Forest disturbance monitoring method, system, device and medium based on image difference method

Abstract

The application discloses a forest disturbance monitoring method, system, device and medium based on an image difference method, and relates to the field of forest monitoring.The method comprises the following steps: performing spectral index feature extraction on historical remote sensing images of a region to be monitored to obtain a remote sensing feature set; dividing the remote sensing feature set of the historical remote sensing images into two categories, i.e., disturbed forest and undisturbed forest, counting the reflectivity mean values of eight bands corresponding to the two types, selecting bands with larger differences in pixel value differences from a preset number of bands, and selecting three bands from the preset number of bands as set difference bands; obtaining a spectral feature map corresponding to a current period remote sensing image of the region to be monitored and a spectral feature map corresponding to a last period remote sensing image; and determining a current disturbed forest region of the region to be monitored by using an image difference method according to the current period spectral feature map and the last period spectral feature map and according to the preset difference bands.The application realizes large-range and fine real-time monitoring of forest disturbance.

Description

Technical Field

[0001] This invention relates to the field of forest monitoring technology, and in particular to a method, system, equipment and medium for monitoring forest disturbance based on image difference method. Background Technology

[0002] Forests are the largest terrestrial ecosystems and a vital component of the Earth's biosphere, playing a crucial role in human survival and development. Forest ecosystems are not only the most productive terrestrial ecosystems but also a vast carbon sink. Dynamic changes in forest resources can significantly impact the carbon cycle and carbon storage within forest ecosystems.

[0003] Forest disturbance refers to temporary changes in environmental conditions that lead to rapid and significant alterations in the ecosystem, causing substantial impacts and typically accompanied by significant biomass loss. Forest disturbance generally includes both natural and anthropogenic disturbances. Natural disturbances are primarily caused by fire, wind, snow, floods, and earthquakes, with volcanic eruptions (including forest fires and wind erosion), snow damage, and other disturbances being the most common. Anthropogenic disturbances mainly include deforestation, logging, clearing, and overgrazing. Currently, anthropogenic disturbances have a greater impact on forests than natural disturbances. The types of forest disturbance are influenced by regional environments, and different environments exhibit different types of disturbance. In primary forests, natural disturbances predominate, while in areas close to human habitation, anthropogenic disturbances become dominant. In recent years, forest disturbances have disrupted ecosystem balance, causing varying degrees of damage to community or species structure, generally manifested in biomass loss, reduced forest cover, and soil erosion. Therefore, it is necessary to conduct remote sensing monitoring of forest disturbance. Accurate monitoring of forest disturbance can not only deepen people's understanding of the structure and function of forest ecosystems, but also provide more scientific basic information for global carbon sink estimation.

[0004] Traditional forest disturbance monitoring often relies on manual field surveys, which are time-consuming, labor-intensive, and inefficient, making them unsuitable for large-scale forest disturbance monitoring. With the development of satellite technology, remote sensing imagery has gradually become the primary means of forest disturbance monitoring. Remote sensing methods can save significant manpower, material resources, and financial resources, shorten the survey cycle, and effectively avoid the influence of human and environmental factors. The selection of remote sensing data will have a certain impact on the monitoring results for different study areas. Currently, the most widely used remote sensing data are Landsat data and MODIS data. MODIS data has a large coverage area and high temporal resolution, but its spatial resolution is low, and the accuracy of its monitoring results cannot be guaranteed. Landsat data has high spatial resolution, but its coverage is limited and easily affected by weather conditions, making it difficult to conduct large-scale forest disturbance remote sensing monitoring. Summary of the Invention

[0005] The purpose of this invention is to provide a method, system, device and medium for monitoring forest disturbance based on image difference method, which realizes large-scale and refined real-time monitoring of forest disturbance.

[0006] To achieve the above objectives, the present invention provides the following solution:

[0007] A forest disturbance monitoring method based on image interpolation includes:

[0008] Spectral index features are extracted from historical remote sensing images of the area to be monitored to obtain a spectral feature set of the historical remote sensing images; each sample in the spectral feature set includes the spectral features of each pixel in the historical remote sensing images.

[0009] The remote sensing feature set of the historical remote sensing images is divided into two categories: disturbed forest and undisturbed forest.

