Land cover remote sensing information acquisition method for complex natural scene

Abstract

The invention relates to the technical field of remote sensing image analysis and processing, in particular to a complex natural scene-oriented land cover remote sensing information acquisition method. The method comprises the following steps: firstly, acquiring all tree clusters, further acquiring a difference separation weight of each environment index in each tree cluster, and adjusting the distance between different tree areas by combining the difference between corresponding environment information of different tree areas under the same environment index to acquire all sub-tree clusters; and finally, carrying out image enhancement, labeling and classification on the tree areas in the sub-tree clusters. The method comprises the following steps: firstly, carrying out preliminary clustering on tree areas based on morphological characteristics of trees, further analyzing differentiation performance of environments where different trees are located in the same cluster, determining the distinguishing capability of each environment index on tree species, then, carrying out reclassification on the trees in the cluster, and carrying out enhancement on the tree areas to different degrees; and the acquisition effect of the remote sensing information of the land cover is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image analysis and processing, and particularly relates to a method for collecting remote sensing information of land cover for complex natural scenes. Background Art

[0002] Land cover refers to various natural and artificial coverings on the earth's surface. Remote sensing technology can monitor large areas in a short time and obtain remote sensing information of land cover, which is an important basis for cover change analysis, ecosystem detection, and resource management. Among them, analyzing the tree species in a certain area based on remote sensing images of land cover can help evaluate the composition of the ecosystem in this area for precise management and protection. Therefore, accurately annotating the vegetation species is crucial when collecting remote sensing information of land cover.

[0003] Currently, methods such as image segmentation or deep learning can be used to classify tree species in remote sensing images of land cover. However, in complex natural environments, due to the heterogeneity of natural scenes and terrain changes, the characteristic manifestations of land cover in remote sensing images may be very complex. The appearance differences of different tree species or even the same tree species in different environments will affect the classification accuracy. At the same time, different tree species may also show similar external manifestations (such as color depth, texture shape, etc.) in remote sensing images, which may lead to misclassification of trees, and further result in poor collection effects of remote sensing information of land cover. Summary of the Invention

[0004] In order to solve the technical problem of poor collection effects of remote sensing information of land cover, the purpose of the present invention is to provide a method for collecting remote sensing information of land cover for complex natural scenes. The specific technical solutions adopted are as follows: Obtain all vegetation areas in the remote sensing image of the area to be measured, and obtain the environmental information of each vegetation area under each environmental index, and all tree areas therein; the environmental index at least includes regional altitude, forest density, and distance to water source; In the remote sensing image, cluster all tree areas according to the morphological and color characteristics of each tree area to obtain all tree clusters; within the preset neighborhood of each tree area, obtain the surrounding environmental complexity of each tree area according to the differences in the environmental information of all vegetation areas under each environmental index; In each of the tree clusters, according to the differences in the number of tree regions corresponding to each type of the surrounding environment complexity, the differential separation weights of each environmental index are obtained, and in combination with the differences between the corresponding environmental information of different tree regions under the same environmental index, the distances between different tree regions are adjusted to obtain all sub-tree clusters; in each sub-tree cluster, according to the environmental information of all tree regions under the environmental index corresponding to the maximum differential separation weight in the tree cluster to which they belong, image enhancement and classification annotation are performed on each tree region.

[0005] Further, the method for obtaining the tree clusters includes: The remote sensing image is converted into the CIE color space, and the mean value of the a-axis values of all pixel points in each tree region is used as the greenness value; the number of corner points in each tree region in the remote sensing image is used as the crown shape parameter. The triple constructed by the greenness value, the crown shape parameter, and the region size of each tree region is used as the characteristic parameter of each tree region, and based on the distance clustering algorithm, the characteristic parameters corresponding to all tree regions in the remote sensing image are clustered, and the corresponding tree clusters are obtained according to the clusters of the characteristic parameters.

[0006] Further, the preset neighborhood of each tree region is all the adjacent vegetation regions of the vegetation region to which each tree region belongs.

