A vegetation detection method and system for soil environmental remediation

Abstract

The present application relates to the technical field of soil image processing, and particularly relates to a vegetation detection method and system for soil environment remediation. The present application analyzes color features and position distribution features of pixel points, obtains a vegetation coverage factor of each superpixel block, and screens out vegetation coverage blocks; according to position distribution and chroma fluctuation features of pixel points in each vegetation coverage block, an artificial cultivation feature degree of each vegetation coverage block is obtained; according to a centroid position and the vegetation coverage factor of each superpixel block, a vegetation coverage cluster and a non-vegetation coverage cluster are obtained; in combination with relative positions between each non-vegetation coverage cluster and each vegetation coverage block in an adjacent vegetation coverage cluster, a soil remediation vegetation salience of each superpixel block is obtained, and then a soil remediation weighted salience of each superpixel block is obtained. The present application accurately analyzes vegetation features in a soil environment remediation process, and improves the accuracy of soil environment remediation evaluation.

Description

Technical Field

[0001] This invention relates to the field of soil image processing technology, and more specifically to a vegetation detection method and system for soil environmental remediation. Background Technology

[0002] Soil remediation is a key measure to improve the quality of contaminated soil, and vegetation restoration is one of the important indicators for evaluating the effectiveness of soil remediation. Vegetation plays a crucial role in the soil remediation process, not only serving as an indicator organism for pollution but also reducing pollutant levels through phytoremediation. Therefore, vegetation monitoring is an important means of assessing the effectiveness of soil remediation.

[0003] In existing technologies, the assessment of soil environmental remediation results relies on a single vegetation index or spectral characteristics. However, the assessment process is negatively affected by the inconsistent scale differences between different types of natural vegetation areas. The failure to consider the spatial distribution characteristics of vegetation cover areas leads to poor accuracy in local soil environmental remediation assessments. Summary of the Invention

[0004] To address the technical problem of poor accuracy in soil environmental remediation assessments due to a lack of consideration for the spatial distribution characteristics of vegetation cover areas, the present invention aims to provide a vegetation detection method and system for soil environmental remediation. The specific technical solution adopted is as follows:

[0005] This invention proposes a vegetation detection method for soil environmental remediation, the method comprising:

[0006] Acquire remote sensing images of the soil area;

[0007] Multiple superpixel blocks of a remote sensing image are obtained. Based on the color features and location distribution features of the pixels in each superpixel block, the vegetation cover factor of each superpixel block is obtained, and vegetation cover blocks are selected. Based on the location distribution of the pixels in each vegetation cover block and the color fluctuation features, the artificial cultivation feature degree of each vegetation cover block is obtained.

[0008] Based on the centroid location and vegetation cover factor of each superpixel block, all superpixel blocks are clustered to obtain block clusters; based on the vegetation cover factor distribution of all superpixel blocks in different block clusters, vegetation cover clusters and non-vegetation cover clusters are obtained.

[0009] According to the relative position distribution between each non-vegetation coverage class cluster and each vegetation coverage patch in the adjacent vegetation coverage class cluster, the artificial cultivation characteristic degree of the corresponding vegetation coverage patch, and the vegetation coverage factor, a soil remediation vegetation salience of each superpixel patch is obtained; according to the soil remediation vegetation salience of each superpixel patch, the salience distance between each superpixel patch and different other superpixel patches, and the vegetation coverage factor of the other superpixel patches, a soil remediation weighted salience of each superpixel patch is obtained;

[0010] According to the soil remediation weighted salience of all superpixel patches, vegetation is detected.

[0011] Further, the vegetation coverage factor obtaining method comprises:

[0012] For any superpixel patch, an over-green minus over-red index of each pixel point is obtained, and normalized mapping is performed, as a vegetation color feature;

[0013] An edge detection processing is performed on the superpixel patch by using a CANNY algorithm, and a plurality of edge lines are obtained; a slope variance of all edge pixel points on an edge line where each edge pixel point is located is obtained, as a slope fluctuation feature of each edge pixel point; an edge pixel point with a minimum relative distance from each pixel point is selected, and the slope fluctuation feature of the corresponding edge pixel point is taken as a local texture roughness coefficient of each pixel point;

[0014] According to the saturation channel value of different pixel points in the HSV space, the vegetation color feature, and the local texture roughness coefficient, a vegetation coverage factor of each superpixel patch is obtained, the saturation channel value is negatively correlated with the vegetation coverage factor, and the vegetation color feature and the local texture roughness coefficient are positively correlated with the vegetation coverage factor.

[0015] Further, the vegetation coverage patch obtaining method comprises:

[0016] If the vegetation coverage factor of the superpixel patch is greater than or equal to a preset coverage threshold, the corresponding superpixel patch is taken as a vegetation coverage patch.

[0017] Further, the artificial cultivation characteristic degree obtaining method comprises:

[0018] For each vegetation coverage patch, the number of edge pixel points on each edge line is counted, as the length of the corresponding edge line; the average difference of slopes between each boundary pixel point and an adjacent boundary pixel point is obtained, as a slope change rate of each boundary pixel point;

[0019] The length of the edge line or the slope change rate of the boundary pixel point is taken as analysis data, the difference between adjacent analysis data is calculated in descending order, the maximum value in the corresponding element when the difference is maximum is selected as a data reference value, if the analysis data is greater than the analysis reference value, the edge line corresponding to the analysis data is taken as a long edge line, or the boundary pixel point corresponding to the analysis data is taken as a boundary segmentation point, to obtain a plurality of block boundaries;

[0020] The principal component direction of each long edge line and block boundary is obtained by using the PCA algorithm, and the included angle between the principal component direction and the horizontal direction is obtained. According to the minimum value of the included angle between the principal component direction of each long edge line and different block boundaries in each vegetation cover block, and the fluctuation degree of the color channel value of all pixel points in the HSV space, the artificial cultivation degree of each vegetation cover block is obtained. The minimum value of the included angle and the fluctuation degree are positively correlated with the degree of artificial cultivation.

