Automatic monitoring method and system for slope vegetation coverage area
By using superpixel segmentation of consecutive frame remote sensing images and temporal consistency guidance of seed matching points, the problem of inaccurate monitoring results from a single remote sensing image was solved, enabling efficient and accurate monitoring of slope vegetation cover area.
Patent Information
- Application Number
- CN202511267762.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-05
- Publication Date
- 2026-02-17
- Estimated Expiration
- 2045-09-05
AI Technical Summary
The results of monitoring the vegetation cover area of slopes obtained from a single remote sensing image are inaccurate. They are affected by factors such as changes in light conditions, differences in vegetation phenology, obstruction by surrounding tall vegetation, and landslides, and may easily misidentify newly sprouted vegetation as rocks.
A superpixel segmentation method based on consecutive frame remote sensing images is adopted. Using the superpixel segmentation results of prior information remote sensing images as guidance, the method determines the temporal consistency and color space differences of seed matching points, and combines optical flow calculation and superpixel segmentation algorithm to obtain the superpixel segmentation results of the target remote sensing image, thereby improving the monitoring accuracy.
It improves the accuracy and reliability of slope vegetation cover area monitoring, and by guiding the segmentation of adjacent remote sensing images, it supplements the mobility and temporality of monitoring and reduces misjudgment errors.
Smart Images
Figure CN121170577B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of image segmentation, and particularly relates to a method and system for automatically monitoring the coverage area of slope vegetation. BACKGROUND
[0002] Slope vegetation refers to vegetation growing on the slope surface, including herbaceous plants, shrubs and small trees, etc. The slope vegetation can achieve the purposes of soil and water conservation, maintaining the stability of the ecological system, enhancing the strength of the slope and reducing the risk of erosion. Monitoring the coverage area of the slope vegetation can predict the risks of vegetation degradation and landslides in advance and make timely early warnings.
[0003] Generally, the coverage area of the slope can be identified according to the remote sensing image of the area to be monitored, and then the area of the coverage area of the slope can be calculated. However, due to the influence of factors such as changes in light conditions, differences in vegetation phenology, obstruction of surrounding tall vegetation, landslides and the like, the monitoring result of the coverage area of the slope vegetation obtained from a single remote sensing image is often inaccurate. For example, the vegetation just sprouting is close to the rock and is easily mistaken for rock. SUMMARY
[0004] The present application provides a method and system for automatically monitoring the coverage area of slope vegetation to solve the problem of inaccurate monitoring result of the coverage area of slope vegetation obtained from a single remote sensing image. The technical solution is as follows:
[0005] In the first aspect, an embodiment of the present application provides a method for automatically monitoring the coverage area of slope vegetation, which comprises the following steps:
[0006] Collecting remote sensing images of consecutive frames of the area to be monitored of the coverage area of the slope, performing superpixel segmentation on the first frame of remote sensing image, obtaining the superpixel segmentation result of the first frame of remote sensing image, and the superpixel segmentation result comprises superpixel regions and seed points of the superpixel regions;
[0007] An arbitrary frame remote sensing image starting from the second frame remote sensing image is recorded as a target remote sensing image, and a remote sensing image adjacent to a frame before the target remote sensing image is recorded as prior information remote sensing image; seed matching points corresponding to seed points of a superpixel region in the target remote sensing image are determined according to a superpixel segmentation result of the prior information remote sensing image, and coordinates of the pixel points in the target remote sensing image and the prior information remote sensing image; time consistency of the corresponding seed matching points and the seed points and a time consistency weight of the seed matching points are respectively determined according to differences between color values of the corresponding seed matching points and the seed points in a Lab color space, and differences between the coordinates of the corresponding seed matching points and the seed points in the target remote sensing image and the prior information remote sensing image; the target remote sensing image is subjected to superpixel segmentation, color distance and spatial distance of each pixel point and the seed points are obtained, the adjustment distance of the pixel points and the seed points in the target remote sensing image is determined by combining the time consistency and the time consistency weight of the seed points, and the superpixel segmentation result of the target remote sensing image is obtained by combining the seed matching points in the target remote sensing image;
[0008] The slope vegetation coverage area is monitored according to the superpixel segmentation result of the target remote sensing image.