[0007] Further, the method for obtaining the surrounding environment complexity includes: Within the preset neighborhood of each tree region, the standard deviation of the environmental information of all adjacent vegetation regions under each environmental index is used as the surrounding environment complexity of each tree region under the corresponding environmental index.

[0008] Further, the method for obtaining the differential separation weights includes: In each of the tree clusters, under each environmental index, with the surrounding environment complexity as the horizontal axis parameter and the number of tree regions corresponding to the surrounding environment complexity as the vertical axis parameter, a characteristic curve is fitted; according to the fluctuation degree of the characteristic curve, the differential separation weights under the corresponding environmental index are obtained.

[0009] Further, the method for obtaining the differential separation weights under the corresponding environmental index according to the fluctuation degree of the characteristic curve includes: In the characteristic curve, according to the change rate between adjacent extreme points and the total number of extreme points, the differential separation weights under the corresponding environmental index are obtained.

[0010] Further, the method for obtaining the sub-tree clusters includes: In each of the tree clusters, for each environmental indicator, using the difference separation weight, the difference between each tree area and the environmental information corresponding to the cluster center is weighted, and the weighted sum value under all environmental indicators is used as the corrected distance between each tree area and the cluster center; all tree areas in each tree cluster are secondarily clustered according to the corrected distance to obtain all sub-tree clusters.

[0011] Further, the method for enhancing each tree area includes: In each sub-tree cluster, the mean of the environmental information under the environmental indicator corresponding to the maximum difference separation weight in the tree cluster to which it belongs is normalized, and the normalized result is used as the enhancement coefficient corresponding to all tree areas in the corresponding sub-tree cluster; the corresponding tree area is enhanced using the enhancement coefficient.

[0012] Further, the method for obtaining the tree area includes: Edge detection is performed on each vegetation area to obtain all closed edges, and the area enclosed by each closed edge is used as a suspected tree area; the mean of the area sizes of all suspected tree areas is used as the reference area, and the suspected tree areas smaller than the reference area are used as tree areas.

[0013] The present invention has the following beneficial effects: The present invention first obtains all vegetation areas in the remote sensing image of the area to be measured, and obtains the environmental information of each vegetation area under each environmental index and all tree areas within it, providing an analysis basis. Then, in the remote sensing image, all tree areas are preliminarily clustered according to the morphological and color characteristics of each tree area to obtain all tree clusters. Further, the surrounding environment complexity of each tree area is obtained, and the surrounding environment complexity provides a certain reference for the growth geographical environment performance of each tree area under each environmental index. For subsequent preparation of evaluating the influence of each environmental index on tree growth to obtain the differential separation weight by combining the distribution of tree areas within the same cluster under similar surrounding environment complexities. Then, in each tree cluster, according to the difference in the number of tree areas corresponding to each surrounding environment complexity, the differential separation weight reflecting the ability of each environmental index to help distinguish different tree species in each tree cluster is obtained, and combined with the difference between the corresponding environmental information of different tree areas under the same environmental index, the distance between different tree areas is adjusted to obtain all sub-tree clusters, and the initially obtained tree clusters are finely divided again. Finally, in each sub-tree cluster, according to the environmental information of all tree areas under the environmental index corresponding to the maximum differential separation weight in the tree cluster to which they belong, image enhancement and classification annotation are performed on each tree area. The present invention first preliminarily clusters tree areas based on the morphological characteristics of trees, further analyzes the differential performance of the environments where different trees are located in the same cluster, determines the discrimination ability of each environmental index for tree species, and then finely classifies the trees in the cluster again, and enhances the tree areas in different sub-clusters to different degrees, improving the acquisition effect of land cover remote sensing information. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0015] Figure 1 It is a flowchart of a method for collecting land cover remote sensing information for a complex natural scene provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following specifically describes, with reference to the accompanying drawings and preferred embodiments, a method for collecting remote sensing information on land cover for complex natural scenes according to the present invention, including its specific implementation manner, structure, features, and effects. 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.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0018] The following specifically describes the specific solution of a method for collecting remote sensing information on land cover for complex natural scenes provided by the present invention with reference to the accompanying drawings.

