Snow cover fraction estimation method based on image information adaptive regression

Abstract

The present application relates to the field of satellite remote sensing technology, and in particular to a snow cover remote sensing estimation method based on adaptive regression of image information, the method comprising: calculating a universal proportional snow index of snow images and snow-free images based on snow image information and snow-free image information, and calculating a normalized difference snow index of snow images; constructing a scatter density map of snow index changes before and after snowing based on the universal proportional snow index and the normalized difference snow index of snow images, so as to obtain a high-value area density change curve and a low-value area density change curve based on the scatter density map of snow index changes before and after snowing; extracting low-value area density feature points and high-value area density feature points based on the low-value area density change curve and the high-value area density change curve, calculating the difference and average value of the low-value area density feature points and the high-value area density feature points, and calculating regression model parameters based on the difference and average value, so as to estimate the snow cover of the target area based on the regression model parameters.

Description

Technical Field

[0001] The present application relates to the field of satellite remote sensing technology, and in particular to a snow cover remote sensing estimation method based on adaptive regression of image information. Background Art

[0002] Seasonal changes in snow cover and glaciers affect regional climate, the water cycle, and the surface radiation budget. Accurate and efficient acquisition of snow cover information is a crucial prerequisite for studying regional climate and hydrological changes. Satellite remote sensing technology, with its advantages of fast data acquisition, wide coverage, and limited constraints on ground conditions, has been widely used in snow cover change research. There are two main methods for representing snow cover distribution in satellite remote sensing observations: binarization of snow cover, which directly determines whether a pixel in an image is snow or non-snow; and snow cover ratio, which characterizes the proportion of a single image pixel covered by snow. Binarized snow cover products can directly estimate the snow-covered area of ​​an entire scene, but a pixel in a 30-meter resolution image is often a mixed pixel composed of snow and other non-snow features. Clearly, binary snow cover has significant errors in accurately describing snow cover distribution. Snow cover ratio describes the snow cover within a pixel and offers significant advantages for analyzing detailed changes in snow cover.

[0003] Currently, there are two main methods for extracting snow cover over large areas: empirical regression algorithms based on snow cover index and spectral unmixing methods. The empirical regression algorithm uses fixed coefficients, and the model accuracy varies across different study areas. The spectral unmixing method requires the collection of spectral information from surface endmembers. When applied to different study areas, the spectral information of the main endmembers in that area must be recollected or adjusted and optimized. Determining the optimal pixel combination for each pixel in the image also consumes significant computational resources. Machine learning methods are also currently being applied to estimate snow cover, but the network construction and training process is complex, requiring retraining and learning for different study areas.

[0004] In summary, although some progress has been made in extracting snow cover from optical remote sensing, there are still some shortcomings, including: 1) the spatial coverage and spatial coverage of existing optical satellite snow cover products cannot meet the requirements of high-precision, large-scale regional snow monitoring, and a simple and efficient algorithm suitable for medium and high-resolution imagery is needed; 2) Spectral unmixing consumes high computing resources, and the selection and construction of machine learning networks are complex; 3) the model and method cannot be dynamically adjusted according to different surface types and image information. The main estimation methods currently use fixed coefficients, end members or networks, and their universality and automation are not high. Summary of the Invention

[0005] This application provides a snow cover remote sensing estimation method based on adaptive regression of image information to solve the problem of unstable estimation accuracy of fixed parameter regression models under different snowing scenes.

[0006] The first aspect of the present application provides a remote sensing estimation method for snow cover based on adaptive regression of image information, comprising the following steps: obtaining snow image information and snow-free image information of a target area, and calculating a universal proportional snow index of snow images and snow-free images based on the snow image information and the snow-free image information, and calculating a normalized difference snow index of the snow image; constructing a scatter density map of snow index changes before and after snowing based on the universal proportional snow index and the normalized difference snow index of the snow image, so as to obtain a high-value area density change curve and a low-value area density change curve based on the scatter density map of snow index changes before and after snowing; extracting low-value area density feature points and high-value area density feature points based on the low-value area density change curve and the high-value area density change curve, calculating the difference and average value of the low-value area density feature points and the high-value area density feature points, and calculating regression model parameters based on the difference and the average value, so as to estimate the snow cover of the target area based on the regression model parameters.