[0009] Further, the method of determining the seed matching points corresponding to the seed points of the superpixel region in the target remote sensing image according to the superpixel segmentation result of the prior information remote sensing image, and the coordinates of the pixel points in the target remote sensing image and the prior information remote sensing image comprises the following specific method:
[0010] An arbitrary superpixel region in the prior information remote sensing image is recorded as a target superpixel region, the number of pixel points contained in the target superpixel region is recorded as the area S of the target superpixel region, a point with the same coordinate as the seed point in the target remote sensing image is recorded as a seed matching point, and a local window with a side length of 2n is established with the seed matching point as the center and is recorded as a matching local window of the seed matching point; wherein, n is a positive integer. The optical flow of the matching local window and the corresponding pixel points in the target superpixel region is calculated. represents a down rounding value of .
[0011] The optical flow of the matching local window of the seed matching point and the corresponding pixel points in the target superpixel region is calculated.
[0012] The coordinates of the seed matching points corresponding to the seed points of the superpixel region in the target remote sensing image are updated according to the optical flow of the corresponding superpixel region and the matching local window.
[0013] Further, the method of updating the coordinates of the seed matching points corresponding to the seed points of the superpixel region in the target remote sensing image according to the optical flow of the corresponding superpixel region and the matching local window comprises the following specific method:
[0014] The median value of the displacement components of the optical flow in the horizontal direction of all corresponding pixel points in the target superpixel region in the matching local window of the seed matching point is recorded as the horizontal coordinate change of the seed matching point; the median value of the displacement components of the optical flow in the vertical direction of all corresponding pixel points in the target superpixel region in the matching local window of the seed matching point is recorded as the vertical coordinate change of the seed matching point; the difference between the horizontal coordinate of the seed matching point and the horizontal coordinate change is taken as the updated horizontal coordinate of the seed matching point, and the difference between the vertical coordinate of the seed matching point and the vertical coordinate change is taken as the updated vertical coordinate of the seed matching point.
[0015] Further, the method for determining the time consistency of the corresponding seed matching point and the seed point is:
[0016] The Euclidean distance between the color values of the seed matching point and the corresponding seed point in the Lab color space is recorded as the first distance between the corresponding seed matching point and the seed point.
[0017] The Euclidean distance between the coordinates of the seed matching point in the target remote sensing image and the coordinates of the corresponding seed point in the prior information remote sensing image is recorded as the second distance between the corresponding seed matching point and the seed point.
[0018] The product of the square of the first distance between the corresponding seed matching point and the seed point and a preset first coefficient is recorded as the first product of the corresponding seed matching point and the seed point, the product of the square of the second distance between the corresponding seed matching point and the seed point and a preset second coefficient is recorded as the second product of the corresponding seed matching point and the seed point, and the arithmetic square root of the sum of the first product and the second product of the corresponding seed matching point and the seed point is recorded as the time consistency of the corresponding seed matching point and the seed point.
[0019] Further, the method for determining the time consistency weight of the seed matching point is:
[0020] The matching pixel points in the target remote sensing image and the prior information remote sensing image are obtained, the average value of the absolute values of the differences between the pixel values in the a channel in the Lab color space of all pixel points in the matching local window of the seed matching point and the matching pixel points in the prior information remote sensing image is recorded as the local green difference of the seed matching point.
[0021] The difference between the digital 1 and the normalized value of the local green difference of the seed matching point is recorded as the time consistency weight of the seed matching point.
[0022] Further, the calculation formula of the adjusted distance between the pixel point in the target remote sensing image and the seed point is:
[0023]
[0024] wherein, D u,vdenotes the adjustment distance between the pixel point v and the seed point u in the target remote sensing image; d1 u,v denotes the color distance between the pixel point v and the seed point u in the target remote sensing image; d1 u denotes the maximum value of the color distance corresponding to the seed point u in the target remote sensing image; d2 u,v denotes the spatial distance between the pixel point v and the seed point u in the target remote sensing image; d2 u denotes the maximum value of the spatial distance corresponding to the seed point u in the target remote sensing image; d3 u denotes the time consistency of the seed point u in the target remote sensing image; γ u denotes the time consistency weight of the seed point u in the target remote sensing image.
[0025] Further, the method comprises the following steps of:
[0026] The seed matching point in the target remote sensing image is taken as the seed point, the adjustment distance between the pixel point and the seed matching point in the target remote sensing image is taken as the comprehensive distance between the pixel point and the seed point in the superpixel segmentation algorithm, and the superpixel segmentation algorithm is used to obtain the superpixel segmentation result of the target remote sensing image.