[0019] Please refer to Figure 1 , which shows the flowchart of a method for collecting remote sensing information on land cover for complex natural scenes provided by an embodiment of the present invention, specifically including: Step S1, obtain all vegetation areas in the remote sensing image of the area to be measured, and obtain the environmental information of each vegetation area under each environmental index, as well as all tree areas within it.

[0020] It should be noted that the implementation scenario targeted by the embodiments of the present invention is the collection of remote sensing information on land cover in summer.

[0021] In an embodiment of the present invention, first, determine the area to be measured, and based on the natural resource satellite remote sensing cloud service platform, obtain the high-resolution original remote sensing image of the area to be measured; then, use the improved cloud masking algorithm of Collection 6 to perform preprocessing on the original remote sensing image, such as reprojection, resampling, smoothing, filtering, and null value filtering, to obtain a high-quality remote sensing image for subsequent analysis. It should be noted that the acquisition and preprocessing of remote sensing images are both prior arts and will not be elaborated herein.

[0022] Since the embodiments of the present invention are for collecting remote sensing information on land cover in complex natural scenes, mainly targeting the distribution of land vegetation and classifying and identifying vegetation, but in addition to vegetation in the area to be measured, there may also be land, water bodies, buildings, etc.; therefore, after obtaining the remote sensing image of the area to be measured in the embodiments of the present invention, all vegetation areas in the remote sensing image are further obtained.

[0023] In one embodiment of the present invention, considering that in the remote sensing images in summer, vegetation has a more special green appearance compared to other coverages; since the obtained remote sensing images are RGB images, in the remote sensing images, the channel value of each pixel point in the G channel is obtained, and then the pixel points with the G channel value greater than a preset threshold, such as 180, are used as target pixel points. The target pixel points are subjected to region connectivity detection, and each connected domain is used as a vegetation region.

[0024] It should be noted that in other embodiments, the implementer can also obtain the vegetation region based on technical means such as threshold segmentation or manual annotation segmentation. These are all existing technical means for obtaining the channel value and region connectivity detection, and will not be elaborated here.

[0025] After obtaining the vegetation region, all tree regions within it can be further obtained. The tree region refers to the crown region of a single plant of vegetation, which is prepared for subsequent classification and annotation of tree species types.

[0026] Preferably, in one embodiment of the present invention, considering that there are gaps between trees and there will be obvious color changes between the gaps and the leaves, all edges can be obtained based on edge detection, and then closed edges can be further obtained. The closed edges may correspond to the crown region or the gap region; also considering that the area corresponding to the crown of the tree is relatively large compared to the gap region; therefore, the method for obtaining the tree region includes: Perform edge detection on each vegetation region to obtain all closed edges, and use the region enclosed by each closed edge as a suspected tree region; use the average value of the region areas of all suspected tree regions as the reference area, and use the suspected tree regions with an area smaller than the reference area as the tree regions.

[0027] As an example, taking any vegetation region as an example, first obtain all edges based on the Canny operator, then perform closed detection on each edge to obtain all closed edges, and then obtain all suspected tree regions; then use the total number of pixel points in the region as the region area, and then obtain the reference area, so as to screen out the tree regions.

[0028] It should be noted that in other embodiments, the implementer can also use the watershed algorithm or a segmentation algorithm based on deep learning to obtain all tree regions within each vegetation region. These are all existing technical means for Canny edge detection and closed detection, and will not be elaborated here.

[0029] Considering that different geographical environments may lead to inconsistent tree growth conditions within the vegetation region, in one embodiment of the present invention, the environmental information of each vegetation region under each environmental indicator is further obtained to characterize the geographical environment of each vegetation region, which is prepared for subsequent division of the tree region based on the different performance of trees under different environmental indicators.

[0030] Among them, in a preferred embodiment of the present invention, the environmental indicators at least include regional altitude, forest density, and water source distance.