[0021] Further, the method for obtaining the block clustering cluster comprises:

[0022] The centroid coordinates of each superpixel block and the vegetation cover factor are used to form multi-dimensional data, the Euclidean distance of the multi-dimensional data between the superpixel blocks is obtained, and K-means clustering is performed on all superpixel blocks to obtain a block clustering cluster.

[0023] Further, the method for obtaining the vegetation cover cluster and the non-vegetation cover cluster comprises:

[0024] The mean value of the vegetation cover factor of all superpixel blocks in each block clustering cluster is obtained as the overall vegetation cover level.

[0025] The difference between adjacent overall vegetation cover levels is calculated in descending order, the maximum value in the corresponding element when the difference is maximum is selected as a coverage level reference value, if the overall vegetation cover level of the block clustering cluster is greater than the coverage level reference value, the corresponding block clustering cluster is taken as a vegetation cover cluster, and the remaining block clustering clusters are taken as non-vegetation cover clusters.

[0026] Further, the method for obtaining the soil remediation vegetation significance degree comprises:

[0027] According to the relative position distribution between each non-vegetation cover cluster and each vegetation cover block in the adjacent vegetation cover cluster, and the vegetation cover factor of the corresponding vegetation cover block, the soil environment remediation degree of each non-vegetation cover cluster is obtained.

[0028] The relative distance of the centroid in the corresponding block area between each vegetation cover block and each non-vegetation cover cluster is obtained, the minimum value of the relative distance between each vegetation cover block and different non-vegetation cover clusters is selected, and the corresponding non-vegetation cover cluster is taken as a reference cluster of each vegetation cover block.

[0029] Obtain the product of the artificial cultivation characteristic degree of each vegetation coverage sub-block and the soil environment remediation degree of the corresponding reference class cluster as the soil remediation significant coefficient; obtain the ratio of the soil remediation significant coefficient of each vegetation coverage sub-block and the minimum relative distance as the soil remediation vegetation significant degree of each vegetation coverage sub-block; if the superpixel sub-block is not a vegetation coverage sub-block, the soil remediation vegetation significant degree of the corresponding superpixel sub-block is set to a positive integer 1.

[0030] Further, the method for obtaining the soil environment remediation degree comprises:

[0031] Obtain the relative distance distribution of the center of mass in the corresponding sub-block area between each non-vegetation coverage class cluster and each vegetation coverage sub-block in the adjacent vegetation coverage class cluster as the remediation edge distance; and form a remediation edge distance sequence in ascending order according to the corresponding remediation edge distance between each non-vegetation coverage class cluster and all vegetation coverage sub-blocks in the adjacent vegetation coverage class cluster;

[0032] According to the distribution sequence of the corresponding vegetation coverage sub-block of each element in the remediation edge distance sequence, a vegetation coverage factor sequence of all vegetation coverage sub-blocks is constructed;

[0033] Obtain the correlation coefficient between the remediation edge distance sequence and the vegetation coverage factor sequence, and normalize it as the soil environment remediation degree of each non-vegetation coverage area.

[0034] Further, the method for obtaining the soil remediation weighted significant degree comprises:

[0035] Obtain the significant distance between each superpixel sub-block and different other superpixel sub-blocks, calculate the product accumulation value of the significant distance and the vegetation coverage factor of the corresponding other superpixel sub-blocks, obtain the product between the product accumulation value and the soil remediation vegetation significant value of each superpixel sub-block, and perform positive correlation normalization as the soil remediation weighted significant degree of each superpixel sub-block.

[0036] The application further provides a vegetation detection system for soil environment remediation, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of any one of the vegetation detection methods for soil environment remediation when executing the computer program.

[0037] The application has the following beneficial effects:

[0038] The application obtains a plurality of superpixel blocks of a remote sensing image, obtains a vegetation cover factor of each superpixel block according to color features and position distribution features of pixel points in each superpixel block, and screens out vegetation cover blocks, so that a vegetation related region can be preliminarily screened out, and interference of non-vegetation regions such as ground shadows, bare soil and buildings is reduced; an artificial cultivation feature degree of each vegetation cover block is obtained according to position distribution and chroma fluctuation features of pixel points in each vegetation cover block, so that contribution of artificial vegetation in a soil remediation region can be more accurately evaluated; all superpixel blocks are clustered according to a centroid position and the vegetation cover factor of each superpixel block, so that a block clustering cluster is obtained, and through clustering analysis, superpixel blocks with similar centroid positions and vegetation cover factors are divided into the same cluster, so that spatial distribution features of vegetation can be better reflected; a vegetation cover cluster and a non-vegetation cover cluster are obtained according to vegetation cover factor distribution of all superpixel blocks in different block clustering clusters; a soil remediation vegetation salience of each superpixel block is obtained according to relative position distribution between each non-vegetation cover cluster and each vegetation cover block in an adjacent vegetation cover cluster, artificial cultivation feature degrees of the vegetation cover blocks and the vegetation cover factor; through quantification of the salience, a vegetation region with important significance for soil remediation can be identified; a soil remediation weighted salience of each superpixel block is obtained according to the soil remediation vegetation salience of each superpixel block, a salience distance between each superpixel block and different other superpixel blocks and the vegetation cover factor of the other superpixel blocks, which is helpful to accurately reflect overall effects of soil remediation. Through accurate analysis of vegetation features in a soil environment remediation process, the application improves accuracy of soil environment remediation evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, and the advantages thereof, below, a brief introduction will be given to the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative effort based on these drawings.