[0027] Further, the method comprises the following steps of:
[0028] For each superpixel region in the superpixel segmentation result of the target remote sensing image:
[0029] The number of pixel points contained in the superpixel region is denoted as the area of the superpixel region, the number of pixel points contained in the edge of the superpixel region is denoted as the side length of the superpixel region, the product of the area of the superpixel region and 4π is denoted as the third product of the superpixel region, the square of the side length of the superpixel region is denoted as the first square of the superpixel region, and the ratio of the third product of the superpixel region to the first square is denoted as the compactness of the superpixel region.
[0030] The variance of the pixel values of all pixel points in the superpixel region in the a channel of the Lab color space is denoted as the green difference of the superpixel region.
[0031] The mean value of all numerical values in the gray level co-occurrence matrix of the superpixel region is denoted as the texture complexity of the superpixel region.
[0032] The normalized vegetation index of the target remote sensing image and the prior information remote sensing image is calculated, and the difference between the normalized vegetation indexes of the target remote sensing image and the prior information remote sensing image is denoted as the vegetation feature value of the target remote sensing image.
[0033] The compactness, green difference, texture complexity of the superpixel region and the vegetation feature value of the target remote sensing image where the superpixel region is located are sequentially arranged to obtain a feature vector of the superpixel region.
[0034] According to the feature vector of the superpixel region, the monitoring of the slope vegetation coverage area is completed.
[0035] Further, the monitoring of the slope vegetation coverage area according to the feature vector of the superpixel region comprises the following specific method:
[0036] All superpixel regions obtained by using random forest on the superpixel segmentation result of the target remote sensing image are classified to obtain the classification result of each superpixel region, and the classification result of the superpixel region includes two kinds of vegetation region and non-vegetation region.
[0037] The cumulative sum of the actual area of all superpixel regions with the classification result of the vegetation region in the slope coverage area to be monitored is recorded as the slope vegetation coverage area corresponding to the target remote sensing image.
[0038] In the second aspect, the embodiments of the present application further provide an automatic monitoring system for a slope vegetation coverage area, comprising a memory, a processor and a computer program stored in the memory and running on the processor, and the processor implements the steps of the method in any one of the above aspects when executing the computer program.
[0039] The present application has the following beneficial effects:
[0040] The present application takes the image segmentation result of the former one of the two adjacent remote sensing images as prior knowledge to guide the image segmentation of the latter one of the two adjacent remote sensing images, and supplements the maneuverability and time sequence for the slope vegetation coverage area monitoring, and further improves the accuracy of the detection result. Specifically, first, the seed matching point corresponding to the seed point in the target remote sensing image is determined, and the time consistency weight and the influence weight of the time consistency of the seed matching point are determined. The time consistency weight of the seed matching point is the determination result of the influence weight of the time consistency of the seed matching point and the seed matching point. The time consistency weight and the time consistency are the prior knowledge result of the image segmentation of the former one of the two adjacent remote sensing images, and the adjustment weight and the adjustment amount of the distance between the seed point and the pixel point are determined. Further, the adjustment distance between the pixel point and the seed point in the target remote sensing image is determined by combining the color distance and the spatial distance between each pixel point and the seed point in the superpixel segmentation result of the target remote sensing image, and the superpixel segmentation result of the target remote sensing image is obtained by using the seed matching point in the target remote sensing image as the seed point and the adjustment distance between the pixel point and the seed matching point in the target remote sensing image as the comprehensive distance between the pixel point and the seed point in the superpixel segmentation algorithm. Finally, the slope vegetation coverage area monitoring is completed according to the superpixel segmentation result of the target remote sensing image, and the problem of inaccurate slope vegetation coverage area monitoring result obtained from a single remote sensing image is solved, and the accuracy and reliability of the slope vegetation coverage area monitoring are improved. BRIEF DESCRIPTION OF DRAWINGS
[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.
[0042] Figure 1 The flowchart of the automatic slope vegetation coverage area monitoring method provided by an embodiment of the present application;
[0043] Figure 2 The matching local window acquisition flowchart provided by an embodiment of the present application. DETAILED DESCRIPTION
[0044] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work belong to the protection scope of the present application.