[0010] The average reflectance of a predetermined number of bands corresponding to disturbed forests and the average reflectance of a predetermined number of bands corresponding to undisturbed forests are calculated from the remote sensing feature set of the historical remote sensing images.

[0011] The mean reflectance of a preset number of bands corresponding to the disturbed forest is subtracted from the mean reflectance of a preset number of bands corresponding to the undisturbed forest. Based on the magnitude of the difference, three bands are selected from the preset number of bands as the set difference bands.

[0012] The spectral feature maps of the current period and the previous period of remote sensing images of the area to be monitored are obtained respectively; the remote sensing images are GF-6WFV satellite data;

[0013] Based on the spectral feature map corresponding to the current period's remote sensing image and the spectral feature map corresponding to the previous period's remote sensing image, and according to the set difference band, the current disturbed forest area of ​​the area to be monitored is determined by the image difference method.

[0014] Optionally, based on the spectral feature map corresponding to the current period's remote sensing image and the spectral feature map corresponding to the previous period's remote sensing image, and according to a set difference band, the current disturbed forest area of ​​the area to be monitored is determined using the image difference method, specifically including:

[0015] According to the formula ΔX i =X iT2 -X iT1 Determine the difference between the values ​​of each pixel;

[0016] Determine whether each pixel is a disturbed forest area based on the difference between each pixel;

[0017] Where, ΔX iX represents the difference of the i-th pixel. iT2 X represents the pixel value of the i-th pixel in the spectral feature map corresponding to the current period's remote sensing image. iT1 This represents the pixel value of the i-th pixel in the spectral feature map corresponding to the remote sensing image from the previous period.

[0018] Optionally, before extracting spectral index features from historical remote sensing images of the area to be monitored to obtain the spectral feature set of the historical remote sensing images, the method further includes:

[0019] The historical remote sensing images are sequentially subjected to radiometric calibration, atmospheric correction, orthorectification, image registration, and image stitching and cropping.

[0020] Optionally, the preset quantity is 8, and the 8 bands are blue band, green band, red band, near-infrared band, red-edge 1 band, red-edge 2 band, purple band and yellow band.

[0021] Optionally, the mean reflectance of a preset number of bands corresponding to the disturbed forest is subtracted from the mean reflectance of a preset number of bands corresponding to the undisturbed forest. Based on the magnitude of the difference, three bands are selected from the corresponding preset number of bands as the set difference bands, specifically including:

[0022] Sort a preset number of differences from largest to smallest, and select the top 5 differences;

[0023] Calculate the average of the first 5 differences;

[0024] The top three differences between the first five differences and the average value, from smallest to largest, are designated as the difference bands.

[0025] This invention also discloses a forest disturbance monitoring system based on image interpolation, comprising:

[0026] The spectral index feature extraction module is used to extract spectral index features from historical remote sensing images of the area to be monitored, and obtain the spectral feature set of the historical remote sensing images; each sample in the spectral feature set includes the spectral features of each pixel in the historical remote sensing images.

[0027] The classification module is used to divide the remote sensing feature set of the historical remote sensing images into two categories: disturbed forest and undisturbed forest.

[0028] The mean reflectance statistics module is used to count the mean reflectance of a preset number of bands corresponding to disturbed forests and the mean reflectance of a preset number of bands corresponding to undisturbed forests in the remote sensing feature set of the historical remote sensing images.

[0029] The module for determining the difference band is used to calculate the difference between the average reflectance of a preset number of bands corresponding to the disturbed forest and the average reflectance of a preset number of bands corresponding to the undisturbed forest, and select 3 bands from the corresponding preset number of bands as the set difference bands based on the magnitude of the difference.

[0030] The feature map determination module is used to obtain the spectral feature map corresponding to the current period remote sensing image of the area to be monitored and the spectral feature map corresponding to the previous period remote sensing image; the remote sensing image is GF-6WFV satellite data;

[0031] The disturbed forest area determination module is used to determine the current disturbed forest area of ​​the area to be monitored by using the image difference method based on the spectral feature map corresponding to the current period's remote sensing image and the spectral feature map corresponding to the previous period's remote sensing image, according to the set difference band.

[0032] The present invention also discloses an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the forest disturbance monitoring method based on image interpolation.

[0033] The present invention also discloses a computer-readable storage medium storing a computer program that, when executed by a processor, implements the forest disturbance monitoring method based on image interpolation.

