An automatic identification and segmentation system for planting fields based on remote sensing images

Abstract

The present invention relates to the field of image feature analysis technology, specifically to a planting field automatic identification and segmentation system based on remote sensing images, and the planting field automatic identification and segmentation system based on remote sensing images includes an image texture feature extraction module, a morphology and segmentation intelligent processing module, a boundary refinement correction module and a plot information integration module. In the present invention, by combining the application of shape recognition logic with advanced edge detection technology, accurate matching and classification of plot morphology are achieved, the accuracy of recognition is improved and manual intervention is reduced, thereby enhancing the automation capability of the system, and computational geometry technology is used to accurately correct the boundaries of the initially segmented plots, which not only optimizes the accuracy of the plot boundaries, but also improves the accuracy of planting structure analysis and resource management, significantly improving the practicality and efficiency of remote sensing data in agricultural production monitoring and management, ensuring the real-time and accuracy of data support, and greatly optimizing the management efficiency and accuracy of planting fields.

Description

Technical Field

[0001] The present invention relates to the technical field of image feature analysis, and in particular to an automatic identification and segmentation system for planting fields based on remote sensing images. Background Art

[0002] The field of image feature analysis technology involves identifying, extracting and analyzing features in digital images in order to obtain useful information from them, including the analysis and processing of features such as texture, shape, color, edge, pattern, etc. It is widely used in pattern recognition, computer vision, remote sensing, medical imaging, video surveillance and autonomous driving. Through image feature analysis, functions such as target detection, object recognition, image classification, image segmentation and image matching can be realized, providing technical support and data basis for a variety of practical applications.

[0003] Among them, the automatic identification and segmentation system of planting fields based on remote sensing images is specifically used in agricultural remote sensing. The core is to use the image features in remote sensing images to automatically identify and segment planting fields. By analyzing remote sensing images, the farmland areas can be accurately identified and segmented, thereby realizing the monitoring of agricultural production conditions, the analysis of crop planting structure and the rational management of agricultural resources. It can improve the efficiency and accuracy of agricultural management and provide important data support for agricultural production.

[0004] Existing technologies are not good enough in automatically identifying and segmenting complex agricultural environments. Traditional image processing technology faces challenges of misjudgment and long processing time in identifying diverse crop planting structures, especially when distinguishing crop types with similar appearances. In addition, existing technologies are also unable to handle the fine-grained processing of plot boundaries. Minor errors in boundaries cannot be effectively controlled, often leading to inaccurate identification and segmentation of cultivated fields, affecting the overall efficiency of agricultural production. At the same time, the lack of application of advanced computational geometry and morphological analysis tools limits its application potential in modern precision agriculture, making the allocation and management of agricultural resources suboptimal. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a planting field automatic identification and segmentation system based on remote sensing images.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: A planting field automatic identification and segmentation system based on remote sensing images comprises:

[0007] The image texture feature extraction module collects remote sensing images, separates texture layers through wavelet transform, analyzes the directionality, roughness and contrast of textures in image data, integrates texture data, and generates texture feature mapping results;

[0008] The morphology and segmentation intelligent processing module applies shape recognition logic to identify the differential land parcel morphology according to the texture feature mapping result, matches and classifies the land parcel morphology through Hough transform, merges the classification results, and generates land parcel morphology recognition data and preliminary land parcel segmentation results;

[0009] The boundary refinement correction module performs edge detection based on the preliminary plot segmentation result, identifies boundary errors, corrects the plot boundary using computational geometry technology, adjusts boundary data, and generates a plot boundary adjustment result;

[0010] The plot information integration module combines the plot boundary adjustment results with the plot morphology recognition data, analyzes the usage status of each plot, verifies the matching degree between the plot boundary and the crop planting situation, optimizes the division accuracy, and generates the final segmentation result of the planting field.

[0011] As a further solution of the present invention, the step of separating the texture layer is:

[0012] Collect remote sensing image data, obtain original images from satellites or aerial photography equipment, and generate original image data results;

[0013] Applying wavelet transform to the original image data result,

[0014]

[0015] Among them, ψ w,λ Represents the wavelet function, which is used to extract image features at different scales w and positions λ. w and λ represent the scale and position parameters of the wavelet transform, respectively. I(x, y) represents the pixel value of the image data. σ 2 is the noise variance, which is used to improve the anti-noise performance of wavelet transform and generate wavelet transform processing results;

[0016] Separate the texture layer from the wavelet transform processing result and apply the formula,

[0017]

[0018] Among them, T(w,λ) is the data value of the wavelet transform result, indicating the possibility that each pixel belongs to the target texture layer, e is the natural logarithm base, and α is the adjustment parameter used to adjust the sensitivity of logistic regression to generate the texture layer separation result.