[0045] Please refer to Figure 1 which shows a flow chart of a method for automatically monitoring the vegetation coverage area of a slope according to an embodiment of the present application, the method comprising the following steps:
[0046] In step S001, remote sensing images of consecutive frames of a region to be monitored for the coverage area of a slope are collected, superpixel segmentation is performed on a first remote sensing image, and a superpixel segmentation result of the first remote sensing image is obtained, the superpixel segmentation result comprising superpixel regions and seed points of the superpixel regions.
[0047] A multispectral remote sensing camera is mounted on a UAV, and the UAV is repeatedly flown over the region to be monitored for the coverage area of the slope.
[0048] The remote sensing images are denoised to avoid the influence of noise on subsequent analysis and improve the accuracy of monitoring the vegetation coverage area of the slope.
[0049] In this embodiment, bilateral filtering is used to denoise each remote sensing image. In actual application, other methods such as Gaussian filtering and median filtering can be used to denoise the remote sensing images, which are not limited herein.
[0050] Due to the influence of factors such as changes in light conditions, differences in vegetation phenology, obstruction by surrounding tall vegetation, and landslides, the monitoring result of the vegetation coverage area of the slope obtained from a single remote sensing image is often inaccurate, for example, it is easy to misjudge the just-budding vegetation as rock. Therefore, the image segmentation result of the previous remote sensing image in the two adjacent remote sensing images is used as prior knowledge to guide the image segmentation of the subsequent remote sensing image in the two adjacent remote sensing images, and the timeliness and maneuverability of monitoring the vegetation coverage area of the slope are supplemented, and the accuracy of the detection result is further improved.
[0051] The first remote sensing image collected is subjected to superpixel segmentation, and a superpixel segmentation result of the first remote sensing image is obtained, a total of K superpixel regions and seed points of each superpixel region.
[0052] In the superpixel segmentation, the number of superpixel regions should be greater than or equal to 100 and less than or equal to 2000, the number of iterations of the superpixel segmentation should be greater than or equal to 5 and less than or equal to 20, K represents the first preset threshold, the value of the first preset threshold is 100 in this embodiment, and the number of iterations of the superpixel segmentation is set to 15.
[0053] At this point, the remote sensing images of the continuous frames and the superpixel segmentation result of the first frame remote sensing image are obtained.
[0054] In step S002, any one remote sensing image starting from the second frame remote sensing image is recorded as a target remote sensing image, and the remote sensing image adjacent to the previous frame of the target remote sensing image is recorded as a prior information remote sensing image; according to the superpixel segmentation result of the prior information remote sensing image, the coordinates of the pixel points in the target remote sensing image and the prior information remote sensing image, the seed matching point corresponding to the seed point of the superpixel region in the target remote sensing image is determined; according to the difference between the color values of the corresponding seed matching point and the seed point in the Lab color space, the difference between the coordinates in the target remote sensing image and the prior information remote sensing image respectively, the temporal consistency of the corresponding seed matching point and the seed point is determined respectively, and the temporal consistency weight of the seed matching point is determined; the target remote sensing image is superpixel segmented to obtain the color distance and the spatial distance between each pixel point and the seed point, the adjustment distance between the pixel point and the seed point in the target remote sensing image is determined by combining the temporal consistency and the temporal consistency weight of the seed point, and the superpixel segmentation result of the target remote sensing image is obtained by combining the seed matching point in the target remote sensing image.
[0055] In step S002, any one remote sensing image starting from the second frame remote sensing image is recorded as a target remote sensing image, and the remote sensing image adjacent to the previous frame of the target remote sensing image is recorded as a prior information remote sensing image.
[0056] According to the superpixel segmentation result of the prior information remote sensing image, the corresponding local segmentation region of each superpixel region in the target remote sensing image, the seed matching point corresponding to the seed point of each superpixel region in the target remote sensing image, and the optical flow of the corresponding superpixel region and the local segmentation region are determined.
[0057] Preferably, as an embodiment of the present application, any one superpixel region in the prior information remote sensing image is recorded as a target superpixel region, the number of pixel points contained in the target superpixel region is recorded as the area of the target superpixel region; the point in the target remote sensing image with the same coordinate as the pixel point in the prior information remote sensing image is recorded as the matching point of the pixel point in the prior information remote sensing image in the target remote sensing image, the matching point of the seed point of the target superpixel region in the target remote sensing image is recorded as the seed matching point, a local window with a side length of 2*2 is established with the seed matching point as the center, and the local window is recorded as the matching local window of the seed matching point.