[0034] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0035] This invention obtains remote sensing images of the area to be monitored from GF-6WFV satellite data. GF-6WFV satellite data has the characteristics of monitoring a large area and high resolution. By combining the image difference method, the current disturbed forest area of ​​the area to be monitored is determined, realizing large-scale and refined real-time monitoring of forest disturbance. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This invention provides a schematic flowchart of a forest disturbance monitoring method based on image interpolation.

[0038] Figure 2 A comparative diagram of image registration for ground feature 1 in 2020 and 2021 is provided for embodiments of the present invention;

[0039] Figure 3 A comparative diagram of image registration for feature 2 in 2020 and 2021 is provided for embodiments of the present invention;

[0040] Figure 4 A comparative diagram of image registration for ground feature 3 in 2020 and 2021 is provided for embodiments of the present invention;

[0041] Figure 5 A schematic diagram of spectral curves of different ground features in the area to be monitored is provided for embodiments of the present invention;

[0042] Figure 6 The present invention provides feature maps corresponding to spectral index features; wherein, (a) represents the normalized vegetation index feature map, (b) represents the blue band feature map, (c) represents the green band feature map, (d) represents the red band feature map, (e) represents the near-infrared band feature map, (f) represents the red-edge 1 band feature map, (g) represents the red-edge 2 band feature map, (h) represents the violet band feature map, and (i) represents the yellow band feature map;

[0043] Figure 7 The present invention provides a distribution map of pixel value differences for each variation type in the corresponding bands.

[0044] Figure 8 A schematic diagram of the grayscale frequency distribution using the red band difference method is provided for embodiments of the present invention;

[0045] Figure 9 A schematic diagram of the grayscale frequency distribution of the red-edge 1-band difference method is provided for embodiments of the present invention;

[0046] Figure 10 A schematic diagram of the grayscale frequency distribution using the yellow band difference method is provided for embodiments of the present invention;

[0047] Figure 11 A schematic diagram of grayscale value frequency distribution using the NDVI interpolation method is provided for embodiments of the present invention;

[0048] Figure 12 A schematic diagram illustrating the monitoring effect of the red band difference method is provided for embodiments of the present invention;

[0049] Figure 13 A schematic diagram illustrating the monitoring effect of the red-edge 1-band interpolation method is provided for embodiments of the present invention;

[0050] Figure 14 A schematic diagram illustrating the monitoring effect of the yellow band difference method is provided for embodiments of the present invention;

[0051] Figure 15 A schematic diagram illustrating the monitoring effect of the NDVI difference method is provided for embodiments of the present invention;

[0052] Figure 16 This invention provides a schematic diagram of a forest disturbance monitoring system based on image interpolation. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] The purpose of this invention is to provide a method, system, device and medium for monitoring forest disturbance based on image difference method, which realizes large-scale and refined real-time monitoring of forest disturbance.

[0055] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0056] Example 1

[0057] like Figure 1 As shown, this embodiment provides a forest disturbance monitoring method based on image interpolation, which specifically includes the following steps:

[0058] Step 101: Extract spectral index features from historical remote sensing images of the area to be monitored to obtain a spectral feature set of the historical remote sensing images; each sample in the spectral feature set includes the spectral features of each pixel in the historical remote sensing images.

[0059] The historical remote sensing images are GF-6WFV satellite data of the area to be monitored.

[0060] Step 101 specifically includes:

[0061] The historical remote sensing images are sequentially subjected to radiometric calibration, atmospheric correction, orthorectification, image registration, and image stitching and cropping to obtain preprocessed historical remote sensing images.

[0062] Preprocessing of remote sensing images is a prerequisite for their effective application. Remote sensing imaging is affected by various factors such as atmosphere, topography, and acquisition angle, resulting in distortion, noise, and blurring between the acquired images and the actual surface images. These factors significantly impact the quality and accuracy of information extraction. Given that the vegetation index constructed in this invention has a certain influence on forest disturbance monitoring results, the processing of GF-6WFV data in this invention mainly includes radiometric calibration, atmospheric correction, orthorectification, image registration, image stitching, and cropping.

[0063] Step 1011: Radiation calibration.

