplify the number of pixels of the line while maintaining the shape and features of the original curve, to achieve boundary smoothing, and finally to obtain the boundary of hilly cultivated land in tropical areas.

[0034] The beneficial effect is: compared with the prior art, the method for extracting the fuzzy boundaries of hilly agricultural plots in hot zones based on remote sensing images of the present invention adopts a deep learning method, based on high-resolution remote sensing images, through the development of an adaptive fuzzy boundary image scaling algorithm, the construction of a multi-branch cascade deep neural network, and the refinement of post-processing of the extraction results, etc., to fully extract and apply the spectral, texture, and spatial information of the image, and realize the intelligent extraction of the fuzzy boundaries of hilly cultivated land. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The specific embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings, wherein:

[0036] Figure 1 It is a flow chart of a method for extracting fuzzy boundaries of hilly agricultural plots in hot areas based on remote sensing images of the present invention;

[0037] Figure 2 The Laplace variance distribution of the three sampled images of different sizes of the present invention;

[0038] Figure 3 The image boundary clear and fuzzy partition result diagram of the present invention;

[0039] Figure 4 An example of a grayscale histogram of a clear and blurred image boundary of the present invention;

[0040] Figure 5 This is the result map of the extraction of the boundary of hilly cultivated land in tropical areas according to the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0042] It should be noted that when a component is referred to as being "fixed to" another component, it may be directly on the other component or there may be a central component. When a component is considered to be "connected to" another component, it may be directly connected to the other component or there may be a central component at the same time. When a component is considered to be "set on" another component, it may be directly set on the other component or there may be a central component at the same time. When a component is referred to as being "set in the middle", it does not only mean being set in the middle, as long as it is not set at both ends within the range defined by the middle. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0044] like Figure 1 As shown, the present application discloses a method for extracting fuzzy boundaries of hilly agricultural plots in hot areas based on remote sensing images, comprising the following steps:

[0045] Step S1, obtaining remote sensing image data, using image processing, deep learning, and statistical models to construct a fuzzy index for the land parcel boundary, and grading and scaling the image according to the index to obtain an image with obvious boundary features;

[0046] Step S2, using the spectrum, texture, and spatial information of remote sensing images, construct a multi-branch cascade deep neural network model to identify land parcel boundaries;

[0047] Step S3: Post-process the recognition results using the maximum suppression binarization, connectivity detection, and smoothness detection algorithms to obtain clear boundaries of hilly cultivated land in tropical areas.

[0048] Specifically, in step S1, the specific steps of constructing the fuzzy index of the land boundary and grading and scaling the image according to the fuzzy index include:

[0049] S11. Cut the remote sensing image into image blocks of uniform size (such as 1024*1024), calculate the matrix variance of the Laplace operator of all image blocks, and divide the image into two parts with clear boundaries and fuzzy boundaries;

[0050] S12. Use the image with clear boundary to train the RCF boundary recognition model, then infer the boundary fuzzy image based on the obtained model to obtain the boundary grayscale map, and analyze the grayscale value distribution of the boundary map of all images;

[0051] S13. Combining the Laplace variance and grayscale distribution, a fuzziness index is constructed using a statistical model. The calculation formula of the fuzziness index is as follows:

[0052] B=α×B G +(1-α)×B L (1)

[0053] Where B is the fuzzy index of the plot boundary; α is used to weight the grayscale distribution index and Laplace variance. In actual operation, α is taken as 0.7; B L is the Laplace blur index; B G is the grayscale fuzzy index, where B L The calculation formula is as follows:

[0054]

[0055] In the formula, B L is the Laplace blur index; σ 2 is the variance of the Laplace operator matrix of the image block; v max and v min are the maximum and minimum variance values ​​of the image set, respectively.

[0056] σ 2 The calculation formula is as follows:

[0057]

[0058] In the formula, σ 2 is the variance; M and N are the length and width of image I respectively; L is the matrix of image I and

[0059] The logarithmic function can compress large variance values ​​and make the changes smooth. Adding 1 is to avoid negative infinity or zero in logarithmic calculation. L The value range is between [0,1]. The closer the value is to 1, the clearer the image is; conversely, the closer the value is to 0, the blurrier the image is.

[0060] B G The calculation formula is as follows:

[0061]

[0062] In the formula, B G is the grayscale blur index; P left , P mid and P right They represent the peaks of the left, middle and right grayscale values ​​(0 to 255) in the grayscale histogram of the image respectively; C is a constant used to control the grayscale blur index to fall between [0,1], and is set to 9 based on experience in practice, and B G The closer the value is to 1, the clearer the image is; conversely, G The closer the value is to 0, the blurrier the image;

[0063] S14. According to the fuzzy index and the feedback of the model recognition, all images are graded, and the images with blurred land boundary are adaptively scaled according to the grade.