[0058] The matching local window acquisition flowchart is shown in Figure 2
[0059] The optical flow of the matching local window of the seed matching point and the corresponding pixel point in the target superpixel region is calculated using the optical flow method.
[0060] wherein S represents an area of the target superpixel region, represents a down-round value of .
[0061] It can be understood that the matching local window of the seed matching point in the target remote sensing image is a corresponding local region of the target superpixel region in the prior information remote sensing image.
[0062] According to the optical flow of the corresponding superpixel region and the matching local window, the coordinates of the seed point of the target superpixel region in the corresponding seed matching point in the target remote sensing image are updated.
[0063] The median value of the displacement component of the optical flow in the horizontal direction of all corresponding pixel points in the matching local window of the seed matching point and the target superpixel region is recorded as the horizontal coordinate change of the seed matching point; the median value of the displacement component of the optical flow in the vertical direction of all corresponding pixel points in the matching local window of the seed matching point and the target superpixel region is recorded as the vertical coordinate change of the seed matching point; the coordinates of the seed matching point are updated, the difference between the horizontal coordinate of the seed matching point and the horizontal coordinate change is taken as the updated horizontal coordinate of the seed matching point, and the difference between the vertical coordinate of the seed matching point and the vertical coordinate change is taken as the updated vertical coordinate of the seed matching point.
[0064] According to the same method, the seed matching point of each seed point in the prior information remote sensing image in the target remote sensing image can be obtained.
[0065] According to the difference between the color values of the corresponding seed matching point and the seed point in the Lab color space, the difference between the coordinates in the target remote sensing image and the prior information remote sensing image, the time consistency of the corresponding seed matching point and the seed point is determined.
[0066] The Euclidean distance between the color values of the seed matching point and the corresponding seed point in the Lab color space is recorded as the first distance between the corresponding seed matching point and the seed point; the Euclidean distance between the coordinates of the seed matching point in the target remote sensing image and the coordinates of the corresponding seed point in the prior information remote sensing image is recorded as the second distance between the corresponding seed matching point and the seed point; the product of the square of the first distance between the corresponding seed matching point and the seed point and the first coefficient is recorded as the first product between the corresponding seed matching point and the seed point; the product of the square of the second distance between the corresponding seed matching point and the seed point and the second coefficient is recorded as the second product between the corresponding seed matching point and the seed point; the arithmetic square root of the sum of the first product and the second product between the corresponding seed matching point and the seed point is recorded as the time consistency between the corresponding seed matching point and the seed point.
[0067] Wherein, the first coefficient and the second coefficient are both preset parameters, the sum of the first coefficient and the second coefficient is 1, the value of the first coefficient is 0.4, and the value of the second coefficient is 0.6 in the embodiment.
[0068] Since the vegetation coverage area may change before and after the target remote sensing image and the prior information remote sensing image are aligned, the time consistency weight of the seed matching point and the seed point is determined according to the change of the green color presented by the corresponding positions of the target remote sensing image and the prior information remote sensing image.
[0069] The feature point matching of the target remote sensing image and the prior information remote sensing image is performed, the matching pixel points in the target remote sensing image and the prior information remote sensing image are obtained, the mean value of the absolute value of the difference of the pixel value of the a channel in the Lab color space of all the pixel points in the matching local window of the seed matching point and the matching pixel points in the prior information remote sensing image is recorded as the local green difference of the seed matching point. The difference between the number 1 and the normalized value of the local green difference of the seed matching point is recorded as the time consistency weight of the seed matching point.
[0070] It can be understood that the time consistency weight of the seed matching point is the determination result of the time consistency weight of the seed matching point and the seed point.
[0071] Wherein, the feature point matching is realized by using the SIFT (Scale-Invariant Feature Transform) algorithm in the embodiment, and the feature point matching is a known technology and will not be described herein. In the actual application process, as other implementation manners, the implementer can select the algorithm for realizing the feature point matching according to the actual situation, and the present application does not make special limitation.
[0072] The superpixel segmentation of the target remote sensing image is performed, the color distance and the spatial distance of each pixel point and the seed point are obtained, and the adjustment distance of the pixel point and the seed point in the target remote sensing image is determined according to the color distance and the spatial distance of the pixel point and the seed point, and the time consistency and the time consistency weight of the seed point.