[0064] Factors such as solar altitude angle, meteorological conditions, and shooting time at different times can affect remote sensing imaging. Radiometric correction is typically required before monitoring forest disturbance. Radiometric calibration is the process of converting the DN values ​​recorded by the sensor into absolute radiance or surface reflectance. This process eliminates errors caused by the sensor itself, obtaining the true radiance values ​​at the sensor inlet. This invention uses calibration parameters from the GF-6WFV sensor obtained from the China Resources Satellite Application Center in 2020 and 2021 to radiometrically calibrate the data (historical remote sensing images). Based on the ENVI 5.3 extension tool China Satellites, the GF-6WFV.til file is loaded, automatically mosaicking the three image files virtually. The Radiometric Calibration tool in ENVI 5.3 is selected to convert the image DN values ​​into radiance values. The formula is:

[0065] L = Gain * DN + Offset;

[0066] In the formula, L is the radiance value, in W·(m2·sr-1·μm-1); DN is the satellite payload observation value; Gain and Offset are the sensor gain and offset values, respectively, in W·(m2·sr-1·μm-1).

[0067] Step 1012: Atmospheric correction.

[0068] Atmospheric correction is the process of eliminating radiation errors in the atmospheric reflection, absorption, and scattering processes, converting the apparent reflectivity of the upper atmosphere into the true reflectivity of the Earth's surface. This invention uses the FLAASH atmospheric correction module in ENVI 5.3, which was developed with the support of the US Air Force Research Laboratory using the MODTRAN radiative transfer calculation method. The input data is radiometrically calibrated data; the center latitude, longitude, and date of the image can be automatically identified through the header file, while other parameters need to be added manually. The sensor type is UNKNOWN-MSI, the sensor altitude is 645 km, the average altitude of the imaging area is obtained from the statistical mean of DEM data, the pixel size is 16 m, the atmospheric model is selected based on latitude, longitude, and image area (referencing latitude, longitude, and imaging timetables), the aerosol model is selected as Rural, and since the GF-6WFV data lacks shortwave infrared bands, the aerosol inversion method is None, with an initial visibility of 40 km.

[0069] Step 1013: Orthorectification.

[0070] Remote sensing images are subject to geometric distortion due to sensor and terrain factors during the imaging process. This distortion affects the quality of remote sensing images and their practical applications. Therefore, orthorectification is performed on remote sensing images to eliminate geometric distortion. This invention utilizes the orthorectification tool in ENVI. Using atmospherically corrected GF-6WFV imagery, 30m resolution DEM data is selected, and 0.5m resolution offset Tianditu imagery from 91 satellites is used. Highly representative control points are uniformly selected for orthorectification, with the calibration error controlled within 0.6 pixels.

[0071] Step 1014: Image registration.

[0072] Before conducting forest disturbance monitoring, it is essential to ensure sufficiently high registration accuracy between images from different periods to reduce false change information. Image registration involves correcting ground control point pixels at the same location across multiple temporal images, thereby eliminating geometric positional errors between different temporal images. This invention uses 2020 images as the reference image to register 2021 images, employing the nearest neighbor resampling method and a polynomial correction model. A total of 530 corresponding control points were selected, with the root mean square (RMS) of each control point controlled within 0.6 pixels, resulting in a total registration error (RMSE) of 0.4 pixels. Information on some control points is shown in Table 1, and the registration results are shown in [Table 1]. Figure 2 -Figure 4, Figures 2-4 The middle image (a) is a remote sensing image from 2020. Figures 2-4 (b) is a remote sensing image from 2021.

[0073] Table 1. Statistical Table of Image Registration Control Points

[0074]

[0075] Step 1015: Image stitching and cropping.

[0076] To obtain complete remote sensing images of the study area, this invention uses the Seamless Mosaic tool of ENVI 5.3 to stitch together two images from 2020 and 2021. The stitched images are then cropped using the administrative vector boundary of the study area to obtain historical remote sensing images of the area to be monitored in 2020 and 2021.

[0077] Extract spectral index features, vegetation red edge band features, texture features, and topographic features from preprocessed historical remote sensing images.

[0078] The feature map corresponding to the spectral index feature is as follows Figure 6 As shown.

[0079] Step 1016: Extraction and analysis of spectral index features.