[0031] It should be noted that the altitude can be obtained based on open-source DEM data, the forest density can be obtained by calculating the NDVI value based on remote sensing images, and then the water bodies around the area to be measured can be obtained based on the open-source water resource management platform, so as to estimate the water source distance. The acquisition processes are all prior arts and will not be elaborated here; the implementer can also customize the types and quantities of environmental indicators, such as adding tree height, soil pH, etc.

[0032] Step S2: In the remote sensing image, according to the morphological characteristics and color characteristics of each tree area, cluster all tree areas to obtain all tree clusters; within the preset neighborhood of each tree area, under each environmental indicator, according to the differences in the environmental information of all vegetation areas, obtain the surrounding environmental complexity of each tree area.

[0033] Considering that the remote sensing image is an image from a top-down perspective, mainly presenting the tree crowns, and the crown shapes of different tree species are different. Firstly, it is reflected in the different sizes of the crowns. For example, the crowns of arbor trees are relatively larger than those of shrub trees. Secondly, it is reflected in the aggregation of the crowns. For example, the branches of pine trees extend outward and the extension degrees of different branches are different, so its crown has a more layered sense from a top-down perspective, while the branches of apple trees are more aggregated and it appears more round and smooth from a top-down perspective; also considering that although different trees all show green characteristics, there are still significant differences in the leaf colors of different trees; therefore, in the embodiment of the present invention, all tree areas are initially clustered in the remote sensing image to obtain all tree clusters; among them, the tree areas in each tree cluster are tree areas suspected of being the same tree species.

[0034] Preferably, in an embodiment of the present invention, considering that the a-axis in the CIELAB color space can represent the green characteristics of the tree area, which can help to more clearly observe the color differences between different tree areas; the area of the tree area can reflect the size of the crown, and the corner points are characteristic points of significant turning or changing points, usually located at the edges or shape-changing places of objects. Then, the number of corner points in the crown area that is more aggregated and round should be less than that in the crown area with a more layered sense. Therefore, the crown shape can be evaluated based on the number of corner points in the tree area, and then the crown performance of each tree area can be characterized by combining the color characteristics and the area; also considering that clustering can group similar trees into a cluster; based on this, the method for obtaining the tree cluster includes: Convert the remote sensing image into the CIE color space, and take the mean value of the a-axis values of all pixel points in each tree area as the greenness value; take the number of corner points in each tree area in the remote sensing image as the crown shape parameter. Construct a triple consisting of the greenness value, crown shape parameters, and area size of each tree area as the characteristic parameters of each tree area. Based on the distance clustering algorithm, cluster the characteristic parameters corresponding to all tree areas in the remote sensing image, and obtain the corresponding tree clusters according to the clusters of the characteristic parameters.

[0035] As an example, first obtain the greenness value of each tree area, then perform Harris corner detection on each tree area to obtain all corner points, so as to obtain the crown shape parameters of each tree area; then construct a triple to obtain the characteristic parameters of each tree area. Considering that the ISODATA clustering algorithm can be used without pre-setting the number of clusters, cluster the characteristic parameters corresponding to all tree areas in the remote sensing image based on the ISODATA clustering algorithm to obtain several clusters, where all the characteristic parameters in each cluster corresponding to the tree areas form a tree cluster. For ease of understanding, the implementer can also directly construct a spatial coordinate system, map the triple corresponding to each tree area into the spatial coordinate system, and then cluster the tree areas based on the ISODATA clustering algorithm.

[0036] It should be noted that in other examples, the implementer can also use the mode instead of the mean to obtain the greenness value; considering that the crown size of different tree areas may affect the number of corner points, after obtaining the number of corner points of each tree area, it can be further divided by the area of the corresponding tree area to obtain the number of corner points per unit area, eliminating the influence of the crown size and accurately obtaining the crown shape parameters; the implementer can also use other clustering algorithms such as the K-means clustering algorithm, which, together with obtaining the a-axis value of each pixel point based on CIELAB, Harris corner detection, and ISODATA clustering, are all existing technologies and will not be elaborated here.