[0064] like Figure 2 In order to divide the image samples into two parts, one with clear boundaries and one with blurred boundaries, for the subsequent boundary extraction model training, the present invention first cuts the remote sensing image into image blocks of uniform size (such as 1024*1024), then calculates the matrix variance of the Laplace operator of all image blocks, and finally calculates the variance σ 2 and the maximum variance value v of the image set max and the minimum variance value v min , construct the Laplace fuzzy index B L Among them, the maximum variance value v max and the minimum variance value v min The variance of three images of different sizes is obtained by sampling and calculating, as shown in the figure.

[0065] like Figure 3 The image boundary clear and fuzzy partition results are shown. Laplace blur index B L The value range is between [0,1]. The closer the value is to 1, the clearer the image is; conversely, the closer the value is to 0, the blurrier the image is. L The calculation result can divide the image into two parts: clear boundary and fuzzy boundary, as shown in the figure. The image with fuzzy boundary cannot directly extract the boundary of the plot, and it is difficult to distinguish the specific content and distribution range of various land types; while the image with clear boundary can directly extract the boundary of the plot, and the distribution range of various land types is relatively clear. Dividing the image samples into fuzzy and clear can be used as the input of the subsequent boundary recognition model, providing a basis for constructing the fuzziness index.

[0066] like Figure 4The image shown is an example of a grayscale histogram with clear and blurred boundaries. The RCF boundary recognition model is trained using the image with clear boundaries, and then the blurred boundary image is inferred based on the obtained model to obtain a boundary grayscale map. As shown in the figure, the horizontal axis of the histogram represents the grayscale value (0 to 255), and the vertical axis represents the number of pixels with this grayscale value. Analysis of the grayscale value distribution of the image boundary map shows that the clearer the image, the more accurate the boundary extraction result, which is manifested in the grayscale map as fewer gray pixels and more black and white pixels. Conversely, the blurrier the image, the more gray pixels on the grayscale map. By extracting the peaks P of the left, middle, and right grayscale values ​​(0 to 255) in the image grayscale histogram left , P mid and P right Construct grayscale fuzzy index B G , which provides the premise for calculating the fuzzy index B of the plot boundary.

[0067] Specifically, in step 2, the multi-branch cascade deep neural network model performs land boundary recognition including the following steps:

[0068] S21. Based on the semantic segmentation network, Mamba is introduced to divide remote sensing images into subsets of land types such as cultivated land, forest land, grassland, garden land, water area, and construction land. The division of other non-agricultural plots is conducive to improving the extraction accuracy of agricultural plot boundaries. For different subsets, multiple weights are trained using the edge extraction network, and the corresponding weights are adaptively selected for reasoning during reasoning;

[0069] S22. The edge extraction network adopts a dual-task branch structure, using both the boundary information and attribute information of the land parcel. The boundary information is used to supervise the edge extraction task. Since all locations without edges belong to the same land object, the edge extraction task can alleviate the problem of fragmented segmentation results in the semantic segmentation task. The attribute information is used to supervise the semantic segmentation task. Since there is no edge inside the land object, the closer the location to the land object boundary is, the more likely it is to be an edge. Therefore, the semantic segmentation task can alleviate the pseudo-edge problem in the edge extraction task.

[0070] S23. Use the trained fuzzy boundary extraction model library to determine the boundaries of plots of various types and extract the boundaries of hilly agricultural plots from them.

[0071] Specifically, the post-processing in step S3 specifically includes the following steps:

[0072] S31. The boundary result of deep neural network recognition is a grayscale image, which needs to be post-processed before it can be used as a plot boundary. The NMS maximum suppression algorithm is used for binarization to simplify the grayscale distribution and refine the lines for subsequent processing;

[0073] S32. Perform connectivity detection based on the Region Growing algorithm, scan from the line breakpoints, and consider the directionality of fitting nearby pixels to achieve breakpoint continuity;

[0074] S33. The Douglas-Peucker algorithm is used to simplify the number of pixels of the line while maintaining the shape and features of the original curve, to achieve boundary smoothing, and finally to obtain the boundary of hilly cultivated land in tropical areas.