[0080] In remote sensing image classification, spectrum is the most fundamental feature. Most of the information extraction from remote sensing images relies on spectral features derived from ground cover information, as different ground covers exhibit varying solar reflectance within their respective spectral ranges. This invention uses eight original bands from GF-6WFV data—B1-B8 bands (blue, green, red, near-infrared, red-edge 1, red-edge 2, violet, and yellow bands)—as the spectral features for forest information extraction. Spectral curves for different ground covers were plotted based on selected samples of vegetation, soil, water bodies, and built-up land, as shown below. Figure 5 As shown.

[0081] from Figure 5 It can be seen that vegetation and non-vegetation samples differ significantly across different wavebands. Therefore, relevant vegetation indices can be constructed using different wavebands to extract forests. This invention introduces the commonly used Normalized Difference Vegetation Index (NDVI), described and calculated as follows:

[0082] The Normalized Difference Vegetation Index (NDVI) is an effective indicator of vegetation growth variation and spatial distribution, and is closely related to vegetation cover. NDVI is calculated as the ratio of the difference between the near-infrared band and the sum of the near-infrared and red bands, with values ​​generally ranging from -1 to 1. Higher vegetation cover results in a higher NDVI value. The calculation formula is as follows:

[0083] NDVI = (NIR - R) / (NIR + R);

[0084] In the formula, NIR represents the reflectance value in the near-infrared band, and R represents the reflectance value in the red band.

[0085] Step 102: Divide the remote sensing feature set of the historical remote sensing images into two categories: disturbed forest and undisturbed forest.

[0086] Meanwhile, the Normalized Difference Vegetation Index (NDVI) is an effective indicator for monitoring vegetation growth.

[0087] Step 103: Calculate the average reflectance of a preset number of bands corresponding to disturbed forests and the average reflectance of a preset number of bands corresponding to undisturbed forests in the remote sensing feature set of the historical remote sensing images.

[0088] Step 104: Calculate the difference between the mean reflectance of a preset number of bands corresponding to the disturbed forest and the mean reflectance of a preset number of bands corresponding to the undisturbed forest, and select 3 bands from the preset number of bands as the set difference bands based on the magnitude of the difference.

[0089] Step 104 involves selecting bands with relatively large differences in average reflectance among 8 (preset number) bands, where the differences between these differences are small, as the set difference bands. Specifically, this includes:

[0090] Sort a preset number of differences from largest to smallest and select the top 5 differences.

[0091] Calculate the average of the first 5 differences.

[0092] The top three differences between the first five differences and the average value, from smallest to largest, are designated as the difference bands.

[0093] Step 105: Obtain the spectral feature map corresponding to the current period remote sensing image and the spectral feature map corresponding to the previous period remote sensing image of the area to be monitored; the remote sensing image is GF-6WFV satellite data.

[0094] Step 106: Based on the spectral feature map corresponding to the current period's remote sensing image and the spectral feature map corresponding to the previous period's remote sensing image, and according to the set difference band, the current disturbed forest area of ​​the area to be monitored is determined by the image difference method.

[0095] Sample points were randomly selected from disturbed and undisturbed forests within the monitoring area. The mean reflectance values ​​for the eight bands corresponding to these two types were calculated. The results are as follows: Figure 7 As shown, from Figure 7 As can be seen, the red band, red-edge 1 band, and yellow band show the largest differences in pixel values ​​between disturbed forest and undisturbed samples. Therefore, it is hoped that the ability of these three bands to extract forest disturbances can be explored.

[0096] The Normalized Difference Vegetation Index (NDVI) is widely used in ecological and vegetation remote sensing and is an effective indicator for monitoring vegetation growth and changes in the ecological environment. Therefore, it is also worth considering using NDVI difference to extract change information.

[0097] Based on the above analysis results, this invention employs four methods—red band difference, red-edge 1 band difference, yellow band difference, and NDVI difference—to extract forest disturbance information within the study area, and compares the accuracy of the results from the four methods. The difference ΔX between each pixel value is also considered. i The calculation formula is as follows:

[0098] According to the formula ΔX i =X iT2 -X iT1 Determine the difference between the values ​​of each pixel;

[0099] Determine whether each pixel is a disturbed forest area based on the difference between each pixel;

[0100] Where, ΔX i X represents the difference of the i-th pixel.iT2 X represents the pixel value of the i-th pixel in the spectral feature map corresponding to the current period's remote sensing image. iT1 This represents the pixel value of the i-th pixel in the spectral feature map corresponding to the remote sensing image from the previous period.