[0007] Optionally, the extracting of low-value area density feature points and high-value area density feature points based on the low-value area density change curve and the high-value area density change curve includes: judging whether the low-value area density curve meets a preset flat condition; if the low-value area density curve does not meet the preset flat condition, then when there is only a single peak in the low-value area density curve, if the single peak is located in the left half of the low-value area density curve, then extracting the point from the convex to the convex in the low-value area density curve with the largest curvature from the single peak as the low-value area density feature point; if the single peak is located in the right half of the low-value area density curve, then extracting the point from the convex to the convex in the low-value area density curve with the largest curvature from the single peak to the left as the low-value area density feature point. Area density feature point; if there are two peaks in the low-value area density curve, the point corresponding to the trough between the two peaks is extracted as the low-value area density feature point; if there are more than two peaks in the low-value area density curve, when the highest peak is the first peak or the last peak, the trough closest to the highest peak is determined as the low-value area density feature point, if the highest peak and the second highest peak do not meet the proximity condition, the average value of the low-value area density feature point corresponding to the highest peak and the low-value area density feature point corresponding to the second highest peak is taken as the low-value area density feature point, if the highest peak and the second highest peak meet the proximity condition, the point corresponding to the trough between the highest peak and the second highest peak is taken as the low-value area density feature point.

[0008] Optionally, after determining whether the low-value area density curve meets the preset smooth condition, it includes: if the low-value area density curve meets the preset smooth condition, calculating the average value of the universal proportional snow index of the snow-free image, and adding the average value to the first preset threshold to obtain the low-value area density feature point.

[0009] Optionally, the method of extracting low-value area density feature points and high-value area density feature points based on the low-value area density change curve and the high-value area density change curve also includes: when there is only a single peak in the high-value area density change curve, judging whether the universal proportional snow index of the highest peak of the high-value area density curve is less than or equal to a second preset threshold; if the universal proportional snow index of the highest peak of the high-value area density curve is less than or equal to the second preset threshold, then when the highest peak of the high-value area density curve is located in the left half of the curve, taking the first preset proportion of the current highest density value as the high-value area density feature point; if the highest peak is located in For the right half of the curve, the second preset proportion of the current highest density value is taken as the density characteristic point of the high-value area; if the universal proportional snow index of the highest peak of the high-value area density curve is greater than the second preset threshold and less than or equal to the third preset threshold, the universal proportional snow index corresponding to the density point at the first preset position to the right of the highest peak is determined as the density characteristic point of the high-value area; if the universal proportional snow index of the highest peak of the high-value area density curve is greater than the third preset threshold, the universal proportional snow index corresponding to the density point at the second preset position to the left of the highest peak is determined as the density characteristic point of the high-value area.

[0010] Optionally, the extracting of low-value area density feature points and high-value area density feature points based on the low-value area density change curve and the high-value area density change curve also includes: when there are multiple peaks in the high-value area density change curve, if the difference between the universal proportional snow index of the second highest peak of the high-value area density change curve and the universal proportional snow index between the highest peak of the high-value area density change curve is greater than a fourth preset threshold, then taking the average value of the universal proportional snow index of the highest peak and the universal proportional snow index between the second highest peak as the high-value area density feature point.

[0011] Optionally, the method of calculating the universal proportional snow coverage index of the snow image and the snow-free image based on the snow image information and the snow-free image information, and after calculating the normalized difference snow coverage index of the snow image, includes: if the growth value of the universal proportional snow coverage index of the snow image relative to the snow-free image is less than a fifth preset threshold, or the universal proportional snow coverage index of the snow image is less than a sixth preset threshold, then determining that the current snow coverage rate is 0.

[0012] Optionally, estimating the snow cover ratio of the snow image based on the regression model parameters includes: calculating the snow cover ratio of the snow image using a preset snow cover ratio calculation formula, wherein the preset snow cover ratio calculation formula is:

[0013]

[0014] Where FSC is the snow cover fraction, tanh() is the hyperbolic tangent function, URSI is the value of the universal proportional snow index (URSI) for snowy images, and a and b are regression model parameters.

[0015] Optionally, calculating the regression model parameters according to the difference and average value of the low-value area density feature points and the high-value area density feature points includes: calculating the regression model parameters using a preset regression model parameter calculation formula, wherein the preset regression model parameter calculation formula is:

[0016] a=-12.294×URSI diff +11.323

[0017] b=-a×(0.9895×URSI mean )

[0018] Among them, a and b are regression model parameters, URSI diff The URSI is the difference between the density feature points in the low-value area and the density feature points in the high-value area. mean It is the average value of the density feature points in the low-value area and the density feature points in the high-value area.