[0019] As a further solution of the present invention, the step of obtaining the texture feature mapping result is:

[0020] Extract texture features from the texture layer separation results, calculate the directionality, roughness and contrast of each separated texture layer, and use the formula:

[0021] Fdir (L) = ∑L sin(θ) cos(φ)

[0022] Determine the directionality, where L is the pixel value in the texture layer separation result, θ is the direction angle of the target pixel, and φ is the direction offset angle, increasing the dimension of the calculation and generating feature analysis results;

[0023] The characteristic analysis results are integrated and the formula is applied.

[0024]

[0025] The influence of directionality, roughness and contrast is integrated, where k, m, and n are weight coefficients, which adjust the influence proportion of directionality, roughness and contrast in the total score respectively to generate the texture data integration result;

[0026] According to the texture data integration result, a texture feature map is created, and a mapping formula is used to perform standardized feature expression.

[0027]

[0028] Among them, M(x) is the texture feature mapping value, S is the texture data integration result, max(S) is the maximum value in S, ∈ is a small constant to avoid division by zero error, and the texture feature mapping result is generated.

[0029] As a further solution of the present invention, the steps for obtaining the land parcel morphology recognition data are as follows:

[0030] Analyze the texture feature mapping result, calculate the shape parameters a, b, c of the plot, where a represents the side length ratio, b represents the angle deviation, and c represents the area ratio, and generate the morphological feature parameter result;

[0031] Combined with the morphological characteristic parameter results, quantitative analysis is performed.

[0032] R(a,b,c)=a 2 ÷(b+g)+c×h

[0033] Among them, R(a, b, c) is the score of quantifying the plot morphology, g and h are adjustment parameters used to adjust the influence of angle deviation and area ratio to generate morphological recognition analysis results;

[0034] According to the morphological recognition analysis results, the differences of all plots are distinguished to generate plot morphological recognition data.

[0035] As a further solution of the present invention, the steps for obtaining the preliminary land parcel segmentation result are:

[0036] Combined with the land morphology recognition data, Hough transform is performed.

[0037] H(θ,ρ)=∑ x,y f(x,y)·|ρ-xcos(θ)-ysin(θ)| 0.5

[0038] Among them, θ is the angle parameter, ρ is the distance parameter, || 0.5 It is used to improve the sensitivity to small errors. f(x,y) is the morphological feature value of the land parcel morphological recognition data, and generates the Hough transform matching result;

[0039] Analyze the Hough transform matching results and use the classification formula:

[0040]

[0041] Where C(i) represents the highest classification score obtained after classifying the i-th plot, which is used to determine the final category of the plot, j represents the index of the classification, and w k Weight parameter, weighting the impact of the difference feature, H ijk It represents the Hough transform result contributed by the kth feature of the i-th plot in the j-th category. The natural logarithm base of e is used for logical function calculation to generate the classification processing result.

[0042] The classification results are combined to summarize similar plots and generate preliminary plot segmentation results.

[0043] As a further solution of the present invention, the step of identifying the boundary error is:

[0044] Receive and process the preliminary plot segmentation result, determine the boundary of each plot, and generate preliminary plot boundary data result;

[0045] Based on the preliminary plot boundary data results, edge detection technology is applied to

[0046]

[0047] Among them, E(x,y) represents the edge strength, and They represent the gradient of the image brightness function f(x,y) in the x direction and the y direction respectively, ∈ is a small constant used to enhance the detection sensitivity of tiny edges, f(x,y) is the image brightness, and generates edge detection results;

[0048] Compare the edge detection result with the preset plot boundary and calculate the deviation,

[0049]

[0050] Among them, D represents the average boundary error, detected_edge(x,y) represents the detected edge position, expected_edge(x,y) represents the expected edge position, and N represents the total number of boundary points to generate boundary error data.

[0051] As a further solution of the present invention, the steps for obtaining the land parcel boundary adjustment result are:

[0052] Analyze the boundary error data, apply computational geometry techniques, and use the error minimization formula.