[0073] Specifically, the calculation formula of the adjustment distance of the pixel point and the seed point in the target remote sensing image is as follows:
[0074]
[0075] Wherein, D u,v represents the adjustment distance of the pixel point v and the seed point u in the target remote sensing image; d1 u,v represents the color distance of the pixel point v and the seed point u in the target remote sensing image; d1 u represents the maximum value of the color distance corresponding to the seed point u in the target remote sensing image; d2 u,vdenotes the spatial distance between a pixel v and a seed point u in a target remote sensing image;d2 u denotes the maximum spatial distance corresponding to a seed point u in a target remote sensing image;d3 u denotes the temporal consistency of a seed point u in a target remote sensing image;γ u denotes the weight of the temporal consistency of a seed point u in a target remote sensing image.
[0076] The seed matching point in the target remote sensing image is taken as the seed point, the adjusted distance between the pixel point and the seed matching point in the target remote sensing image is taken as the comprehensive distance between the pixel point and the seed point in the superpixel segmentation algorithm, and the superpixel segmentation algorithm is used to obtain the superpixel segmentation result of the target remote sensing image.
[0077] Thus, the superpixel segmentation result of the target remote sensing image is obtained.
[0078] Step S003, according to the superpixel segmentation result of the target remote sensing image, the monitoring of the slope vegetation coverage area is completed.
[0079] For each superpixel region in the superpixel segmentation result of the target remote sensing image, the number of pixel points contained in the superpixel region is recorded as the area of the superpixel region, the number of pixel points contained in the edge of the superpixel region is recorded as the side length of the superpixel region, and the product of the area of the superpixel region and 4π is recorded as the third product of the superpixel region; the square of the side length of the superpixel region is recorded as the first square of the superpixel region; and the ratio of the third product of the superpixel region to the first square is recorded as the compactness of the superpixel region.
[0080] For each superpixel region in the superpixel segmentation result of the target remote sensing image, the variance of the pixel value of the a channel in the Lab color space of all pixel points in the superpixel region is recorded as the green difference of the superpixel region.
[0081] For each superpixel region in the superpixel segmentation result of the target remote sensing image, the gray level co-occurrence matrix of the superpixel region is obtained, and the mean value of all values in the gray level co-occurrence matrix of the superpixel region is recorded as the texture complexity of the superpixel region.
[0082] According to the target remote sensing image, the reflectivity of the near-infrared band and the infrared band is obtained, and according to the reflectivity of the near-infrared band and the infrared band, the normalized difference vegetation index NDVI of the target remote sensing image is calculated.
[0083] The normalized difference vegetation index of the prior information remote sensing image can be obtained in the same way.
[0084] The difference between the normalized difference vegetation indices of the target remote sensing image and the prior information remote sensing image is recorded as the vegetation feature value of the target remote sensing image.
[0085] The compactness, green difference, texture complexity of the superpixel region and the vegetation feature value of the target remote sensing image where the superpixel region is located are sequentially arranged to obtain a feature vector of the superpixel region.
[0086] All superpixel regions obtained by using the random forest on the superpixel segmentation result of the target remote sensing image are classified to obtain a classification result of each superpixel region, and the classification result of the superpixel region includes two kinds of vegetation region and non-vegetation region.
[0087] When the random forest is used to classify the superpixel region, the number of decision trees is set to 100, the maximum depth of the tree is set to 15, the minimum sample required for splitting is set to 2, and the minimum sample number of the leaf node is set to 1.
[0088] The cumulative sum of the actual area corresponding to all superpixel regions with a classification result of vegetation region in the slope coverage area to be monitored is recorded as the slope vegetation coverage area corresponding to the target remote sensing image.
[0089] According to the same method, the slope vegetation coverage area corresponding to each frame of remote sensing image starting from the second frame of remote sensing image can be obtained respectively.
[0090] At this point, the monitoring of the slope vegetation coverage area is completed.
[0091] Based on the same inventive concept as the above method, the embodiments of the present application also provide an automatic monitoring system for the slope vegetation coverage area, which comprises a memory, a processor and a computer program stored in the memory and running on the processor, and the processor implements the steps of any one of the above automatic monitoring methods for the slope vegetation coverage area.
[0092] The above is only the preferred embodiment of the present application, and is not used to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the principle of the present application shall be included in the protection scope of the present application.