[0040] Figure 1 A flow chart of a vegetation detection method for soil environment remediation provided by an embodiment of the present application;

[0041] Figure 2 A flow chart of a vegetation cover factor acquisition method provided by an embodiment of the present application;

[0042] Figure 3 A flow chart of an artificial cultivation feature degree acquisition method provided by an embodiment of the present application;

[0043] Figure 4A flow chart of a method for obtaining soil remediation vegetation coverage is provided in an embodiment of the application. DETAILED DESCRIPTION

[0044] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined object, the specific embodiments, structure, features and effects of a vegetation detection method and system for soil environment remediation according to the present application are described in detail below in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0045] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0046] The specific scheme of a vegetation detection method and system for soil environment remediation provided by the present application is described below in combination with the drawings.

[0047] Please refer to Figure 1 which shows a method flow chart of a vegetation detection method for soil environment remediation provided by an embodiment of the present application, specifically including:

[0048] Step S1: Obtain a remote sensing image of a soil area.

[0049] In an embodiment of the present application, the remote sensing image can quickly cover a large range of soil area, provide a global perspective, facilitate the identification of the distribution of vegetation coverage and non-vegetation coverage area, and help to evaluate the growth state of vegetation in the soil environment remediation area; first, the remote sensing image of the soil area environment is collected by using aerial photography equipment when the weather is fine.

[0050] It should be noted that in order to facilitate subsequent image processing, the image is filtered and denoised by non-local mean filtering, which helps to improve the quality of the image, highlight the outline and details of the image, and make the image clearer; the non-local mean filtering algorithm is a well-known technical means to those skilled in the art, which will not be described here.

[0051] Step S2: Obtain a plurality of superpixel blocks of the remote sensing image, obtain a vegetation coverage factor of each superpixel block according to the color features and position distribution features of the pixel points in each superpixel block, and screen out vegetation coverage blocks; obtain an artificial cultivation feature degree of each vegetation coverage block according to the position distribution and chroma fluctuation features of the pixel points in each vegetation coverage block.

[0052] The remote sensing image contains a large number of pixel points, and direct processing of the pixel points is not only large in calculation amount, but also easy to introduce noise; in order to reduce the calculation complexity, the image is segmented into a plurality of super pixel blocks, which is helpful to simplify the image analysis and understanding process; a plurality of super pixel blocks of the remote sensing image are obtained.

[0053] It should be noted that, in an embodiment of the present application, the remote sensing image is processed by a super pixel segmentation algorithm to obtain a plurality of super pixel blocks of the remote sensing image; wherein the super pixel segmentation can aggregate the pixel points in the image into a plurality of larger pixel blocks according to the similarity criteria of color, texture or shape, i.e. super pixel blocks, which can greatly reduce the dimension of image processing while retaining important information of the image. The specific super pixel segmentation algorithm is a known technical means to those skilled in the art, which will not be described here.

[0054] The soil itself is darker in color than the vegetation, and has higher saturation, and artificial objects often have regular geometric shapes, and their texture distribution is more uniform than that of the vegetation area; by analyzing the color features and position distribution features, the vegetation in the remote sensing image usually has specific spectral characteristics, and the color features can effectively distinguish the vegetation from the non-vegetation area; the position distribution features are helpful to understand the possibility of vegetation growth and the texture distribution condition; according to the color features and position distribution features of the pixel points in each super pixel block, the vegetation coverage factor of each super pixel block is obtained.

[0055] Preferably, in an embodiment of the present application, the method for obtaining the vegetation coverage factor is as follows: Figure 2 which shows a flow chart of a method for obtaining a vegetation coverage factor, comprising:

[0056] Step S201: for any super pixel block, obtain the over-green minus over-red index of each pixel point, and perform normalized mapping as the vegetation color feature.

[0057] The over-green minus over-red index is helpful to further highlight the green feature of the vegetation, while suppressing the interference of non-vegetation areas such as soil and bare land, and can effectively enhance the contrast between the vegetation and the non-vegetation area.

[0058] It should be noted that the specific over-green minus over-red index is a known technical means to those skilled in the art, which will not be described here.

[0059] Step S202: the CANNY algorithm is used to perform edge detection processing on the super pixel block to obtain a plurality of edge lines; the slope variance of all edge pixel points on the edge line where each edge pixel point is located is obtained as the slope fluctuation feature of each edge pixel point; the edge pixel point with the minimum relative distance from each pixel point is selected, and the slope fluctuation feature of the corresponding edge pixel point is taken as the local texture roughness coefficient of each pixel point.

[0060] The CANNY algorithm is a classical edge detection algorithm, which can effectively extract edge information in an image. By performing edge detection on the superpixel blocks, clear edge lines can be obtained. The slope change of the pixel points on the edge lines reflects the smoothness or roughness of the edges. By assigning the slope fluctuation characteristics of the nearest edge pixel points to each pixel point, the roughness information of the edges can be extended to the entire region, thereby describing the roughness of the local texture.