[0053]

[0054] Among them, L opt represents the sum of squares of the total error, ω i Represents the weight adjusted according to the characteristics of the plot, which is used to improve the accuracy of the correction. i,actual and P i,expected Respectively represent the actual and expected boundary points, and generate the optimization results of the land parcel boundary;

[0055] Update the boundary data of the plot according to the plot boundary optimization result, match the boundary with the actual demand, and adopt the update formula.

[0056] P new =P actual +β·(P expected -P actual )

[0057] Among them, P actual represents the actual detected boundary point position, P expected It represents the expected boundary point location, which is determined based on planning and design standards. β represents the adjustment coefficient, which controls the adjustment range and has a value range between 0 and 1. It generates updated boundary data.

[0058] The updated boundary data results are integrated to form a land parcel boundary adjustment result, including the adjustment process and final result of each land parcel boundary.

[0059] As a further solution of the present invention, the steps for obtaining the final segmentation result of the planting field are:

[0060] Integrate the plot boundary adjustment result with the plot morphology recognition data, match each plot data, and generate an integrated plot data set;

[0061] Using the integrated plot data set, calculate the usage index U(i) of each plot,

[0062]

[0063] Among them, ak represents the attribute value of plot i at feature k, including soil quality score, distance to water source or other plot characteristics, w k is the weight coefficient of the target feature, which is used to adjust the influence of each attribute in the total usage index and generate the usage index of each plot;

[0064] The matching degree between the boundary and crop planting conditions of each plot is analyzed using the matching degree analysis formula.

[0065]

[0066] Among them, M(i) represents the matching degree, d j is a binary value indicating whether the crop planting situation matches the expected situation, 1 indicates a match, 0 indicates a mismatch, and λ j is the impact factor or area weight of region j, which is used to improve the accuracy of the analysis and generate the matching analysis results;

[0067] According to the utilization index and matching degree analysis results of each plot, the boundaries of the plots are finally adjusted and optimized to ensure that the division of each plot matches the actual needs of crop planting and generate the final segmentation results of the planting fields.

[0068] Compared with the prior art, the advantages and positive effects of the present invention are:

[0069] In the present invention, the application of shape recognition logic is combined with advanced edge detection technology to achieve accurate matching and classification of plot morphology, improve recognition accuracy and reduce manual intervention, thereby enhancing the automation capability of the system, and use computational geometry technology to perform precise boundary correction on the initially segmented plots, which not only optimizes the accuracy of plot boundaries, but also improves the accuracy of planting structure analysis and resource management, significantly improves the practicability and efficiency of remote sensing data in agricultural production monitoring and management, ensures the real-time and accuracy of data support, and greatly optimizes the management efficiency and accuracy of planting plots. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 is a system flow chart of the present invention;

[0071] Figure 2 The flowchart of separating the texture layer of the present invention is as follows;

[0072] Figure 3 A flowchart for obtaining texture feature mapping results of the present invention;

[0073] Figure 4 This is a flow chart for obtaining land parcel morphology recognition data of the present invention;

[0074] Figure 5This is a flow chart for obtaining the preliminary land parcel segmentation result of the present invention;

[0075] Figure 6 This is a flow chart for identifying boundary errors of the present invention;

[0076] Figure 7 A flowchart for obtaining the land boundary adjustment result of the present invention;

[0077] Figure 8 The present invention is a flow chart for obtaining the final segmentation result of the planting field. DETAILED DESCRIPTION

[0078] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0079] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are 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 limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0080] Embodiment 1

[0081] See also Figure 1 , a planting field automatic identification and segmentation system based on remote sensing images includes:

[0082] The image texture feature extraction module collects remote sensing images, separates texture layers through wavelet transform, analyzes the directionality, roughness and contrast of textures in image data, integrates texture data, and generates texture feature mapping results;

[0083] The morphology and segmentation intelligent processing module uses shape recognition logic to identify differential plot morphology based on texture feature mapping results, matches and classifies plot morphology through Hough transform, merges classification results, and generates plot morphology recognition data and preliminary plot segmentation results;

[0084] The boundary refinement and correction module performs edge detection based on the preliminary plot segmentation results, identifies boundary errors, uses computational geometry technology to correct the plot boundaries, adjusts boundary data, and generates plot boundary adjustment results;

[0085] The plot information integration module combines the plot boundary adjustment results with the plot morphology recognition data, analyzes the usage status of each plot, verifies the matching degree between the plot boundary and the crop planting situation, optimizes the division accuracy, and generates the final segmentation result of the planting field.