[0019] Optionally, the calculating of the universal proportional snow cover index of the snow image and the snow-free image based on the snow image information and the snow-free image information, and the calculating of the normalized difference snow cover index of the snow image, includes: calculating the universal proportional snow cover index using a first calculation formula, and calculating the normalized difference snow cover index using a second calculation formula, wherein the first calculation formula is:

[0020]

[0021] The second calculation formula is:

[0022]

[0023] Among them, R Green is the green band reflectivity, R NIR is the reflectivity in the near-infrared band, R SWIR is the shortwave infrared reflectivity.

[0024] In the above embodiment, based on the information of snowy and snow-free images, a universal proportional snow cover index is calculated for the snowy and snow-free images, and a normalized difference snow cover index is calculated for the snowy images. A scatter density map of the change in snow cover index before and after snowfall is constructed based on the universal proportional snow cover index and the normalized difference snow cover index of the snowy images. A high-value area density change curve and a low-value area density change curve are obtained based on the scatter density map of the change in snow cover index before and after snowfall. Based on the low-value area density change curve and the high-value area density change curve, low-value area density feature points and high-value area density feature points are extracted. The difference and average of the low-value area density feature points and the high-value area density feature points are calculated. The regression model parameters are calculated based on the difference and average, and the snow cover ratio of the target area is estimated based on the regression model parameters. Thus, the problem of unstable estimation accuracy of the fixed parameter regression model under different snowfall scenarios is solved, the dependence of the regression model on training data is reduced, the applicability and accuracy of snow cover ratio estimation can be improved, and dynamic monitoring of snow cover ratio in different research areas can be achieved.

[0025] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0027] Figure 1 This is a flowchart of a snow cover remote sensing estimation method based on image information adaptive regression according to an embodiment of the present application;

[0028] Figure 2 This is a flowchart of an implementation of snow cover extraction according to one embodiment of the present application;

[0029] Figure 3 A schematic diagram of a low-value area scatter plot and a high-value area scatter plot and a corresponding low-value area density curve and a high-value area density curve according to one embodiment of the present application;

[0030] Figure 4 A schematic diagram of an abnormal point and a flat curve according to an embodiment of the present application;

[0031] Figure 5 URSI is a low-value area feature point according to an embodiment of the present application. L Extracted flow chart;

[0032] Figure 6 Schematic diagram of a low-value region density curve according to one embodiment of the present application;

[0033] Figure 7URSI is a high-value area feature point according to an embodiment of the present application. H Extracted flow chart;

[0034] Figure 8 Schematic diagram of a high-value area density curve according to one embodiment of the present application. DETAILED DESCRIPTION

[0035] The following describes in detail embodiments of the present application, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0036] The following describes a method for remote sensing estimation of snow cover based on adaptive regression of image information according to an embodiment of the present application with reference to the accompanying drawings. In response to the problem of unstable estimation accuracy of the fixed parameter regression model mentioned in the background art under different snowy scenes, the present application provides a method for remote sensing estimation of snow cover based on adaptive regression of image information. In this method, based on snow image information and snow-free image information, a universal proportional snow cover index for snowy images and snow-free images is calculated, and a normalized difference snow cover index for snowy images is calculated. A scatter density map of the change in snow cover index before and after snowing is constructed based on the universal proportional snow cover index and the normalized difference snow cover index for snowy images. A high-value area density change curve and a low-value area density change curve are obtained based on the scatter density map of the change in snow cover index before and after snowing. Based on the low-value area density change curve and the high-value area density change curve, low-value area density feature points and high-value area density feature points are extracted. The difference and average of the low-value area density feature points and the high-value area density feature points are calculated. The regression model parameters are calculated based on the difference and average, so as to estimate the snow cover of the target area based on the regression model parameters. This solves the problem of unstable estimation accuracy of the fixed parameter regression model under different snowing scenarios, reduces the regression model's dependence on training data, improves the applicability and accuracy of snow cover estimation, and enables dynamic monitoring of snow cover in different research areas.

[0037] Since the empirical regression method of the snow cover index is simple, efficient, and can be produced on a large scale, the empirical regression method with fixed parameters has inconsistent accuracy in different scenarios and is not universal. Therefore, this application designs a snow cover estimation method based on a nonlinear regression method, which adaptively adjusts the regression model parameters in different scenarios according to the image change characteristics before and after snowfall, and can realize dynamic monitoring of snow cover in different study areas.