[0061] It should be noted that in some embodiments of the present application, the relative distance can be obtained by using existing distance calculation methods such as Euclidean distance or Manhattan distance. The specific means are well known to those skilled in the art, and will not be described here.

[0062] It should be noted that in an embodiment of the present application, the slope is obtained by obtaining the ratio of the difference in the y-coordinate and the difference in the x-coordinate between each edge pixel point and the previous adjacent edge pixel point, as the slope of each edge pixel point. The specific CANNY algorithm is well known to those skilled in the art, and will not be described here.

[0063] Step S203: Obtain a vegetation coverage factor of each superpixel block according to the saturation channel value of different pixel points in the HSV space, the vegetation color feature, and the local texture roughness coefficient. The saturation channel value is negatively correlated with the vegetation coverage factor, and the vegetation color feature and the local texture roughness coefficient are positively correlated with the vegetation coverage factor.

[0064] In an embodiment of the present application, for each superpixel block, the formula of the vegetation coverage factor is:

[0065]

[0066] wherein C represents the vegetation coverage factor of each superpixel block; represents the local texture roughness coefficient of the i-th pixel point in each superpixel block; S i represents the saturation channel value of the i-th pixel point in the HSV space; n represents the number of pixel points in the superpixel block; E i represents the vegetation color feature of the i-th pixel point; norm() represents a normalization function.

[0067] In the formula of the vegetation coverage factor, the smaller the local texture roughness coefficient of the i-th pixel point in each superpixel block, the more uniform the texture distribution, the greater the saturation channel value of the i-th pixel point in the HSV space, the greater the possibility of being soil, the more the existence of object interference, the more the vegetation color feature is disturbed by other objects, the relatively lower the vegetation color feature, and the smaller the vegetation coverage factor.

[0068] The vegetation coverage factor can reflect the vegetation coverage degree in a certain region, and the greater the vegetation coverage factor is, the more likely the vegetation coverage block is; the vegetation coverage block is screened out through analysis of the vegetation coverage factor.

[0069] Preferably, in an embodiment of the present application, the method for obtaining the vegetation coverage block comprises:

[0070] If the vegetation coverage factor of the superpixel block is greater than or equal to the preset coverage threshold, the corresponding superpixel block is taken as the vegetation coverage block.

[0071] It should be noted that, in an embodiment of the present application, the preset coverage threshold is set to 0.65; in other embodiments of the present application, the preset coverage threshold can be set according to specific circumstances, which is not limited or described here.

[0072] In the process of soil environmental remediation, artificial planting vegetation is a common method. Compared with the naturally growing vegetation area, the plant species of the artificial vegetation area is relatively single, the color is relatively consistent, the distribution is more uniform, and the plants often present in neat rows; therefore, the artificial cultivation feature degree of each vegetation coverage block is obtained according to the position distribution and the chroma fluctuation characteristics of the pixel points in each vegetation coverage block.

[0073] Preferably, in an embodiment of the present application, the method for obtaining the artificial cultivation feature degree is described in Figure 3 which shows a flowchart of a method for obtaining the artificial cultivation feature degree, comprising:

[0074] Step S301: For each vegetation coverage block, the number of edge pixel points on each edge line is counted as the length of the corresponding edge line; the average difference of the slopes between each boundary pixel point and the adjacent boundary pixel point is obtained as the slope change rate of each boundary pixel point.

[0075] The length of the edge line reflects the continuity and saliency of the edge. In the vegetation coverage block, the long edge line can correspond to the boundary or internal structure of the vegetation area, which is used to quantify the saliency of the edge line; the boundary refers to the edge line intersecting between the blocks, and the slope change rate of the boundary pixel point reflects the smoothness or tortuosity of the boundary.

[0076] Step S302: The length of the edge line or the slope change rate of the boundary pixel point is taken as the analysis data, the difference between the adjacent analysis data is calculated in the order from large to small, the maximum value in the corresponding element when the difference is maximum is selected as the data reference value, if the analysis data is greater than the analysis reference value, the edge line corresponding to the analysis data is taken as the long edge line, or the boundary pixel point corresponding to the analysis data is taken as the boundary segmentation point, and a plurality of block boundaries are obtained.

[0077] By calculating the difference between adjacent analysis data, the mutation points of data distribution can be found, which usually correspond to significant feature changes; selecting the maximum value when the difference is maximum as the reference value can ensure that the selected edge lines or boundary points have significance.

[0078] Step S303: Obtain the principal component direction of each long edge line and the block boundary by using the PCA algorithm, and obtain the included angle between the principal component direction and the horizontal direction; according to the minimum difference of the included angle between the principal component direction of each long edge line and different block boundaries in each vegetation cover block and the fluctuation degree of the chroma channel value of all pixel points in the HSV space, obtain the artificial cultivation feature degree of each vegetation cover block, and the minimum difference of the included angle and the fluctuation degree are positively correlated with the artificial cultivation degree.

[0079] PCA can extract the main change direction of data, which is used to quantify the directional features of edge lines or boundaries; in the vegetation cover block, the principal component direction can reflect the overall direction of the vegetation arrangement; the included angle reflects the deviation degree of the direction of the edge line or the boundary from the horizontal direction, and the fluctuation degree of the chroma channel value reflects the uniformity of the color of the vegetation, and the more consistent the color distribution, the greater the artificial cultivation degree.

[0080] It should be noted that in an embodiment of the present application, the fluctuation degree can be represented by calculating the variance, the greater the variance, the greater the fluctuation degree, and the smaller the variance, the smaller the fluctuation degree; in other embodiments of the present application, the fluctuation degree can also be represented by calculating the range or standard deviation, and the specific means are well known to those skilled in the art, which will not be described here.