[0086] The texture feature mapping results include texture orientation map, texture roughness score and contrast level. The plot morphology recognition data include the recognized shape type, measured size range and plot center coordinates. The preliminary plot segmentation results include preliminary plot boundary configuration, preliminary plot segmentation area number and area estimation results of each area. The plot boundary adjustment results include the adjusted plot boundary line, corrected plot vertex position and plot boundary alignment error analysis results. The final segmentation results of the planting field include the ultimate partition line of the planting field, partition type identification and accuracy verification results.

[0087] See also Figure 2 , the steps to separate the texture layer are:

[0088] Collect remote sensing image data, obtain original images from satellites or aerial photography equipment, and generate original image data results;

[0089] Applying wavelet transform to the raw image data results in

[0090]

[0091] Among them, ψ w,λ Represents the wavelet function, which is used to extract image features at different scales w and positions λ. w and λ represent the scale and position parameters of the wavelet transform, respectively. I(x, y) represents the pixel value of the image data. σ 2 is the noise variance, which is used to improve the anti-noise performance of wavelet transform and generate wavelet transform processing results;

[0092] Separate the texture layer from the wavelet transform processing result and apply the formula,

[0093]

[0094] Among them, T(w,λ) is the data value of the wavelet transform result, indicating the possibility that each pixel belongs to the target texture layer, e is the natural logarithm base, and α is the adjustment parameter used to adjust the sensitivity of logistic regression to generate the texture layer separation result.

[0095] Assume that we are dealing with a simple 2x2 pixel image region, whose pixel values ​​are:

[0096]

[0097] Assume that the wavelet function ψ w,λ The response at this scale and location is:

[0098]

[0099] Assume σ 2 With a value of 0.01 (this is a common noise variance assumption in wavelet transforms), the wavelet transform value at each position is calculated:

[0100]

[0101] α is an integer parameter used to adjust the sensitivity of logistic regression. Assuming it is 0.05, substitute the result of T(w,λ)=484.8 calculated above into the formula to obtain:

[0102]

[0103] L(w,λ) is the result after wavelet transform and logistic regression model processing. The result is a probability value between 0 and 1, which is used to indicate the possibility that each pixel belongs to the target texture layer.

[0104] See also Figure 3 , the steps to obtain the texture feature mapping results are:

[0105] Extract texture features from the texture layer separation results, calculate the directionality, roughness and contrast of each separated texture layer, and use the formula,

[0106] F dir (L) = ∑L·sin(θ)·cos()

[0107] Determine the directionality, where L is the pixel value in the texture layer separation result, θ is the direction angle of the target pixel, and is the direction offset angle, increasing the dimension of the calculation and generating the feature analysis result;

[0108] Integrate the feature analysis results and apply the formula,

[0109]

[0110] The influence of directionality, roughness and contrast is integrated, where k, m, and n are weight coefficients, which adjust the influence proportion of directionality, roughness and contrast in the total score respectively to generate the texture data integration result;

[0111] According to the texture data integration results, a texture feature map is created, and the mapping formula is used to express the standardized features.

[0112]

[0113] Among them, M(x) is the texture feature mapping value, S is the texture data integration result, max(S) is the maximum value in S, and a small constant is used to avoid division by zero errors to generate the texture feature mapping result.

[0114] Assume that the pixel direction angle θ is 45° and the direction offset angle is 30°.

[0115] Assume that the texture layer separation result L is:

[0116]

[0117] Using the assumed angles θ = 45° and θ = 30° (converted to radians for calculation):

[0118] sin(45°)≈0.707,cos(30°)≈0.866

[0119] Calculate F dir (L),

[0120] F dir (L)=0.9·0.707·0.866+0.85·0.707·0.866

[0121] +0.95 0.707 0.866+0.92 0.707 0.866

[0122] F dir (L)=1.653

[0123] Assume k = 0.3, m = 0.4, n = 0.3,

[0124] Assuming the roughness and contrast are 0.7 and 0.8 respectively, then,

[0125] S=0.3·1.653 2 +0.4 0.7 2 +0.3 0.8 2

[0126] =0.3·2.7324+0.4·0.49+0.3·0.64

[0127] =0.81972+0.196+0.192=1.20772

[0128] Assume 0.001 to avoid division by zero error,

[0129] Assuming max(S) is 2.0 in S, using S = 1.20772, then,

[0130]

[0131] The mapping result is obtained by comprehensively considering the directionality, roughness and contrast characteristics of the texture, standardizing the calculated comprehensive eigenvalue S, and generating a texture feature mapping value for each pixel or area to describe the texture characteristics of each pixel or area in the image.