[0038] Specifically, Figure 1 A flowchart of a snow cover remote sensing estimation method based on adaptive regression of image information provided in an embodiment of the present application.

[0039] like Figure 1 As shown in FIG, the snow cover remote sensing estimation method based on image information adaptive regression includes the following steps:

[0040] In step S101, snow image information and snow-free image information of the target area are obtained, and based on the snow image information and snow-free image information, the universal proportional snow index of the snow image and the snow-free image is calculated, and the normalized difference snow index of the snow image is calculated.

[0041] Optionally, in some embodiments, based on the snow image information and the snow-free image information, a universal proportional snow cover index for the snow image and the snow-free image is calculated, and a normalized difference snow cover index for the snow image is calculated, including: calculating the universal proportional snow cover index using a first calculation formula, and calculating the normalized difference snow cover index using a second calculation formula, wherein the first calculation formula is:

[0042]

[0043] The second calculation formula is:

[0044]

[0045] Among them, R Green is the green band reflectivity, R NIR is the reflectivity in the near-infrared band, R SWIR is the shortwave infrared reflectivity.

[0046] Optionally, in some embodiments, based on the snow image information and the snow-free image information, a universal proportional snow coverage index of the snow image and the snow-free image is calculated, and after calculating the normalized difference snow coverage index of the snow image, if the growth value of the universal proportional snow coverage index of the snow image relative to the snow-free image is less than the fifth preset threshold, or the universal proportional snow coverage index of the snow image is less than the sixth preset threshold, then it is determined that the current snow coverage rate is 0.

[0047] This application uses the Landsat-8OLI surface reflectance image with strip number 193027 on April 21, 2019 as an example.

[0048] Obtain Landsat-8OLI surface reflectance data (including snow image information and snow-free clear sky image information), NASADEM data, and 2020 Globeland30 land cover type data in the target area, and perform preprocessing such as splicing, reprojection, resampling, and cropping on the data. Among them, the Landsat-8OLI surface reflectance data includes the snow image with strip number 193027 on April 21, 2019, and the snow-free clear sky image with strip number 193027 on November 17, 2020 at the same location. Use the cloud mask method Fmask4.0 to mask the cloud and cloud shadow range in the snow surface reflectance data. Mask the water body according to the 2020 Globeland30 land cover type data, calculate the mountain shadow according to the solar azimuth and solar altitude angle provided by NASADEM and Landsat images, and obtain the completely invisible mountain shadow range. The masked data is considered to be valid data and can be used for subsequent calculations and processing, such as Figure 2 shown.

[0049] The first and second calculation formulas are further used to calculate the universal proportional snow index (URSI) of the snow image and the corresponding snow-free clear sky image, as well as the normalized difference snow index (NDSI) of the snow image. For each valid pixel in the snow image, if the increase in its URSI value relative to the corresponding pixel in the snow-free image is greater than or equal to 0.03 or the URSI value of the pixel itself is greater than or equal to 0.5, it is considered to be a potential snow pixel. Pixels judged to be potential snow pixels will participate in the subsequent snow cover estimation; if the increase in the universal proportional snow index of the snow image relative to the snow-free image is less than 0.03 (i.e., the fifth preset threshold), or the universal proportional snow index of the snow image is less than 0.5 (i.e., the sixth preset threshold), the current snow cover is determined to be 0.

[0050] In step S102, a scatter density map of the change of the snow index before and after snowing is constructed according to the universal proportional snow index and the normalized difference snow index of the snow image, so as to obtain a high-value area density change curve and a low-value area density change curve based on the scatter density map of the change of the snow index before and after snowing.

[0051] Using the obtained URSI and NDSI of the snow image and the snow-free image, a scatter plot of the low-value area is drawn based on the URSI of the snow image (called SnowedURSI) and the URSI of the snow-free image (called Snow-free URSI). Figure 3 (1)). Based on the snow image URSI and the snow image NDSI (called Snowed NDSI), a scatter plot of high value areas is drawn (as shown in Figure 3 (3)). Calculate the Gaussian kernel density corresponding to the two scatter plots and draw the density curve of the low-value area and the density curve of the high-value area (as shown in Figure 3(2) and 3(4)).