Claims
1. A method for automatic monitoring of the area of vegetation cover of a slope, characterized in that, The method comprises the following steps: Collecting remote sensing images of continuous frames of a region to be monitored in a slope coverage area, performing superpixel segmentation on a first frame of the remote sensing images to obtain a superpixel segmentation result of the first frame of the remote sensing images, wherein the superpixel segmentation result comprises superpixel regions and seed points of the superpixel regions; Taking any frame of remote sensing images starting from a second frame of remote sensing images as a target remote sensing image, taking a frame of remote sensing images adjacent to a previous frame of the target remote sensing image as prior information remote sensing images; determining a corresponding seed matching point of a seed point of a superpixel region in the target remote sensing image according to a superpixel segmentation result of the prior information remote sensing images, coordinates of pixel points in the target remote sensing image and the prior information remote sensing images; determining time consistency of the corresponding seed matching point and the seed point according to a difference between color values of the corresponding seed matching point and the seed point in a Lab color space, a difference between coordinates of the corresponding seed matching point and the seed point in the target remote sensing image and the prior information remote sensing images, respectively, and a time consistency weight of the seed matching point, wherein the time consistency weight of the seed matching point is determined by: obtaining matching pixel points in the target remote sensing image and the prior information remote sensing images, taking a mean value of absolute values of differences between pixel values of a channel a in the Lab color space of all pixel points in a matching local window of the seed matching point and the matching pixel points in the prior information remote sensing images as a local green difference of the seed matching point; taking a difference between a number 1 and a normalized value of the local green difference of the seed matching point as the time consistency weight of the seed matching point, wherein the matching local window of the seed matching point is a local window of a preset side length centered on the seed matching point; performing superpixel segmentation on the target remote sensing image to obtain color distance and spatial distance of each pixel point and the seed point, combining time consistency of the seed point and the time consistency weight to determine an adjustment distance of the pixel point and the seed point in the target remote sensing image, taking the seed matching point in the target remote sensing image as the seed point, taking the adjustment distance of the pixel point and the seed matching point in the target remote sensing image as a comprehensive distance of the pixel point and the seed point in a superpixel segmentation algorithm, and using the superpixel segmentation algorithm to obtain a superpixel segmentation result of the target remote sensing image; Monitoring a slope vegetation coverage area according to the superpixel segmentation result of the target remote sensing image; The adjustment distance of the pixel point and the seed point in the target remote sensing image is calculated according to the following formula: wherein, denotes the spatial distance of a pixel point in the target remote sensing image to the seed point; denotes the color distance of a pixel point in the target remote sensing image to the seed point; denotes the maximum value of the color distance of the seed point in the target remote sensing image corresponding to the seed point; denotes the spatial distance of a pixel point in the target remote sensing image to the seed point; denotes the maximum value of the spatial distance of the seed point in the target remote sensing image corresponding to the seed point; denotes the temporal consistency of the seed point in the target remote sensing image denotes the temporal consistency weight of the seed point in the target remote sensing image . 2. The method for automatic monitoring of the area of vegetation cover of a slope according to claim 1, characterized in that, The specific method of determining the corresponding seed matching point of the seed point of the superpixel region in the target remote sensing image according to the superpixel segmentation result of the prior information remote sensing images, the coordinates of the pixel points in the target remote sensing image and the prior information remote sensing images comprises: The prior information is denoted as a target superpixel region in a remote sensing image, and the number of pixel points contained in the target superpixel region is denoted as the area of the target superpixel region ; The point in the target remote sensing image which is the same as the seed point coordinate in the prior information remote sensing image is recorded as a seed matching point; a local window with a side length of established with the seed matching point as the center is recorded as a matching local window of the seed matching point; the optical flow of the matching local window and the corresponding pixel point in the target superpixel region is calculated; wherein, represents the down-round value of . Calculating an optical flow of a matching local window of the seed matching point and a corresponding pixel point in a target superpixel region; Updating the coordinates of the corresponding seed matching point of the seed point of the superpixel region in the target remote sensing image according to the optical flow of the corresponding superpixel region and the matching local window.