[0037] Considering that there may be various complex terrains or complex natural scenes in the area to be measured, such as the transitional zone between mountains and hills, and the suitable growth environments of different tree species are different, resulting in possible differences in the types and growth conditions of trees in different vegetation areas; also considering that the environmental information of each vegetation area under each environmental indicator has been obtained, further, within the preset neighborhood of each tree area, under each environmental indicator, according to the differences in the environmental information of all vegetation areas, obtain the surrounding environmental complexity of each tree area. The surrounding environmental complexity provides a certain reference for the growth geographical environment performance of each tree area under each environmental indicator, and prepares for evaluating the impact of each environmental indicator on tree growth to obtain the differential separation weight by combining the distribution of tree areas within the same cluster under similar surrounding environmental complexities under each environmental indicator.

[0038] Preferably, in an embodiment of the present invention, the preset neighborhood of each tree area is first determined, and the preset neighborhood of each tree area is all the neighboring vegetation areas of the vegetation area to which each tree area belongs, so as to analyze the surrounding environment complexity of each tree area subsequently.

[0039] Preferably, in an embodiment of the present invention, considering that the standard deviation can understand the distribution characteristics of data, under each environmental index, if the standard deviation of the environmental information corresponding to all neighboring vegetation areas is larger, it indicates that the degree of dispersion of the environmental information is larger and the environmental information is more complex; therefore, the method for obtaining the surrounding environment complexity includes: Within the preset neighborhood of each tree area, the standard deviation of the environmental information of all neighboring vegetation areas under each environmental index is used as the surrounding environment complexity of each tree area under the corresponding environmental index. It should be noted that the acquisition of the standard deviation is already an existing technology, and the implementer can also use discrete parameters such as variance to measure the surrounding environment complexity, which will not be elaborated here.

[0040] In other embodiments, the implementer can also expand the analysis scope. For example, the remote sensing images in the surrounding areas such as the 8-neighborhood of the area to be measured are mosaicked with the remote sensing image of the area to be measured, and all vegetation areas and corresponding environmental information are obtained, so as to analyze and obtain the surrounding environment complexity of each tree area; mosaicking remote sensing images is already an existing technical means, which will not be elaborated here.

[0041] Step S3, in each tree cluster, according to the difference in the number of tree areas corresponding to each surrounding environment complexity, the difference separation weight of each environmental index is obtained, and in combination with the difference between the environmental information corresponding to different tree areas under the same environmental index, the distance between different tree areas is adjusted to obtain all sub-tree clusters; in each sub-tree cluster, according to the environmental information of each tree area under the environmental index corresponding to the maximum difference separation weight in the tree cluster to which it belongs, image enhancement and classification annotation are performed on each tree area.

[0042] Considering that in the initial clustering and classification, trees with different tree species but similar canopy performances may be classified into the same tree cluster; and considering that the growth environments of trees of the same tree species are relatively similar, in each tree cluster, the distribution of the surrounding environment complexity of tree areas under each environmental index can help reflect the differences in the geographical environment or growth environment corresponding to trees with similar canopies; that is, under different surrounding environment complexities under a certain environmental index, if the difference in the number of tree areas in the same tree cluster is more obvious, it indicates that the tree areas in this cluster may be concentrated in a certain geographical environment, and the stronger the distinguishing ability of this environmental index for distinguishing tree species with similar growth environments, the more convenient it is to separate different tree species; Therefore, in the embodiments of the present invention, in each tree cluster, according to the difference in the number of tree areas corresponding to each surrounding environment complexity, the differential separation weight of each environmental indicator is obtained; the differential separation weight reflects the ability of each environmental indicator to help distinguish different tree species in each tree cluster.

[0043] Preferably, in an embodiment of the present invention, considering that in each tree cluster, for each environmental indicator, with the surrounding environment complexity as the horizontal axis parameter and the number of tree areas corresponding to the surrounding environment complexity as the vertical axis parameter, a characteristic curve is fitted. By analyzing the degree of fluctuation of the characteristic curve, it can help analyze the distribution of the number of tree areas corresponding to each surrounding environment complexity under each environmental indicator, and further evaluate the differential separation weight. Based on this, the method for obtaining the differential separation weight includes: In each tree cluster, for each environmental indicator, with the surrounding environment complexity as the horizontal axis parameter and the number of tree areas corresponding to the surrounding environment complexity as the vertical axis parameter, a characteristic curve is fitted; according to the degree of fluctuation of the characteristic curve, the differential separation weight under the corresponding environmental indicator is obtained.