[0052] Specifically, pixels with URSI values ​​between 0 and 0.6 in the snowy image were identified as low-value areas. A scatter plot of the low-value areas was plotted, with the URSI values ​​of these pixels in the snowy image (Snowed URSI) as the horizontal axis and the URSI values ​​of the pixels in the snow-free image (Snow-free URSI) as the vertical axis. The number of pixels in the scatter plot was then uniformly sampled to 20,000, and the Gaussian kernel density corresponding to the low-value area scatter plot was calculated. Starting from the minimum value of the Snowed URSI, the maximum value of the Gaussian kernel density was extracted at intervals of 0.005, and a density curve was plotted. This curve is called the low-value area density curve.

[0053] Pixels with URSI values ​​between 0.6 and 1.2 in the snowy image were identified as high-value areas. A scatter plot of these high-value areas was plotted, with the URSI value (Snowed URSI) of these pixels in the snowy image as the horizontal axis and the NDSI value (Snowed NDSI) of these pixels in the snowy image as the vertical axis. The number of pixels in the scatter plot was then uniformly sampled to 20,000, and the Gaussian kernel density corresponding to the high-value area scatter plot was calculated. Starting from the minimum Snowed URSI value, the maximum value of the Gaussian kernel density was extracted at intervals of 0.005, and a density curve was plotted. This curve is called the high-value area density curve.

[0054] In step S103, the low-value area density feature points and the high-value area density feature points are extracted based on the low-value area density change curve and the high-value area density change curve, the difference and average value of the low-value area density feature points and the high-value area density feature points are calculated, and the regression model parameters are calculated according to the difference and the average value to estimate the snow cover rate of the target area based on the regression model parameters.

[0055] Optionally, in some embodiments, based on the low-value area density change curve and the high-value area density change curve, the low-value area density feature points and the high-value area density feature points are extracted, including: judging whether the low-value area density curve meets the preset flat condition; if the low-value area density curve does not meet the preset flat condition, then when there is only a single peak in the low-value area density curve, if the single peak is located in the left half of the low-value area density curve, the point from convex to convex in the low-value area density curve with the largest curvature is extracted from the single peak as the low-value area density feature point; if the single peak is located in the right half of the low-value area density curve, the point from convex to convex in the low-value area density curve with the largest curvature is extracted from the single peak to the left as the low-value area density feature point. density feature points; if there are double peaks in the low-value area density curve, the points corresponding to the troughs between the double peaks are extracted as the low-value area density feature points; if there are more than two peaks in the low-value area density curve, when the highest peak is the first peak or the last peak, the trough closest to the highest peak is determined as the low-value area density feature point; if the highest peak and the second highest peak do not meet the proximity condition, the average value of the low-value area density feature point corresponding to the highest peak and the low-value area density feature point corresponding to the second highest peak is taken as the low-value area density feature point; if the highest peak and the second highest peak meet the proximity condition, the point corresponding to the trough between the highest peak and the second highest peak is taken as the low-value area density feature point.

[0056] Optionally, in some embodiments, after determining whether the low-value area density curve meets the preset smooth condition, it includes: if the low-value area density curve meets the preset smooth condition, calculating the average value of the universal proportional snow index of the snow-free image, and adding the average value to the first preset threshold to obtain the low-value area density feature point.

[0057] After filtering the abnormal points in the low-value area density curve and the high-value area density curve, as well as the peaks and troughs of the low-value area density flat curve, the low-value area density feature points URSI are extracted. L .

[0058] The abnormal point filtering in the density curve is to delete the abnormal points in the two density curves (such as the sudden drop point in the rising curve). Figure 4 (1) As shown. For the curve with Snowed URSI values ​​between 20% and 80% in the low-value area scatter plot, the difference between the maximum density Dmax and the minimum density Dmin is calculated, and the ratio of the difference to the minimum density Dmin is calculated. If the ratio is less than 0.2, it is considered that the entire curve meets the preset flat condition, that is, the curve is a flat curve (such as Figure 4 (2)), and filter out its peaks and troughs.

[0059] Low-value area density feature point URSI L Extraction process as Figure 5 The specific implementation process is as follows:

[0060] (1) The low-value area density curve has only one steep and obvious single peak: If the single peak is located in the left half of the curve (e.g. Figure 6 (1)), then search from the peak to the right and extract the first inflection point with obvious change, that is, the point where the curve changes from convex to concave, and the point with the largest curvature is the low-value area density feature point URSI L ; If the single peak is located in the right half of the curve (such as Figure 6 (2)), then search from the single peak to the left and extract the first inflection point with obvious change as the low-value area density feature point URSI L .

[0061] (2) The density curve of the low-value area has two steep peaks (e.g. Figure 6 (3)), the trough between the two peaks is extracted as the low-value area density feature point URSI L .