[0075] like Figure 5 The schematic diagram of the boundary extraction results of hilly cultivated land in tropical areas is shown in the figure. From the original image, it can be seen that the forest, grassland, cultivated land and other agricultural land in the area are cross-distributed and densely vegetated, which makes the boundaries of the plots unclear and difficult to distinguish the specific distribution range of different land types, and also interferes with the further identification of planted crops. After the trained neural network recognition, the boundaries of the plots in the image are initially obtained. Although the boundaries of the plots at this time have a certain outline, the boundary results are still a grayscale image and are relatively fuzzy. Therefore, based on the previous step, the maximum suppression binarization, connectivity detection, smoothness detection and other algorithms are further used to post-process the recognition results, and finally the clear boundaries of hilly cultivated land in tropical areas are obtained, which provides a basis for the accurate identification of various crop types, and is also conducive to the relevant departments in the region to grasp the local agricultural information in a timely manner and promote the intelligent management of plots.

[0076] The above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Any modification or equivalent substitution that does not depart from the spirit and scope of the present invention shall be included in the scope of the technical solution of the present invention.

Claims

1. A method for extracting fuzzy boundaries of hilly agricultural plots in hot areas based on remote sensing images, characterized in that: The following steps are involved: Step S1, obtaining remote sensing image data, using image processing, deep learning, and statistical models to construct a fuzzy index for the land parcel boundary, and grading and scaling the image according to the index to obtain an image with obvious boundary features; Step S2, using the spectrum, texture, and spatial information of remote sensing images, construct a multi-branch cascade deep neural network model to identify land parcel boundaries; Step S3: Post-process the recognition results using the maximum suppression binarization, connectivity detection, and smoothness detection algorithms to obtain clear boundaries of hilly cultivated land in tropical areas.

2. According to the method for extracting fuzzy boundaries of hot hilly agricultural plots based on remote sensing images according to claim 1, it is characterized in that: In step S1, the specific steps of constructing the fuzzy index of the land parcel boundary and grading and scaling the image according to the fuzzy index include: S11. Cut the remote sensing image into image blocks of uniform size, calculate the matrix variance of the Laplace operator of all image blocks, and divide the image into two parts with clear boundaries and fuzzy boundaries; S12. Use the image with clear boundary to train the RCF boundary recognition model, then infer the boundary fuzzy image based on the obtained model to obtain the boundary grayscale map, and analyze the grayscale value distribution of the boundary map of all images; S13. Combining the Laplace variance and grayscale distribution, a fuzziness index is constructed using a statistical model. The calculation formula of the fuzziness index is as follows: B=α×B G +(1-a)×B L (1) Where B is the fuzzy index of the plot boundary; α is used to weight the grayscale distribution index and Laplace variance. In actual operation, α is taken as 0.7; B L is the Laplace blur index; B G is the grayscale fuzzy index, where B L The calculation formula is as follows: In the formula, B L is the Laplace blur index; σ 2 is the variance of the Laplace operator matrix of the image block; v max and v min are the maximum and minimum variance values ​​of the image set, respectively. σ 2 The calculation formula is as follows: In the formula, σ 2 is the variance; M and N are the length and width of image I respectively; L is the matrix of image I and B L The value range of is between [0,1]. B G The calculation formula is as follows: In the formula, B G is the grayscale blur index; P left , P mid and P right They represent the peaks of the left, middle and right grayscale values ​​(0 to 255) in the image grayscale histogram respectively; C is a constant used to control the grayscale blur index to fall between [0,1]; S14. According to the fuzzy index and the model recognition feedback, all images are graded, and the images with blurred land boundary are adaptively scaled according to the grade.

3. The method for extracting fuzzy boundaries of hilly agricultural plots in hot areas based on remote sensing images according to claim 1 is characterized in that: In step 2, the multi-branch cascade deep neural network model performs land boundary recognition, including the following steps: S21. Based on the semantic segmentation network, Mamba is introduced to divide the remote sensing images into subsets of different land types. For different subsets, multiple weights are trained using the edge extraction network. During reasoning, the corresponding weights are adaptively selected for reasoning. S22. The edge extraction network adopts a dual-task branch structure and uses the boundary information and attribute information of the plot at the same time. The boundary information is used to supervise the edge extraction task, and the attribute information is used to supervise the semantic segmentation task. S23. Use the trained fuzzy boundary extraction model library to determine the boundaries of plots of various types and extract the boundaries of hilly agricultural plots from them.

4. The method for extracting fuzzy boundaries of hilly agricultural plots in hot areas based on remote sensing images according to claim 1 is characterized in that: The post-processing in step S3 specifically includes the following steps: S31. Use the NMS maximum suppression algorithm to perform binarization, simplify the grayscale distribution, and refine the lines for subsequent processing; S32. Perform connectivity detection based on the Region Growing algorithm, scan from the line breakpoints, and consider the directionality of fitting nearby pixels to achieve breakpoint continuity; S33. The Douglas-Peucker algorithm is used to simplify the number of pixels of the line while maintaining the shape and features of the original curve, to achieve boundary smoothing, and finally to obtain the boundary of hilly cultivated land in tropical areas.

Images