[0101] For the NDVI interpolation method, when ΔX i When ΔX < 0, it indicates that the pixel value of the later remote sensing image is less than that of the previous year, representing forest disturbance. For the red band, red-edge 1 band, and yellow band difference method, when ΔX i When the value is greater than 0, it indicates forest disturbance.

[0102] The threshold determination in the image difference method specifically includes: During the acquisition of remote sensing images, the solar illumination angle, atmospheric conditions, and orbital changes cause radiometric and geometric differences between different images. Although the images undergo radiometric calibration and atmospheric correction, these errors cannot be completely eliminated. In the image difference method, threshold determination is particularly important; an unreasonable threshold setting will affect the accuracy of forest disturbance monitoring. Traditional threshold determination methods require a large number of samples, resulting in a large workload and low efficiency. This invention attempts to use the distribution function method from statistics, treating the image difference result as a random variable. Numerical statistics are performed on the entire difference image, and the threshold for the type of change is determined based on the cumulative frequency. Specifically, this includes: statistical analysis of pixel values ​​in the entire difference image. Since the number of pixels in the changed areas is much smaller than the number of pixels in the unchanged areas, most pixels in the difference image have small gray values ​​concentrated in the peak area of ​​the distribution map, indicating no change. The two ends of the histogram mainly represent pixels that have changed. In probability theory, different confidence levels are often used as the critical points for significance testing, with commonly used confidence levels being 90%, 95%, or 99%. This invention conducts experiments at cumulative frequencies of 90%, 95%, and 99%, and ultimately selects the cumulative frequency as the pixel grayscale value corresponding to 99% as the threshold for determining a certain change in the image; that is, the probability of determining that the pixel value of the image difference represents a certain type of change is 1%.

[0103] The method of this invention is simple and fast, and is not easily affected by outliers. Its principle is as follows: Let X be a random variable, x be any pixel gray value, and P represent probability. Then the cumulative probability distribution function is expressed as: F(X)=P{X≤x}.

[0104] F(X) = P{X≤x} is used to represent the probability of a random value falling within any interval. X is the pixel value, and x is the threshold value at which the value changes. In probability theory, the values ​​corresponding to different confidence levels are often used as the critical points for significance testing. Commonly used confidence levels are 90%, 95%, or 99%.

[0105] The accuracy of this invention is evaluated as follows:

[0106] Currently, the most widely used accuracy evaluation method is the confusion matrix method based on image classification. A confusion matrix is ​​a comparison matrix used to represent the number of pixels in a specific class of an image classification problem compared to the number of pixels in the true reference class. In the confusion matrix, the rows of the array typically represent the predicted data obtained from image classification, and the columns represent the true reference data. In image classification, it is mainly used to compare the true values ​​and the classified values, displaying the accuracy of each class and the overall classification accuracy in the confusion matrix.

[0107] (1) Overall accuracy (OA) refers to the proportion of correctly classified pixels to the total number of pixels, which is the total number of all real reference pixels. It reflects the overall accuracy of the classification results.

[0108] (2) The Kappa coefficient is calculated by multiplying the total number of true reference pixels by the sum of the diagonals of the confusion matrix, subtracting the product of the number of true reference pixels in a particular class and the total number of classified pixels in that class, and then dividing by the square of the total number of pixels minus the sum of the product of the number of true reference pixels in a particular class and the total number of classified pixels in that class, over all classes. The Kappa coefficient comprehensively measures classification error. Its formula is expressed as:

[0109]

[0110] In the formula, N represents the total number of all real reference pixels; X kk X represents the sum of the diagonals of the confusion matrix; k+ and X +k These represent the product of the actual reference pixels in a certain class and the total number of classified pixels in that class, respectively; M represents the total number of pixels used for evaluation.

[0111] (3) Commission Error refers to the proportion of pixels that are classified as the class of interest to the user but actually belong to another class.

[0112] (4) Omission errors refer to the proportion of pixels that belong to the true surface classification but were not classified into the corresponding category by the classifier.