[0062] (3) When the density curve of the low-value area has more than two steep peaks (e.g. Figure 6 (4) If the highest peak is located at the first peak or the last peak, find the trough closest to it as the low-value area density feature point URSI L If the highest peak is in the middle, the average value of URSI at the highest peak and the second highest peak is taken as the URSI of the low-value area density feature point. L .

[0063] (4) For all cases except the above three, the URSI average value of the entire snow-free image is calculated, and the average value plus 0.15 (i.e., the first preset threshold) is used as the URSI of the low-value area density feature point. L .

[0064] Optionally, in some embodiments, the low-value area density feature points and the high-value area density feature points are extracted based on the low-value area density change curve and the high-value area density change curve, and further include: when there is only a single peak in the high-value area density change curve, judging whether the universal proportional snow index of the highest peak of the high-value area density curve is less than or equal to a second preset threshold value; if the universal proportional snow index of the highest peak of the high-value area density curve is less than or equal to the second preset threshold value, then when the highest peak of the high-value area density curve is located in the left half of the curve, taking the first preset proportion of the current highest density value as the high-value area density feature point, if the highest peak of the high-value area density curve is located in the left half of the curve, taking the first preset proportion of the current highest density value as the high-value area density feature point, If the peak is located in the right half of the curve, the second preset proportion of the current highest density value is taken as the density characteristic point of the high-value area; if the universal proportional snow index of the highest peak of the high-value area density curve is greater than the second preset threshold and less than or equal to the third preset threshold, the universal proportional snow index corresponding to the density point at the first preset position to the right of the highest peak is determined as the density characteristic point of the high-value area; if the universal proportional snow index of the highest peak of the high-value area density curve is greater than the third preset threshold, the universal proportional snow index corresponding to the density point at the second preset position to the left of the highest peak is determined as the density characteristic point of the high-value area.

[0065] Optionally, in some embodiments, the low-value area density feature points and the high-value area density feature points are extracted based on the low-value area density change curve and the high-value area density change curve, and also include: when there are multiple peaks in the high-value area density change curve, if the difference between the universal proportional snow index of the second highest peak of the high-value area density change curve and the universal proportional snow index between the highest peak of the high-value area density change curve is greater than a fourth preset threshold, then the average value of the universal proportional snow index of the highest peak and the universal proportional snow index between the second highest peak is taken as the high-value area density feature point.

[0066] High-value area density feature point URSI H Extraction process as Figure 7 The specific implementation process is as follows:

[0067] (1) When the URSI value of the highest peak of the high-value area density curve is less than or equal to 0.82 (i.e., the second preset threshold), if the highest peak is located in the left half of the curve (e.g., Figure 8 (1), the maximum URSI value at the first preset ratio (i.e. 60%) of the highest density is taken as the URSI of the high-value area density feature point H ; If the highest peak is in the right half of the curve (such as Figure 8

[0068] (2)), the maximum URSI value at the second preset ratio (i.e. 90%) of the highest density is taken as the URSI of the high-value area density feature point H .

[0069] (2) The URSI value of the highest peak of the high-value area density curve is greater than 0.82 and less than or equal to 1.02 (i.e., the third preset threshold) (e.g. Figure 8 (3)), then find the URSI value corresponding to the density point at the first preset position to the right of the highest peak as the high-value area density feature point URSI H In the embodiment of the present application, the density point at the first preset position is the fifth density point to the right of the highest peak.

[0070] (3) When the URSI value of the highest peak of the high-value area density curve is greater than 1.02 (e.g. Figure 8 (4)), then find the URSI value corresponding to the density point at the second preset position to the left of the highest peak as the high-value area density feature point URSI H In the embodiment of the present application, the density point at the second preset position is the tenth density point to the left of the highest peak.

[0071] (4) In the case where the high-value area density curve has more than one steep and obvious peak, if the URSI distance between the highest peak and the second highest peak is greater than the fourth preset threshold (the fourth preset threshold is 0.2 in the embodiment of the present application), and the universal proportional snow index of the second highest peak is higher than half of the universal proportional snow index of the highest peak (that is, the difference between the universal proportional snow index of the second highest peak and the universal proportional snow index between the highest peak is greater than the fourth preset threshold), then the average value of the universal proportional snow index of the highest peak and the universal proportional snow index between the second highest peak is taken as the URSI of the high-value area density feature point. H .