3. The method for automatic monitoring of the area of vegetation cover of a slope according to claim 2, characterized in that, The specific method of updating the coordinates of the corresponding seed matching point of the seed point of the superpixel region in the target remote sensing image according to the optical flow of the corresponding superpixel region and the matching local window comprises: The median value of the displacement components of the optical flow of all corresponding pixel points in the target superpixel region in the horizontal direction in the matching local window of the seed matching point is recorded as the horizontal coordinate change of the seed matching point; the median value of the displacement components of the optical flow of all corresponding pixel points in the target superpixel region in the vertical direction in the matching local window of the seed matching point is recorded as the vertical coordinate change of the seed matching point; The difference between the horizontal coordinate of the seed matching point and the horizontal coordinate change is taken as the updated horizontal coordinate of the seed matching point, and the difference between the vertical coordinate of the seed matching point and the vertical coordinate change is taken as the updated vertical coordinate of the seed matching point.
4. The method for automatic monitoring of the area of vegetation cover of a slope according to claim 1, characterized in that, The method for determining the time consistency of the corresponding seed matching point and the seed point comprises the following steps: The Euclidean distance between the color value of the seed matching point and the corresponding seed point in the Lab color space is recorded as the first distance between the corresponding seed matching point and the seed point. The Euclidean distance between the coordinates of the seed matching point in the target remote sensing image and the coordinates of the corresponding seed point in the prior information remote sensing image is recorded as the second distance between the corresponding seed matching point and the seed point. The product of the square of the first distance between the corresponding seed matching point and the seed point and a preset first coefficient is recorded as the first product of the corresponding seed matching point and the seed point, the product of the square of the second distance between the corresponding seed matching point and the seed point and a preset second coefficient is recorded as the second product of the corresponding seed matching point and the seed point, and the arithmetic square root of the sum of the first product and the second product of the corresponding seed matching point and the seed point is recorded as the time consistency of the corresponding seed matching point and the seed point.
5. The method for automatic monitoring of the area of vegetation cover of a slope according to claim 1, characterized in that, The method for monitoring the slope vegetation coverage area according to the superpixel segmentation result of the target remote sensing image comprises the following steps: For each superpixel region in the superpixel segmentation result of the target remote sensing image: The number of pixel points contained in the superpixel region is denoted as the area of the superpixel region, the number of pixel points contained in the edge of the superpixel region is denoted as the side length of the superpixel region, the product of the area of the superpixel region and the side length of the superpixel region is denoted as the third product of the superpixel region, the square of the side length of the superpixel region is denoted as the first square of the superpixel region, and the ratio of the third product of the superpixel region to the first square is denoted as the compactness of the superpixel region. The number of pixel points contained in the superpixel region is denoted as the area of the superpixel region, the number of pixel points contained in the edge of the superpixel region is denoted as the side length of the superpixel region, the product of the area of the superpixel region and the side length of the superpixel region is denoted as the third product of the superpixel region, the square of the side length of the superpixel region is denoted as the first square of the superpixel region, and the The variance of the pixel value of all pixel points in the a channel in the Lab color space in the superpixel region is recorded as the green difference of the superpixel region. The mean value of all values in the gray level co-occurrence matrix of the superpixel region is recorded as the texture complexity of the superpixel region. The normalized vegetation index of the target remote sensing image and the prior information remote sensing image is calculated. The difference between the normalized vegetation indexes of the target remote sensing image and the prior information remote sensing image is recorded as the vegetation feature value of the target remote sensing image. The compactness, the green difference, the texture complexity of the superpixel region, and the vegetation feature value of the target remote sensing image where the superpixel region is located are sequentially arranged to obtain the feature vector of the superpixel region. The method for monitoring the slope vegetation coverage area according to the feature vector of the superpixel region comprises the following steps:
6. The method for automatic monitoring of the area of vegetation cover of a slope according to claim 5, characterized in that, All superpixel regions obtained by using the random forest on the superpixel segmentation result of the target remote sensing image are classified to obtain the classification result of each superpixel region, and the classification result of the superpixel region includes two types of vegetation region and non-vegetation region. The cumulative sum of the actual areas of all superpixel regions with the classification result of the vegetation region in the corresponding target remote sensing image is recorded as the corresponding slope vegetation coverage area of the target remote sensing image. 7. An automatic monitoring system of the area of vegetation cover of a slope, comprising a memory, a processor and a computer program stored in said memory and running on said processor, characterized in that, The processor implements the steps of the method according to any one of claims 1-6 when executing the computer program.
Citation Information
Patent Citations
Target tree species remote sensing identification method based on air-space-ground cooperation
CN119559520A
GIS-based traffic project land expropriation and demolition digital management system and method
CN120410828A