[0081] In an embodiment of the present application, for each vegetation cover block, the formula of the artificial cultivation feature degree is:

[0082]

[0083] Wherein, P represents the artificial cultivation feature degree of each vegetation cover block; Var (H) represents the variance of the chroma channel value of all pixel points in the HSV space in each vegetation cover block, that is, the fluctuation degree; θ j represents the included angle between the principal component direction of the jth long edge line and the horizontal direction; α represents the included angle between the principal component direction of the block boundary and the horizontal direction; min | θ j represents the minimum difference of the included angle between the principal component direction of the jth long edge line and different block boundaries; m represents the number of long edge lines.

[0084] In the formula of the artificial cultivation feature degree, 0.01 is added in the formula to avoid that the denominator of the formula is 0, and the formula is meaningless; The greater the chroma fluctuation feature is, the more uneven the distribution is, and the less likely the artificial cultivation is; the smaller the minimum difference between the principal component directions corresponding to the angles of the jth long edge line and different patch boundaries is, the more consistent the principal component directions are, the more likely the cultivation is in the same direction, and the more likely the artificial vegetation is.

[0085] Step S3: clustering all the superpixel patches according to the centroid positions and the vegetation cover factors of each superpixel patch to obtain patch clustering clusters; and obtaining vegetation cover clusters and non-vegetation cover clusters according to the vegetation cover factor distributions of all the superpixel patches in different patch clustering clusters.

[0086] The centroid position is the geometric center of each superpixel patch, and the vegetation cover factor indicates the vegetation density or coverage ratio in the superpixel patch. Combining the two types of data forms multi-dimensional data, which is helpful to simultaneously consider the spatial position and the vegetation coverage. Through clustering, regions with similar vegetation coverage patterns are identified. All the superpixel patches are clustered according to the centroid positions and the vegetation cover factors of each superpixel patch to obtain patch clustering clusters.

[0087] Preferably, in an embodiment of the present application, the method for obtaining the patch clustering clusters comprises:

[0088] Multi-dimensional data is formed by the centroid coordinates and the vegetation cover factors of each superpixel patch, the Euclidean distance of the multi-dimensional data between the superpixel patches is obtained, all the superpixel patches are K-means clustered to obtain patch clustering clusters.

[0089] It should be noted that the K-means clustering can divide the superpixel patches into K clusters, so that the points in the same cluster have high similarity, and the points in different clusters have large differences, which is helpful to identify regions with similar vegetation coverage and spatial distribution patterns. The value of K is determined by the elbow rule, and specific K-means clustering is a technical means familiar to those skilled in the art, which will not be described here.

[0090] It should be noted that in other embodiments of the present application, Manhattan distance between multi-dimensional data and other existing distance calculation methods can also be calculated. The specific Euclidean distance and Manhattan distance are technical means familiar to those skilled in the art, which will not be described here.

[0091] The vegetation cover factor reflects the vegetation coverage degree in the superpixel patch and quantifies the vegetation coverage of the superpixel patch. The higher the vegetation cover factor is, the higher the vegetation coverage degree in the patch is. The vegetation cover clusters and the non-vegetation cover clusters are obtained according to the vegetation cover factor distributions of all the superpixel patches in different patch clustering clusters.

[0092] Preferably, in an embodiment of the present application, the method for obtaining the vegetation cover clusters and the non-vegetation cover clusters comprises:

[0093] Obtaining the mean of the vegetation coverage factors of all superpixel patches in each sub-block clustering cluster as the overall vegetation coverage level;

[0094] Calculating the difference of the adjacent overall vegetation coverage levels in descending order, selecting the maximum value in the corresponding element when the difference is maximum as the coverage level reference value; if the overall vegetation coverage level of the sub-block clustering cluster is greater than the coverage level reference value, the corresponding sub-block clustering cluster is taken as the vegetation coverage cluster, and the remaining sub-block clustering clusters are taken as the non-vegetation coverage cluster.

[0095] Step S4: According to the relative position distribution between each non-vegetation coverage cluster and each vegetation coverage patch in the adjacent vegetation coverage cluster, the artificial cultivation characteristic degree and the vegetation coverage factor of the corresponding vegetation coverage patch, obtaining the soil repair vegetation significance of each superpixel patch; according to the soil repair vegetation significance of each superpixel patch, the significant distance between each superpixel patch and different other superpixel patches, and the vegetation coverage factor of the other superpixel patches, obtaining the soil repair weighted significance of each superpixel patch.

[0096] The relative position distribution is the spatial relationship between the non-vegetation coverage cluster and the vegetation coverage patch in the adjacent vegetation coverage cluster, reflecting the spatial influence of the vegetation coverage patch on the non-vegetation coverage cluster, and the closer the vegetation coverage patch, the greater the contribution to the soil repair of the non-vegetation coverage cluster; the artificial cultivation characteristic degree and the vegetation coverage factor both reflect the possibility of planting plants by human beings; according to the relative position distribution between each non-vegetation coverage cluster and each vegetation coverage patch in the adjacent vegetation coverage cluster, the artificial cultivation characteristic degree and the vegetation coverage factor of the corresponding vegetation coverage patch, obtaining the soil repair vegetation significance of each superpixel patch.