[0132] See also Figure 4 , the steps to obtain the land parcel morphology recognition data are:

[0133] Analyze the texture feature mapping results, calculate the shape parameters a, b, c of the plot, where a represents the side length ratio, b represents the angle deviation, and c represents the area ratio, and generate the morphological feature parameter results;

[0134] Combined with the morphological characteristic parameter results, quantitative analysis is performed.

[0135] R(a,b,c)=a 2 ÷(b+g)+c×h

[0136] Among them, R(a, b, c) is the score of quantifying the plot morphology, g and h are adjustment parameters used to adjust the influence of angle deviation and area ratio to generate morphological recognition analysis results;

[0137] According to the results of morphological recognition analysis, the differences of all plots are distinguished and the plot morphological recognition data is generated.

[0138] Parameter Description:

[0139] a: indicates the length ratio of the plot. Assuming the length and width of a plot are 10m and 5m, then

[0140] b: represents the angle deviation. For example, if a plot deviates 30 degrees from the true north, then b = 30;

[0141] c: represents the area ratio. Assuming that the area ratio of this plot to the other plot is 1.5, then c = 1.5;

[0142] g and h are adjustment parameters, assuming g = 0.1 and h = 0.5, to make the formula appropriately sensitive to angle deviation and area ratio.

[0143] Substituting a=2, b=30 and c=1.5 into the formula,

[0144]

[0145] A score of 0.883 indicates that the plot has a high degree of conformity in terms of morphological characteristics, and a score close to 1 indicates that the plot is highly suitable under a given evaluation model, facilitating subsequent classification.

[0146] See also Figure 5, the steps to obtain the preliminary plot segmentation results are:

[0147] Combined with the land morphology recognition data, Hough transform is performed.

[0148] H(θ,ρ)=∑ x,y f(x,y)·|ρ-xcos(θ)-ysin(θ)| 0.5

[0149] Among them, θ is the angle parameter, ρ is the distance parameter, || 0.5 It is used to improve the sensitivity to small errors. f(x,y) is the morphological feature value of the land parcel morphological recognition data, and generates the Hough transform matching result;

[0150] Analyze the Hough transform matching results and use the classification formula.

[0151]

[0152] Where C(i) represents the highest classification score obtained after classifying the i-th plot, which is used to determine the final category of the plot, j represents the index of the classification, and w k The weight parameter weights the impact of the difference feature.

[0153] H ijk It represents the Hough transform result contributed by the kth feature of the i-th plot in the j-th category. The natural logarithm base of e is used for logical function calculation to generate the classification processing result.

[0154] The classification results are merged to summarize similar plots and generate preliminary plot segmentation results.

[0155] Parameter Description:

[0156] θ: Angle parameter, used to determine the inclination angle of the straight line in Hough transform;

[0157] ρ: distance parameter, used to determine the length of a straight line from the origin of the coordinate system in Hough transform;

[0158] f(x,y): morphological feature value of land parcel morphological recognition data, at the image coordinate (x,y), it is assumed that 1 represents the presence of features and 0 represents the absence of features.

[0159] The constant 0.5 is used to increase sensitivity to small errors.

[0160] Assumptions:

[0161] θ = 45° (needs to be converted to radians for calculation);

[0162] ρ = 100;

[0163] f(10,20)=1 (at image coordinate (10,20), indicating a feature)

[0164] Assume x = 10, y = 20, and use θ = 45° (that is, θ ≈ 0.785 radians):

[0165] H(θ,ρ)=f(10,20)·|100-10cos(0.785)-20sin(0.785)| 0.5

[0166] cos(0.785)≈0.707, sin(0.785)≈0.707

[0167] H(θ,ρ)=1·|100-10×0.707-20×0.707| 0.5 ≈|100-21.21| 0.5 ≈8.88

[0168] H(θ,ρ)=8.88 indicates that there is an obvious accumulation or clustering in the transformation space, which means that under this particular combination of θ and ρ, the straight line or curve is well aligned with multiple points in the image data. This indicates that the corresponding plots or features in the image show obvious structure or morphology along a certain direction (defined by θ) and distance (defined by ρ).