[0044] Among them, in a preferred embodiment of the present invention, considering that the more the number of extreme points in the characteristic curve and the greater the change rate between adjacent extreme points, it can indicate that the fluctuation change of the characteristic curve is greater. Therefore, the method for obtaining the differential separation weight under the corresponding environmental indicator according to the degree of fluctuation of the characteristic curve includes: In the characteristic curve, according to the change rate between adjacent extreme points and the total number of extreme points, the differential separation weight under the corresponding environmental indicator is obtained.

[0045] As an example, in the characteristic curve, first, all extreme points are obtained. Then, the absolute value of the difference between the vertical axis parameters between adjacent extreme points, that is, the absolute value of the difference between the corresponding numbers of tree areas, is divided by the absolute value of the difference between the horizontal axis parameters between adjacent extreme points, that is, the absolute value of the difference between the corresponding surrounding environment complexities, to obtain the change rate between the corresponding adjacent extreme points. Then, the sum of all change rates is calculated, the sum value is multiplied and combined with the total number of extreme points, and then the product is mapped into the sigmoid function for normalization to obtain the differential separation weight under the environmental indicator corresponding to the characteristic curve.

[0046] It should be noted that fitting the characteristic curve, obtaining extreme points, and normalization are all existing technical means and will not be elaborated here.

[0047] After obtaining the differential separation weights of each environmental indicator in each tree cluster, the characteristics of the similar geographical environments where the same species of trees survive can be further combined to analyze the differences between the corresponding environmental information of different tree regions under each environmental indicator, and the distances between different tree regions can be adjusted in combination with the differential separation weights, so as to perform secondary clustering within the tree cluster to obtain all sub-tree clusters therein, where the tree regions in the sub-tree clusters are the tree regions of the same tree species.

[0048] Preferably, in an embodiment of the present invention, considering that in each tree cluster, the characteristics of the cluster center can reflect the average characteristics of all tree regions, so when the difference between the corresponding environmental information of a tree region and the cluster center under each environmental indicator is greater, it indicates that the two may not be in a similar growth environment. At the same time, if the ability of this environmental indicator to distinguish tree species, that is, the differential separation weight, is greater, it means that the difference in environmental information provided under this environmental indicator has a greater reference value for distinguishing tree species in different growth environments. Based on this, the distances between each tree region and the cluster center can be adjusted, so that the differences between tree regions in different survival environments are greater, in order to accurately distinguish or classify; therefore, the method for obtaining sub-tree clusters includes: In each tree cluster, under each environmental indicator, use the differential separation weight to weight the difference between the corresponding environmental information of each tree region and the cluster center, and obtain the corrected distance between each tree region and the cluster center according to the weighted sum value under all environmental indicators; perform secondary clustering on all tree regions in each tree cluster according to the corrected distance to obtain all sub-tree clusters.

[0049] As an example, in each tree cluster, under each environmental indicator, use the square of the absolute value of the difference to measure the difference between the corresponding environmental information of each tree region and the cluster center, and then multiply the differential separation weight by the square of the absolute value of the difference of the corresponding environmental information to obtain the weighted sum value under each environmental indicator. Take the square root of the weighted sum value as the corrected distance between each tree region and the cluster center; then, according to the corrected distances of all tree regions in the tree cluster relative to the cluster center, map all tree regions into a spatial model, and further perform secondary clustering on the tree regions in the tree cluster based on the ISODATA clustering algorithm to obtain all sub-tree clusters.