[0113] Threshold determination for the four interpolation methods of this invention:

[0114] In statistics, the distribution of image grayscale values ​​can be represented by a frequency distribution histogram. This invention statistically analyzed the frequency distribution of grayscale values ​​in four types of difference images, and the results are as follows: Figures 8-11As shown, the distributions approximate a normal distribution. Since the number of pixels in the changed areas is much smaller than the number of pixels in the unchanged areas, most pixels are concentrated in the peak portion of the distribution map, indicating no change. The two ends of the histogram mainly represent pixels that have changed. Based on the frequency distribution histogram and cumulative frequency distribution map results, after repeated experiments, the cumulative frequency was finally selected as the gray value at 99% as the threshold for determining whether a certain change has occurred in the image. That is, the probability of determining that the pixel value of the image difference is a certain type of change is 1%. The results are shown in Table 2. ΔX i The pixel gray value of the image difference result is... Figures 8-11 The x-coordinate in the diagram.

[0115] Table 2 Criteria for Determining Forest Disturbance

[0116]

[0117] Forest disturbance monitoring results and accuracy evaluation:

[0118] In ENVI 5.3 software, image interpolation was performed on the red band, red-edge 1 band, yellow band, and NDVI index, respectively. The interpolation results obtained according to the judgment criteria are as follows: Figures 12-15 As shown in the figure, the black areas represent forest disturbance areas, and the white areas represent undisturbed areas. The figure shows that image interpolation can monitor forest disturbance, but a large number of "spurious disturbance" events occur. The spectral characteristics of vegetation are affected by the surrounding environment. In the western part of the Jianghuai watershed and the southeastern part of Chuzhou City, where forests are scarce, the monitoring results show a large number of forest disturbance events. This may be because changes in large areas of farmland caused "spurious disturbance" events. Rainfall in 2021 was greater than in 2020, and some farmland was flooded, resulting in larger image differences and thus "spurious disturbance" events.

[0119] Based on the Class II forest resource survey data, high-resolution data from Gaofen-1 and Gaofen-2 satellites, and high-resolution imagery from Google Earth, and combined with visual interpretation, a total of 730 sample points were identified for accuracy evaluation. Among these, 293 sample points were identified as having forest disturbance, and 437 as having undisturbed forest. Four monitoring indicators provided by the confusion matrix and the Kappa coefficient were used for accuracy evaluation. The accuracy evaluation results are shown in Table 3. Table 3 shows that among the four monitoring results, the NDVI difference method had the highest accuracy at 89.04% with a Kappa coefficient of 0.765. This was followed by the red band difference method at 88.49% with a Kappa coefficient of 0.754. The overall accuracy of the red-edge 1 band difference method and the yellow band difference method was similar, at 86.99% and 86.71% respectively, with Kappa coefficients of 0.716 and 0.715 respectively. However, the four interpolation methods showed high omission errors, indicating a lower effectiveness in monitoring disturbed forest areas, with some disturbed areas not being identified. Conversely, the low misclassification errors suggested better monitoring of undisturbed forest areas. This may be because undisturbed forest areas are much larger than disturbed areas, leading to better identification. Therefore, while image interpolation can quickly identify disturbed forest areas and provide a reference for forest resource management departments, it also suffers from a large number of false disturbances.

[0120] Table 3. Accuracy Evaluation of Forest Disturbance Monitoring Results Based on Image Interference Method

[0121]

[0122]

[0123] Example 2

[0124] like Figure 16 As shown, this embodiment provides a forest disturbance monitoring system based on image interpolation, including:

[0125] The spectral index feature extraction module 201 is used to extract spectral index features from historical remote sensing images of the area to be monitored, and obtain the spectral feature set of the historical remote sensing images; each sample in the spectral feature set includes the spectral features of each pixel in the historical remote sensing images.

[0126] The classification module 202 is used to divide the remote sensing feature set of the historical remote sensing image into two categories: disturbed forest and undisturbed forest.

[0127] The reflectance mean statistics module 203 is used to count the reflectance mean of a preset number of bands corresponding to disturbed forests and the reflectance mean of a preset number of bands corresponding to undisturbed forests in the remote sensing feature set of the historical remote sensing images.

[0128] The setting difference band determination module 204 is used to calculate the difference between the average reflectance of a preset number of bands corresponding to the disturbed forest and the average reflectance of a preset number of bands corresponding to the undisturbed forest, and select 3 bands from the corresponding preset number of bands as the setting difference bands based on the magnitude of the difference.