[0097] Preferably, in an embodiment of the present application, the method for obtaining the soil repair vegetation significance is as follows: Figure 4 The method for obtaining the soil vegetation significance is shown in a flow chart, which comprises:

[0098] Step S401: According to the relative position distribution between each non-vegetation coverage cluster and each vegetation coverage patch in the adjacent vegetation coverage cluster, and the vegetation coverage factor of the corresponding vegetation coverage patch, obtaining the soil environment repair degree of each non-vegetation coverage cluster.

[0099] Preferably, in an embodiment of the present application, the method for obtaining the soil environment repair degree comprises:

[0100] obtaining a relative distance distribution of the centroid in the corresponding patch area between each non-vegetation coverage cluster and each vegetation coverage patch in the adjacent vegetation coverage cluster as a repair edge distance; and constructing a repair edge distance sequence in ascending order according to the repair edge distance between each non-vegetation coverage cluster and all vegetation coverage patches in the adjacent vegetation coverage cluster;

[0101] constructing a vegetation coverage factor sequence of all vegetation coverage patches according to the distribution sequence of the vegetation coverage patch corresponding to each element in the repair edge distance sequence;

[0102] obtaining a correlation coefficient between the repair edge distance sequence and the vegetation coverage factor sequence, and normalizing the correlation coefficient as a soil environment repair degree of each non-vegetation coverage area.

[0103] It should be noted that in an embodiment of the present application, the correlation coefficient between the sequences can be obtained by calculating the Pearson correlation coefficient. The greater the correlation coefficient, the closer the sequences, and the better the progress of soil environment repair can be reflected. The soil environment repair degree of the non-vegetation coverage area is obtained by normalizing the correlation coefficient using the sigmoid algorithm. In other embodiments of the present application, the correlation coefficient can also be obtained by using the cosine similarity, and the correlation coefficient is standardized to achieve normalization. The specific means is well known to those skilled in the art, and will not be described here.

[0104] Step S402: obtaining the relative distance of the centroid in the corresponding patch area between each vegetation coverage patch and each non-vegetation coverage cluster, selecting the minimum value of the relative distance between each vegetation coverage patch and different non-vegetation coverage clusters, and taking the corresponding non-vegetation coverage cluster as the reference cluster of each vegetation coverage patch.

[0105] The relative distance between the centroids quantifies the spatial relationship between the vegetation coverage patch and the non-vegetation coverage patch. The smaller the relative distance, the closer the patches, and the better the soil repair characteristics of the vegetation coverage patch can be represented.

[0106] Step S403: obtaining the product of the artificial cultivation feature degree of each vegetation coverage patch and the soil environment repair degree of the corresponding reference cluster as the soil repair significant coefficient; obtaining the ratio of the soil repair significant coefficient of each vegetation coverage patch and the minimum value of the relative distance as the soil repair vegetation significant degree of each vegetation coverage patch; and setting the soil repair vegetation significant degree of the corresponding superpixel patch to a positive integer 1 if the superpixel patch is not a vegetation coverage patch.

[0107] The artificial cultivation feature degree reflects an index of the degree of artificial cultivation in the vegetation coverage patch. The greater the artificial cultivation feature degree, the more likely the coverage patch is a human-planted area, and the greater the soil vegetation significant degree.

[0108] In an embodiment of the present application, for each vegetation coverage sub-block, the formula of the soil remediation vegetation prominence is:

[0109]

[0110] wherein X represents the soil remediation vegetation prominence of each vegetation coverage sub-block; P represents the artificial cultivation characteristic degree of each vegetation coverage sub-block; min(d) represents the minimum value of the relative distance of the center of each vegetation coverage sub-block and the corresponding sub-block area of different non-vegetation coverage clusters; and F represents the soil environment remediation degree of the corresponding non-vegetation coverage cluster when the minimum value of the relative distance between each vegetation coverage sub-block and different non-vegetation coverage clusters.

[0111] In the formula of the soil remediation vegetation prominence, the smaller the minimum value of the relative distance, the closer the vegetation coverage sub-block to the center position of the non-vegetation coverage area, the closer distance possibly having similar soil remediation characteristics, and the vegetation growth condition better reflecting the soil remediation progress of the vegetation coverage sub-block. Therefore, the greater the soil environment remediation degree and the artificial cultivation characteristic degree, the greater the soil remediation characteristics.

[0112] According to the soil remediation vegetation prominence of each superpixel sub-block, the significant distance between each superpixel sub-block and different other superpixel sub-blocks, and the vegetation coverage factor of the other superpixel sub-blocks, the soil remediation weighted prominence of each superpixel sub-block is obtained.

[0113] Preferably, in an embodiment of the present application, the method for obtaining the soil remediation weighted prominence comprises:

[0114] The significant distance between each superpixel sub-block and different other superpixel sub-blocks is obtained, the product accumulation value of the significant distance and the vegetation coverage factor of the corresponding other superpixel sub-blocks is calculated, the product between the product accumulation value and the soil remediation vegetation prominence of each superpixel sub-block is obtained, and the positive correlation normalization is performed as the soil remediation weighted prominence of each superpixel sub-block.

[0115] It should be noted that, in the embodiments of the present application, the method for obtaining the significant distance comprises: obtaining the saliency information of the superpixel sub-blocks by the saliency detection method, and weighting the Euclidean distance of the center coordinates between the superpixel sub-blocks by the difference of the saliency information as the significant distance; and the specific means is the technical means well known to those skilled in the art, which is not described here.