[0050] It should be noted that performing spatial mapping and then clustering according to the corrected distance is already a prior art and will not be elaborated here; among them, before performing weighted summation using the differential separation weight, it is necessary to normalize the differential separation weight under each environmental indicator and make the sum of the differential separation weights under all environmental indicators equal to 1; assuming there are three environmental indicators a, b, and c, taking environmental indicator a as an example, the normalization method of its differential separation weight is , where a, b, and c are the serial numbers of the environmental indicators respectively. is the differential separation weight of the a-th environmental indicator, is the differential separation weight of the b-th environmental indicator, is the differential separation weight of the c-th environmental indicator.

[0051] In another embodiment of the present invention, the implementer can also replace the cluster center with other tree areas in each tree cluster. That is, under each environmental indicator, the difference between the environmental information corresponding to two different tree areas is weighted using the differential separation weight, and then the weighted sum value under all environmental indicators is used as the corrected distance between the two different tree areas; the corrected distances between all pairs of different tree areas are obtained, and then mapped into a spatial model for clustering to obtain sub-tree areas, which is the same as the above acquisition method and will not be elaborated here.

[0052] After obtaining all sub-tree clusters, different degrees of enhancement can be further performed on the tree areas of different tree species corresponding to different sub-tree clusters to enhance the distinguishability between the tree areas of different tree species; considering that among the environmental information under the environmental indicator corresponding to the maximum differential separation weight in each tree cluster, it has the strongest ability to distinguish the sub-tree areas of different tree species, so the embodiment of the present invention will enhance each tree area according to the environmental information under the environmental indicator corresponding to the maximum differential separation weight of all tree areas in the tree cluster to which they belong.

[0053] Preferably, in one embodiment of the present invention, the method for enhancing each tree area includes: In each sub-tree cluster, under the environmental indicator corresponding to the maximum differential separation weight in the tree cluster to which it belongs, the mean value of the environmental information of the vegetation areas to which all tree areas belong is normalized, and the normalization result is used as the enhancement coefficient corresponding to all tree areas in the corresponding sub-tree cluster; the corresponding tree areas are enhanced using the enhancement coefficient.

[0054] As an example, specifically, the mean value is mapped into the sigmoid function for normalization. The implementer can also use other normalization functions, and then the enhancement coefficients of all tree areas within each planted tree area can be obtained, and then the corresponding trees are enhanced in terms of brightness or color using the enhancement coefficients, which is already prior art and will not be elaborated here.

[0055] In another embodiment of the present invention, the implementer can also directly regard each sub-tree cluster as a set of tree areas under a certain tree species, and then label the tree areas in different sub-tree clusters with different colors for classification, which will not be elaborated here.

[0056] In summary, the present invention first preliminarily obtains all tree clusters according to the morphological and color characteristics of each tree area; further, under each environmental index, analyzes and obtains the complexity of the surrounding environment of each tree area; then, in each tree cluster, according to the difference in the number of tree areas corresponding to each surrounding environment complexity, obtains the difference separation weight of each environmental index, and combines the differences between the corresponding environmental information of different tree areas under the same environmental index to adjust the distance between different tree areas to obtain all sub-tree clusters; finally, performs image enhancement and annotation classification on each tree area in each sub-tree cluster. The present invention first performs preliminary clustering of tree areas based on the morphological characteristics of trees, further analyzes the differential performance of the environments where different trees in the same cluster are located, determines the discrimination ability of each environmental index for tree species, and then performs refined classification of the trees in the cluster again, and performs different degrees of enhancement on the tree areas in different sub-clusters to improve the acquisition effect of land cover remote sensing information.

[0057] It should be noted that the above sequence of embodiments of the present invention is only for description and does not represent the advantages or disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0058] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. The key point of each embodiment is to illustrate the differences from other embodiments.