[0072] Furthermore, the difference URSI between the low-value area density feature point and the high-value area density feature point is calculated. diff and the mean URSI mean , the calculation formula of this process is as follows:

[0073] URSI diff =URSI H -URSI L ;

[0074] URSI mean =(URSI L +URSI H ) / 2;

[0075] Among them, URSI L It is the low-value area density feature point, URSI H It is the density feature point of high value area.

[0076] Furthermore, in some embodiments, the difference URSI between the low-value area density feature point and the high-value area density feature point is calculated. diff and the mean URSImean Calculating the regression model parameters includes: calculating the regression model parameters using a preset regression model parameter calculation formula, wherein the preset regression model parameter calculation formula is:

[0077] a=-12.294×URSI diff +11.323

[0078] b=-a×(0.9895×URSI mean )

[0079] Among them, a and b are regression model parameters, URSI diff The URSI is the difference between the density feature points in the low-value area and the density feature points in the high-value area. mean It is the average value of the density feature points in the low-value area and the density feature points in the high-value area.

[0080] Furthermore, in some embodiments, estimating the snow cover ratio of the snow image based on the regression model parameters includes: calculating the snow cover ratio of the snow image using a preset snow cover ratio calculation formula, wherein the preset snow cover ratio calculation formula is:

[0081]

[0082] Where FSC is the snow cover fraction, tanh() is the hyperbolic tangent function, URSI is the value of the universal proportional snow index (URSI) for snowy images, and a and b are regression model parameters.

[0083] Compared with the prior art, the embodiments of the present application have the following advantages:

[0084] (1) By making full use of the image spectral information and temporal information, the regression model parameters can be adaptively adjusted, which is no longer limited by the large accuracy variation of the fixed parameter regression model on different surfaces and can be applied on a global large scale;

[0085] (2) The algorithm is simple, efficient, easy to implement, and suitable for massive data production.

[0086] According to the snow cover remote sensing estimation method based on adaptive regression of image information proposed in the embodiment of the present application, a universal proportional snow index is calculated for the snow image and the snow-free image based on the snow image information and the snow-free image information, and a normalized difference snow index is calculated for the snow image. A scatter density map of the snow index change before and after snowfall is constructed based on the universal proportional snow index and the normalized difference snow index of the snow image. Based on the scatter density map of the snow index change before and after snowfall, a high-value area density change curve and a low-value area density change curve are obtained. Based on the low-value area density change curve and the high-value area density change curve, low-value area density feature points and high-value area density feature points are extracted. The difference and average of the low-value area density feature points and the high-value area density feature points are calculated. The regression model parameters are calculated based on the difference and average, and the snow cover of the target area is estimated based on the regression model parameters. Thus, the problem of unstable estimation accuracy of the fixed parameter regression model under different snow scenes is solved, the dependence of the regression model on training data is reduced, the applicability and accuracy of snow cover estimation can be improved, and dynamic monitoring of snow cover in different research areas can be achieved.

[0087] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0088] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Thus, a feature specified as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of this application, "N" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0089] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application belong.

Claims

1. A snow cover remote sensing estimation method based on image information adaptive regression, characterized in that: The following steps are involved: Acquire snow image information and snow-free image information of a target area, and calculate a universal proportional snow cover index of the snow image and the snow-free image based on the snow image information and the snow-free image information, and calculate a normalized difference snow cover index of the snow image; Constructing a scatter density map of the change of the snow index before and after snowfall according to the universal proportional snow index and the normalized difference snow index of the snow image, so as to obtain a high-value area density change curve and a low-value area density change curve based on the scatter density map of the change of the snow index before and after snowfall; Extracting low-value area density feature points and high-value area density feature points based on the low-value area density change curve and the high-value area density change curve, calculating the difference and average value of the low-value area density feature points and the high-value area density feature points, and calculating regression model parameters according to the difference and the average value, so as to estimate the snow cover rate of the target area based on the regression model parameters; The calculating of the universal proportional snow cover index of the snow image and the snow-free image based on the snow image information and the snow-free image information, and calculating the normalized difference snow cover index of the snow image, comprises: The general proportional snow index is calculated using a first calculation formula, and the normalized difference snow index is calculated using a second calculation formula, wherein the first calculation formula is: ； The second calculation formula is: ； Among them, URSI is the Universal Ratio Snow Cover Index, is the green band reflectivity, is the reflectivity in the near-infrared band, is the shortwave infrared reflectivity, is the normalized difference snow cover index.