[0116] In an embodiment of the present application, for each superpixel sub-block, the formula of the soil remediation weighted prominence is represented as:

[0117]

[0118] Wherein, S represents the soil remediation weighted saliency of each superpixel patch; X represents the soil remediation vegetation saliency of each superpixel patch; C k represents the vegetation coverage factor of the kth other superpixel patch; d k represents the significant distance between each superpixel patch and the kth other superpixel patch; K represents the number of other superpixel patches; exp() represents the exponential function with the natural constant as the base.

[0119] In the formula of the soil remediation weighted saliency, the significant distance between each superpixel patch and the kth other superpixel patch is mapped in a negative correlation manner by the exponential function with the natural constant as the base. represents the product accumulation value of the significant distance between each superpixel patch and different other superpixel patches and the vegetation coverage factor of the corresponding other superpixel patch, the greater the product accumulation value, the greater the significant distance between the patches, the greater the difference in performance characteristics between the patches, the greater the vegetation coverage factor of the other superpixel patch, the greater the credibility of the significant distance between the patches, and the greater the saliency of the corresponding superpixel patch reflecting soil remediation.

[0120] Step S5: detecting vegetation according to the soil remediation weighted saliency of each superpixel patch.

[0121] It should be noted that in another embodiment of the present application, after obtaining the soil remediation weighted saliency of all superpixel patches, the vegetation is detected, including: based on the existing CA saliency detection algorithm, analyzing each pixel point in all superpixel patches at different preset scales to obtain the soil remediation saliency update degree of each pixel point at each preset scale; obtaining the average value of the soil remediation saliency update degree of each pixel point at all preset scales as the overall vegetation saliency of each pixel point; the preset scale set is {100%, 80%, 50%, 30%}, in other embodiments of the present application, the preset scale can be set according to specific circumstances, and the specific CA saliency detection algorithm is a technical means familiar to those skilled in the art, which will not be described here.

[0122] Taking the overall vegetation saliency of each pixel point as the analysis data, obtaining the pixel point corresponding to the analysis data greater than the analysis reference value as the vegetation pixel point of interest; performing morphological closing operation processing on all vegetation pixel points of interest to obtain the vegetation region of interest; obtaining the real-time vegetation condition index and vegetation health index of the vegetation region of interest through the normalized vegetation index, obtaining the product of the vegetation condition index and the vegetation health index as the real-time soil environment remediation completion degree of the vegetation region of interest. The specific morphological closing operation processing and the normalized vegetation index are technical means familiar to those skilled in the art, which will not be described here.

[0123] ​Based on this, by processing the accurate soil remediation weighted significance of the vegetation area, identifying the vegetation area of interest, and analyzing the corresponding vegetation health condition, it is helpful to more accurately evaluate the soil environment remediation condition, and the higher the soil environment remediation completion degree is, the better the soil environment remediation effect is.

[0124] To sum up, the present application obtains the vegetation coverage factor of each superpixel block by analyzing the color characteristics and position distribution characteristics of the pixel points, and screens out the vegetation coverage block;According to the position distribution and the chroma fluctuation characteristics of the pixel points in each vegetation coverage block, the artificial cultivation characteristics degree of each vegetation coverage block is obtained;According to the centroid position and the vegetation coverage factor of each superpixel block, the vegetation coverage cluster and the non-vegetation coverage cluster are obtained;The relative position distribution between each non-vegetation coverage cluster and each vegetation coverage block in the adjacent vegetation coverage cluster is combined to obtain the soil remediation vegetation significance of each superpixel block, and then the soil remediation weighted significance of each superpixel block is obtained.The present application accurately analyzes the vegetation characteristics in the soil environment remediation process, and improves the accuracy of soil environment remediation evaluation.

[0125] The present application also provides a vegetation detection system for soil environment remediation, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the steps of any one of the vegetation detection methods for soil environment remediation are realized.

[0126] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0127] Each embodiment in the specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other, and each embodiment mainly describes the difference from other embodiments.

Claims

1. A vegetation detection method for soil environmental remediation, characterized in that, The method includes: Acquire remote sensing images of the soil area; Multiple superpixel blocks of a remote sensing image are obtained. Based on the color features and location distribution features of the pixels in each superpixel block, the vegetation cover factor of each superpixel block is obtained, and vegetation cover blocks are selected. Based on the location distribution of the pixels in each vegetation cover block and the color fluctuation features, the artificial cultivation feature degree of each vegetation cover block is obtained. Based on the centroid location and vegetation cover factor of each superpixel block, all superpixel blocks are clustered to obtain block clusters; based on the vegetation cover factor distribution of all superpixel blocks in different block clusters, vegetation cover clusters and non-vegetation cover clusters are obtained. Based on the relative positional distribution between each non-vegetation cover cluster and each vegetation cover block within adjacent vegetation cover clusters, the artificial cultivation characteristic degree and vegetation cover factor of the corresponding vegetation cover blocks are used to obtain the soil remediation vegetation saliency of each superpixel block; based on the soil remediation vegetation saliency of each superpixel block, the significant distance between each superpixel block and other different superpixel blocks, and the vegetation cover factor of other superpixel blocks, the soil remediation weighted saliency of each superpixel block is obtained. Vegetation was detected based on the soil remediation weighted significance of all superpixel blocks; The method for obtaining the significance of soil remediation vegetation includes: Based on the relative positional distribution between each non-vegetation cover cluster and each vegetation cover block within an adjacent vegetation cover cluster, and the vegetation cover factor of the corresponding vegetation cover block, the soil environmental restoration degree of each non-vegetation cover cluster is obtained. Obtain the relative distance between the centroids within the corresponding block area of ​​each vegetation cover block and each non-vegetation cover cluster, select the minimum value of the relative distance between each vegetation cover block and different non-vegetation cover clusters, and use the corresponding non-vegetation cover cluster as the reference cluster of each vegetation cover block. The product of the artificial cultivation characteristic degree of each vegetation cover block and the soil environmental restoration degree of the corresponding reference cluster is obtained as the soil restoration significance coefficient; the ratio of the soil restoration significance coefficient of each vegetation cover block to the minimum relative distance is obtained as the soil restoration vegetation significance of each vegetation cover block; if the superpixel block is not a vegetation cover block, the soil restoration vegetation significance of the corresponding superpixel block is set to a positive integer 1.