[0169] Assumptions:

[0170] Suppose there are 3 categories, each of which is contributed by 2 features;

[0171] Assume that i represents a specific plot of land;

[0172] Assume that the feature weight w 1 =0.3, w 2 =0.5;

[0173] Assume H i11 =10,H i12 =20, H i21 =15,H i22 =25,H i31 =5,H i32 =30.

[0174] Calculate the score for the j=1th class:

[0175]

[0176] Calculate the score for class j=2:

[0177]

[0178] Calculate the score for class j=3:

[0179]

[0180] Select the maximum value:

[0181] C(i)=max(C(i1), C(i2), C(i3))=max(1,1,1)=1

[0182] A score of 1 indicates that the plot and the characteristics of the corresponding category are perfectly matched in the statistical model, indicating that all characteristics strongly support that the plot belongs to its respective category when evaluated by the logistic function.

[0183] See also Figure 6 , the identification steps of boundary error are:

[0184] Receive and process preliminary plot segmentation results, determine the boundaries of each plot, and generate preliminary plot boundary data results;

[0185] Based on the preliminary plot boundary data results, edge detection technology is applied.

[0186]

[0187] Among them, E(x,y) represents the edge strength, and They represent the gradient of the image brightness function f(x,y) in the x direction and the y direction respectively, ∈ is a small constant used to enhance the detection sensitivity of tiny edges, f(x,y) is the image brightness, and generates edge detection results;

[0188] Compare the edge detection results with the preset plot boundaries and calculate the deviation.

[0189]

[0190] Among them, D represents the average boundary error, detected_edge(x,y) represents the detected edge position, expected_edge(x,y) represents the expected edge position, and N represents the total number of boundary points to generate boundary error data.

[0191] Assume that the image brightness gradient at point (x, y) is:

[0192]

[0193] Set ∈ = 0.1, then,

[0194]

[0195] E(x,y)=36.194 represents the edge strength at point (x,y), with higher values ​​indicating strong edges.

[0196] Assuming there are 3 boundary points, the detected boundaries and expected boundaries are:

[0197] detected_edge = [100,150,120]

[0198] expected_edge = [102,148,123]

[0199] but,

[0200]

[0201] D = 2.33 represents the average boundary error, and a smaller value indicates a better matching accuracy.

[0202] See also Figure 7 , the steps to obtain the results of the land parcel boundary adjustment are:

[0203] Analyze boundary error data, apply computational geometry techniques, and use the error minimization formula.

[0204]

[0205] Among them, L opt represents the sum of squares of the total error, ω i Represents the weight adjusted according to the characteristics of the plot, which is used to improve the accuracy of the correction. i,actual and P i,expected Respectively represent the actual and expected boundary points, and generate the optimization results of the land parcel boundary;

[0206] Update the boundary data of the plot according to the optimization results of the plot boundary, match the boundary with the actual needs, and use the update formula.

[0207] P new =P actual +β·(P expected -P actual )

[0208] Among them, P actual represents the actual detected boundary point position, P expected It represents the expected boundary point location, which is determined based on planning and design standards. β represents the adjustment coefficient, which controls the adjustment range and has a value range between 0 and 1. It generates updated boundary data.

[0209] The updated boundary data results are integrated to form the plot boundary adjustment results, including the adjustment process and final results of each plot boundary.

[0210] Assume there are 3 boundary points, the actual and expected positions are:

[0211] P actual=[100,150,120]

[0212] P expected =[102,148,123]

[0213] Setting i is 1, then,

[0214]

[0215] L opt =17 represents the sum of squares of the total error, minimizing L opt Values ​​can help correct parcel boundaries more accurately.

[0216] Assume that there is an actual position P of a boundary point actual =100, P expected The expected position is 110. Assuming β is 0.5, it means adjusting to the average of the expected position and the current position. Then,

[0217] P new =100+0.5·(110-100)

[0218] =100+0.5·10

[0219] =100+5

[0220] =105

[0221] P new =105 means that after adjustment, the boundary point of the plot will move from the actual position 100 to 105.

[0222] See also Figure 8 , the steps to obtain the final segmentation result of the planting field are:

[0223] Integrate the plot boundary adjustment results with the plot morphology recognition data, match each plot data, and generate an integrated plot data set;

[0224] Using the integrated plot data set, calculate the usage index U(i) of each plot.