Claims

1. A method for collecting remote sensing information of land cover for complex natural scenes, characterized in that, The method includes: Obtaining all vegetation areas in the remote sensing image of the area to be measured, and obtaining the environmental information of each vegetation area under each environmental indicator, and all tree areas therein; the environmental indicators at least include regional altitude, tree density, and water source distance; In the remote sensing image, clustering all tree areas according to the morphological and color characteristics of each tree area to obtain all tree clusters; within the preset neighborhood of each tree area, under each environmental indicator, obtaining the surrounding environmental complexity of each tree area according to the differences in the environmental information of all vegetation areas; In each of the tree clusters, obtaining the differential separation weight of each environmental indicator according to the differences in the number of tree areas corresponding to each surrounding environmental complexity, and combining the differences between the environmental information corresponding to different tree areas under the same environmental indicator to adjust the distances between different tree areas to obtain all sub-tree clusters; in each sub-tree cluster, performing image enhancement and annotation classification on each tree area according to the environmental information of each tree area under the environmental indicator corresponding to the maximum differential separation weight in the tree cluster to which it belongs.

2. The method for collecting remote sensing information on land cover for complex natural scenes according to claim 1, characterized in that The method for obtaining the tree clusters includes: Converting the remote sensing image into the CIE color space, and taking the mean value of the a-axis values of all pixel points of each tree area as the greenness value; taking the number of corner points in each tree area in the remote sensing image as the crown shape parameter; Taking the triple constructed by the greenness value, the crown shape parameter, and the area size of each tree area as the characteristic parameter of each tree area, and clustering the characteristic parameters corresponding to all tree areas in the remote sensing image based on the distance clustering algorithm, and obtaining the corresponding tree clusters according to the clusters of the characteristic parameters.

3. A method for collecting remote sensing information on land cover for complex natural scenes according to claim 1, characterized in that, The preset neighborhood of each tree area is all the adjacent vegetation areas of the vegetation area to which each tree area belongs.

4. The method for collecting remote sensing information of land cover for complex natural scenes according to claim 3, characterized in that, The method for obtaining the surrounding environmental complexity includes: Within the preset neighborhood of each tree area, taking the standard deviation of the environmental information of all adjacent vegetation areas under each environmental indicator as the surrounding environmental complexity of each tree area under the corresponding environmental indicator.

5. A method for remotely sensing information acquisition of land cover for complex natural scenes according to claim 1, characterized in that The method for obtaining the differential separation weight includes: In each of the tree clusters, under each environmental indicator, taking the surrounding environmental complexity as the horizontal axis parameter and the number of tree areas corresponding to the surrounding environmental complexity as the vertical axis parameter to fit the characteristic curve; obtaining the differential separation weight under the corresponding environmental indicator according to the fluctuation degree of the characteristic curve.

6. The method for collecting remote sensing information on land cover for complex natural scenes according to claim 5, characterized in that The method for obtaining the differential separation weight under the corresponding environmental indicator according to the fluctuation degree of the characteristic curve includes: In the characteristic curve, obtaining the differential separation weight under the corresponding environmental indicator according to the change rate between adjacent extreme points and the total number of extreme points.

7. A method for collecting remote sensing information on land cover for complex natural scenes according to claim 1, characterized in that The method for obtaining the sub-tree clusters includes: In each of the tree clusters, for each environmental indicator, using the difference separation weight, the difference between each tree area and the environmental information corresponding to the cluster center is weighted, and the weighted sum value under all environmental indicators is sensed to obtain the corrected distance between each tree area and the cluster center; all tree areas in each tree cluster are secondarily clustered according to the corrected distance to obtain all sub-tree clusters.

8. A method for collecting remote sensing information of land cover for complex natural scenes according to claim 1, characterized in that The method for enhancing each tree area includes: In each sub-tree cluster, the mean value of the environmental information under the environmental indicator corresponding to the maximum difference separation weight in the tree cluster to which it belongs is normalized, and the normalization result is used as the enhancement coefficient corresponding to all tree areas in the corresponding sub-tree cluster; the corresponding tree area is enhanced using the enhancement coefficient.

9. A method for collecting remote sensing information on land cover for complex natural scenes according to claim 1, characterized in that, The method for obtaining the tree area includes: Edge detection is performed on each vegetation area to obtain all closed edges, and the area enclosed by each closed edge is used as a suspected tree area; the mean value of the area of all suspected tree areas is used as the reference area, and the suspected tree areas smaller than the reference area are used as tree areas.

Images