2. The vegetation detection method for soil environmental remediation according to claim 1, characterized in that, The method for obtaining the vegetation cover factor includes: For any superpixel block, obtain the over-green minus over-red index of each pixel, and perform normalization mapping to serve as vegetation color features; The CANNY algorithm is used to perform edge detection on superpixel blocks to obtain multiple edge lines; the slope variance of all edge pixels on the edge line where each edge pixel is located is obtained as the slope fluctuation feature of each edge pixel; the edge pixel with the smallest relative distance to each pixel is selected, and the slope fluctuation feature of the corresponding edge pixel is used as the local texture coarsening coefficient of each pixel. The vegetation cover factor of each superpixel block is obtained based on the saturation channel value, vegetation color features, and local texture roughness coefficient of different pixels in HSV space. The saturation channel value is negatively correlated with the vegetation cover factor, while the vegetation color features and local texture roughness coefficient are positively correlated with the vegetation cover factor.

3. The vegetation detection method for soil environmental remediation according to claim 1, characterized in that, The method for obtaining the vegetation cover blocks includes: If the vegetation cover factor of a superpixel block is greater than or equal to the preset cover threshold, the corresponding superpixel block will be used as a vegetation cover block.

4. The vegetation detection method for soil environmental remediation according to claim 2, characterized in that, The method for obtaining the artificial cultivation characteristics includes: For each vegetation cover block, count the number of edge pixels on each edge line to get the length of the corresponding edge line; obtain the average difference in slope between each boundary pixel and its adjacent boundary pixels to get the slope change rate of each boundary pixel. The length of the edge line or the slope change rate of the boundary pixels are used as the analysis data. The difference between adjacent analysis data is calculated in descending order. The maximum value of the corresponding element with the largest difference is selected as the data reference value. If the analysis data is greater than the analysis reference value, the edge line corresponding to the analysis data is taken as the long edge line, or the boundary pixel corresponding to the analysis data is taken as the boundary segmentation point to obtain multiple block boundaries. The PCA algorithm is used to obtain the principal component directions of each long edge line and block boundary, and the angle between the principal component directions and the horizontal direction is obtained. Based on the minimum difference of the angles between the principal component directions of each long edge line and different block boundaries in each vegetation cover block, and the fluctuation of the chroma channel values ​​of all pixels in the HSV space, the artificial cultivation feature degree of each vegetation cover block is obtained. The minimum difference of the angles and the fluctuation degree are both positively correlated with the degree of artificial cultivation.

5. The vegetation detection method for soil environmental remediation according to claim 1, characterized in that, The method for obtaining the block clusters includes: Multidimensional data is constructed using the centroid coordinates and vegetation cover factor of each superpixel block. The Euclidean distance between the multidimensional data of the superpixel blocks is obtained. K-means clustering is performed on all superpixel blocks to obtain the block clusters.

6. The vegetation detection method for soil environmental remediation according to claim 1, characterized in that, The methods for obtaining the vegetation cover clusters and non-vegetation cover clusters include: The average vegetation cover factor of all superpixel blocks within each block cluster is obtained as the overall vegetation cover level. The differences in vegetation cover levels between adjacent clusters are calculated in descending order. The maximum value of the corresponding element with the largest difference is selected as the reference value for the cover level. If the overall vegetation cover level of a block cluster is greater than the reference value for the cover level, the corresponding block cluster is regarded as a vegetation cover cluster, and the remaining block clusters are regarded as non-vegetation cover clusters.

7. The vegetation detection method for soil environmental remediation according to claim 1, characterized in that, The method for obtaining the degree of soil environmental remediation includes: Obtain the relative distance distribution of the centroids within the corresponding block regions between each non-vegetation cluster and each vegetation block within an adjacent vegetation cluster, and use it as the restoration edge distance; construct the restoration edge distance sequence by arranging the corresponding restoration edge distances between each non-vegetation cluster and all vegetation blocks within an adjacent vegetation cluster in ascending order. Based on the distribution order of vegetation cover blocks corresponding to each element on the restoration edge distance sequence, a vegetation cover factor sequence corresponding to all vegetation cover blocks is constructed. The correlation coefficients between the remediation edge distance sequence and the vegetation cover factor sequence were obtained and normalized to represent the soil environmental remediation degree of each non-vegetation-covered area.

8. The vegetation detection method for soil environmental remediation according to claim 1, characterized in that, The method for obtaining the weighted significance of soil remediation includes: Obtain the salient distance between each superpixel block and other superpixel blocks. Calculate the cumulative sum of the salient distance and the vegetation cover factor of the corresponding other superpixel blocks. Obtain the product between the cumulative sum and the soil remediation vegetation salient value of each superpixel block, and perform positive correlation normalization to obtain the soil remediation weighted salient value of each superpixel block.

9. A vegetation detection system for soil environmental remediation, the system 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, it implements the steps of the vegetation detection method for soil environmental remediation as described in any one of claims 1 to 8.

Images