[0225]

[0226] Among them, a k represents the attribute value of plot i at feature k, including soil quality score, distance to water source or other plot characteristics, w k is the weight coefficient of the target feature, which is used to adjust the influence of each attribute in the total usage index and generate the usage index of each plot;

[0227] The matching degree between the boundary and crop planting conditions of each plot is analyzed using the matching degree analysis formula.

[0228]

[0229] Among them, M(i) represents the matching degree, d j is a binary value indicating whether the crop planting situation matches the expected situation, 1 indicates a match, 0 indicates a mismatch, and λ j is the impact factor or area weight of region j, which is used to improve the accuracy of the analysis and generate the matching analysis results;

[0230] According to the utilization index and matching analysis results of each plot, the final adjustment and optimization of the plot boundaries are carried out to ensure that the division of each plot matches the actual needs of crop planting and generate the final segmentation results of the planting fields.

[0231] Assume a plot has the following characteristics and weights:

[0232] Feature 1: Soil quality score = 80, weight = 0.5,

[0233] Feature 2: Distance to water source (closer distances give higher scores) = 90, weight = 0.5,

[0234] but,

[0235]

[0236] U(i)=85 represents the utilization score of the plot with all its characteristics combined, and a higher score indicates that the plot is in good utilization.

[0237] Assume that a plot has three regions, and the matching status and weight of each region are as follows:

[0238] Region 1: Match = 1, Area Weight = 0.2,

[0239] Region 2: Match = 0, Area Weight = 0.5,

[0240] Region 3: Match = 1, Area Weight = 0.3,

[0241] but,

[0242]

[0243] M(i)=0.5 means that the matching degree of the plot is 50%, indicating that the crop planting conditions match the plot boundaries in half of the area.

[0244] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A planting field automatic identification and segmentation system based on remote sensing images, characterized in that: The system comprises: The image texture feature extraction module collects remote sensing images, separates texture layers through wavelet transform, analyzes the directionality, roughness and contrast of textures in image data, integrates texture data, and generates texture feature mapping results; The morphology and segmentation intelligent processing module applies shape recognition logic to identify the differential land parcel morphology according to the texture feature mapping result, matches and classifies the land parcel morphology through Hough transform, merges the classification results, and generates land parcel morphology recognition data and preliminary land parcel segmentation results; The boundary refinement correction module performs edge detection based on the preliminary plot segmentation result, identifies boundary errors, corrects the plot boundary using computational geometry technology, adjusts boundary data, and generates a plot boundary adjustment result; The plot information integration module combines the plot boundary adjustment results with the plot morphology recognition data, analyzes the usage status of each plot, verifies the matching degree between the plot boundary and the crop planting situation, optimizes the division accuracy, and generates the final segmentation result of the planting field; The steps for obtaining the final segmentation result of the planting field are as follows: Integrate the plot boundary adjustment result with the plot morphology recognition data, match each plot data, and generate an integrated plot data set; Using the integrated plot data set, calculate the usage index of each plot , ; in, represents the attribute value of plot i at feature k, including soil quality score, distance to water source or other plot characteristics, is the weight coefficient of the target feature, which is used to adjust the influence of each attribute in the total usage index and generate the usage index of each plot; The matching degree between the boundary and crop planting conditions of each plot is analyzed using the matching degree analysis formula. ; in, Indicates the matching degree, It is a binary value indicating whether the crop planting situation matches the expected situation, 1 indicates a match and 0 indicates a mismatch. It is a region The impact factor or area weight is used to improve the accuracy of the analysis and generate matching analysis results; According to the utilization index and matching degree analysis results of each plot, the boundaries of the plots are finally adjusted and optimized to ensure that the division of each plot matches the actual needs of crop planting and generate the final segmentation results of the planting fields.

2. The automatic identification and segmentation system for planting fields based on remote sensing images according to claim 1 is characterized in that: The separation steps of the texture layer are: Collect remote sensing image data, obtain original images from satellites or aerial photography equipment, and generate original image data results; Applying wavelet transform to the original image data result, ; in, Represents the wavelet function used to extract the difference scale w and position The image features under Represent the scale and location parameters of wavelet transform, represents the pixel value of the image data, is the noise variance, which is used to improve the anti-noise performance of wavelet transform and generate wavelet transform processing results; Separate the texture layer from the wavelet transform processing result and apply the formula, ; in, is the data value of the wavelet transform result, indicating the possibility that each pixel belongs to the target texture layer, e is the natural logarithm base, and α is the adjustment parameter used to adjust the sensitivity of logistic regression to generate the texture layer separation result.