2. The method according to claim 1, characterized in that The extracting low-value area density feature points and high-value area density feature points based on the low-value area density change curve and the high-value area density change curve includes: Determine whether the low-value area density curve meets a preset flat condition; If the low-value area density curve does not meet the preset flat condition, then when there is only a single peak in the low-value area density curve, if the single peak is located in the left half of the low-value area density curve, the point of the low-value area density curve from convex to convex with the largest curvature is extracted from the single peak as the low-value area density feature point; if the single peak is located in the right half of the low-value area density curve, the point of the low-value area density curve from convex to convex to the left of the single peak is extracted as the low-value area density feature point; If the low-value area density curve has double peaks, extract the points corresponding to the troughs between the double peaks as the low-value area density feature points; If there are more than two peaks in the low-value area density curve, when the highest peak is the first peak or the last peak, the trough closest to the highest peak is determined as the low-value area density feature point; if the highest peak and the second highest peak do not meet the proximity condition, the average value of the low-value area density feature point corresponding to the highest peak and the low-value area density feature point corresponding to the second highest peak is taken as the low-value area density feature point; if the highest peak and the second highest peak meet the proximity condition, the point corresponding to the trough between the highest peak and the second highest peak is taken as the low-value area density feature point.

3. The method according to claim 2, characterized in that After determining whether the low-value area density curve meets the preset flat condition, the method includes: If the low-value area density curve meets the preset flat condition, the average value of the universal proportional snow index of the snow-free image is calculated, and the average value is added to a first preset threshold to obtain the low-value area density feature point.

4. The method according to claim 1, wherein The extracting of low-value area density feature points and high-value area density feature points based on the low-value area density change curve and the high-value area density change curve also includes: When the high-value area density change curve has only a single peak, determining whether the universal proportional snow index of the highest peak of the high-value area density curve is less than or equal to a second preset threshold; If the universal proportional snow index of the highest peak of the high-value area density curve is less than or equal to a second preset threshold, then when the highest peak of the high-value area density curve is located in the left half of the curve, a first preset proportion of the current highest density value is taken as the high-value area density feature point; if the highest peak is located in the right half of the curve, a second preset proportion of the current highest density value is taken as the high-value area density feature point; If the universal proportional snow index of the highest peak of the high-value area density curve is greater than the second preset threshold and less than or equal to the third preset threshold, determine the universal proportional snow index corresponding to the density point at the first preset position to the right of the highest peak as the high-value area density feature point; If the universal proportional snow index of the highest peak of the high-value area density curve is greater than the third preset threshold, the universal proportional snow index corresponding to the density point at the second preset position to the left of the highest peak is determined as the high-value area density feature point.

5. The method according to claim 4, characterized in that The extracting of low-value area density feature points and high-value area density feature points based on the low-value area density change curve and the high-value area density change curve also includes: When there are multiple peaks in the high-value area density change curve, if the difference between the universal proportional snow index of the second highest peak of the high-value area density change curve and the universal proportional snow index between the highest peak of the high-value area density change curve is greater than the fourth preset threshold, the average value of the universal proportional snow index of the highest peak and the universal proportional snow index between the second highest peak is taken as the high-value area density characteristic point.

6. The method according to claim 1, characterized in that After calculating the universal proportional snow cover index of the snow image and the snow-free image based on the snow image information and the snow-free image information, and calculating the normalized difference snow cover index of the snow image, the method includes: If the growth value of the general proportional snow index of the snowy image relative to the snowless image is less than the fifth preset threshold, or the general proportional snow index of the snowy image is less than the sixth preset threshold, it is determined that the current snow cover rate is 0.

7. The method according to claim 1, characterized in that The estimating the snow cover rate of the snow image based on the regression model parameters includes: The snow coverage ratio of the snow image is calculated using a preset snow coverage ratio calculation formula, wherein the preset snow coverage ratio calculation formula is: ； in, is the snow cover ratio, tanh() is the hyperbolic tangent function, URSI is the value of the universal proportional snow index URSI for snow images, and are the regression model parameters.

8. The method according to claim 1, characterized in that The calculating of the regression model parameters according to the difference and average value of the low-value area density feature points and the high-value area density feature points includes: The regression model parameters are calculated using a preset regression model parameter calculation formula, wherein the preset regression model parameter calculation formula is: in, and are regression model parameters, is the difference between the density feature points in the low-value area and the density feature points in the high-value area, It is the average value of the density feature points in the low-value area and the density feature points in the high-value area.

Images