3. The automatic identification and segmentation system for planting fields based on remote sensing images according to claim 2 is characterized in that: The steps of obtaining the texture feature mapping result are: Extract texture features from the texture layer separation results, calculate the directionality, roughness and contrast of each separated texture layer, and use the formula: ; Determine the directionality, where L is the pixel value in the texture layer separation result, is the direction angle of the target pixel, For the direction deviation angle, increase the dimension of calculation and generate characteristic analysis results; The characteristic analysis results are integrated and the formula is applied. ; Comprehensive directionality, roughness and contrast effects, among which, are weight coefficients, which adjust the influence proportion of directionality, roughness and contrast in the total score respectively to generate the texture data integration result; According to the texture data integration result, a texture feature map is created, and a mapping formula is used to perform standardized feature expression. ; in, is the texture feature mapping value, S is the texture data integration result, is the maximum value in, To avoid small constants of division by zero errors, texture feature map results are generated.

4. The automatic identification and segmentation system for planting fields based on remote sensing images according to claim 1 is characterized in that: The steps for obtaining the land parcel morphology recognition data are as follows: Analyze the texture feature mapping result, calculate the shape parameters a, b, c of the plot, where a represents the side length ratio, b represents the angle deviation, and c represents the area ratio, and generate the morphological feature parameter result; Combined with the morphological characteristic parameter results, quantitative analysis is performed. ; in, To quantify the score of plot morphology, To adjust the parameters, used to adjust the influence of angle deviation and area ratio, and generate the morphological recognition analysis results; According to the morphological recognition analysis results, the differences of all plots are distinguished to generate plot morphological recognition data.

5. The automatic identification and segmentation system for planting fields based on remote sensing images according to claim 4 is characterized in that: The steps for obtaining the preliminary plot segmentation result are as follows: Combined with the land morphology recognition data, Hough transform is performed. ; in, is the angle parameter, ρ is the distance parameter, Used to increase sensitivity to small errors, Generate Hough transform matching results for the morphological feature values ​​of the land parcel morphological recognition data; Analyze the Hough transform matching results and use the classification formula: ; in, It represents the highest classification score after classifying the i-th plot, which is used to determine the final category of the plot. j represents the index of the classification. The weight parameter weights the impact of the difference feature. It represents the Hough transform result contributed by the kth feature of the i-th plot in the j-th category. The natural logarithm base of e is used for logical function calculation to generate the classification processing result. The classification results are combined to summarize similar plots and generate preliminary plot segmentation results.

6. The automatic identification and segmentation system for planting fields based on remote sensing images according to claim 1 is characterized in that: The identification steps of the boundary error are: Receive and process the preliminary plot segmentation result, determine the boundary of each plot, and generate preliminary plot boundary data result; Based on the preliminary plot boundary data results, edge detection technology is applied to ; in, represents the edge strength, Represents the image brightness function The gradient in the x and y directions, is a small constant used to enhance the detection sensitivity of tiny edges. is the image brightness, generating edge detection results; Compare the edge detection result with the preset plot boundary and calculate the deviation, ; Where D represents the average boundary error, represents the detected edge position, Represents the expected edge position, N represents the total number of boundary points, and generates boundary error data.

7. The automatic identification and segmentation system for planting fields based on remote sensing images according to claim 6 is characterized by: The steps for obtaining the plot boundary adjustment result are as follows: Analyze the boundary error data, apply computational geometry techniques, and use the error minimization formula. ; in, represents the sum of squares of the total errors, Represents the weight adjusted according to the characteristics of the plot, which is used to improve the accuracy of the correction. Respectively represent the actual and expected boundary points, and generate the optimization results of the land parcel boundary; Update the boundary data of the plot according to the plot boundary optimization result, match the boundary with the actual demand, and adopt the update formula. ; in, Indicates the actual detected boundary point position, It represents the expected boundary point location, which is determined based on planning and design standards. β represents the adjustment coefficient, which controls the adjustment range and has a value range between 0 and 1. It generates updated boundary data. The updated boundary data results are integrated to form a land parcel boundary adjustment result, including the adjustment process and final result of each land parcel boundary.

Images