[0129] The feature map determination module 205 is used to obtain the spectral feature map corresponding to the current period remote sensing image of the area to be monitored and the spectral feature map corresponding to the previous period remote sensing image; the remote sensing image is GF-6WFV satellite data.

[0130] The disturbed forest area determination module 206 is used to determine the current disturbed forest area of ​​the area to be monitored by using the image difference method based on the spectral feature map corresponding to the current period remote sensing image and the spectral feature map corresponding to the previous period remote sensing image, according to the set difference band.

[0131] Example 3

[0132] This embodiment provides an electronic device, including a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the forest disturbance monitoring method based on image difference method described in Embodiment 1.

[0133] This embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the forest disturbance monitoring method based on image interpolation as described in Embodiment 1.

[0134] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0135] Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the device and its core ideas. Furthermore, those skilled in the art will recognize that, based on the ideas of this invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A forest disturbance monitoring method based on image interpolation, characterized in that, include: Spectral index features are extracted from historical remote sensing images of the area to be monitored to obtain a spectral feature set of the historical remote sensing images; each sample in the spectral feature set includes the spectral features of each pixel in the historical remote sensing images. The remote sensing feature set of the historical remote sensing images is divided into two categories: disturbed forests and undisturbed forests. Forest disturbance includes both natural disturbance and human disturbance. Natural disturbance includes volcanic eruptions and snow damage, including forest fires and windfalls. Human disturbance includes deforestation, logging, felling, and grazing. The average reflectance of a predetermined number of bands corresponding to disturbed forests and the average reflectance of a predetermined number of bands corresponding to undisturbed forests are calculated from the remote sensing feature set of the historical remote sensing images. The mean reflectance of a preset number of bands corresponding to disturbed forests is subtracted from the mean reflectance of a preset number of bands corresponding to undisturbed forests. Based on the magnitude of the difference, three bands are selected from the preset number of bands as the set difference bands. Specifically, this includes: sorting the preset number of differences from largest to smallest and selecting the top five differences; calculating the average of the top five differences; and selecting the top three differences between the top five differences and the average from smallest to largest as the set difference bands. The preset number is eight, and the eight bands are blue band, green band, red band, near-infrared band, red-edge 1 band, red-edge 2 band, purple band, and yellow band. The spectral feature maps of the current period and the previous period of remote sensing images of the area to be monitored are obtained respectively; the remote sensing images are GF-6 WFV satellite data. Based on the spectral feature map corresponding to the current period's remote sensing image and the spectral feature map corresponding to the previous period's remote sensing image, and according to the set difference band, the current disturbed forest area of ​​the area to be monitored is determined by the image difference method.

2. The forest disturbance monitoring method based on image interpolation according to claim 1, characterized in that, Based on the spectral feature maps of the current period's remote sensing images and the spectral feature maps of the previous period's remote sensing images, and according to the set difference bands, the current disturbed forest area of ​​the area to be monitored is determined using the image difference method, specifically including: According to the formula Determine the difference between the values ​​of each pixel; Determine whether each pixel is a disturbed forest area based on the difference between each pixel; in, This represents the difference value of the i-th pixel. This represents the pixel value of the i-th pixel in the spectral feature map corresponding to the current period's remote sensing image. This represents the pixel value of the i-th pixel in the spectral feature map corresponding to the remote sensing image from the previous period.

3. The forest disturbance monitoring method based on image interpolation according to claim 1, characterized in that, Before extracting spectral index features from historical remote sensing images of the area to be monitored to obtain the spectral feature set of the historical remote sensing images, the process also includes: The historical remote sensing images are sequentially subjected to radiometric calibration, atmospheric correction, orthorectification, image registration, and image stitching and cropping.

4. A forest disturbance monitoring system based on image interpolation, characterized in that, The forest disturbance monitoring system based on image interpolation uses the forest disturbance monitoring method based on image interpolation as described in claim 1, and the forest disturbance monitoring system based on image interpolation includes: The spectral index feature extraction module is used to extract spectral index features from historical remote sensing images of the area to be monitored, and obtain the spectral feature set of the historical remote sensing images; each sample in the spectral feature set includes the spectral features of each pixel in the historical remote sensing images. The classification module is used to divide the remote sensing feature set of the historical remote sensing images into two categories: disturbed forest and undisturbed forest. The mean reflectance statistics module is used to count the mean reflectance of a preset number of bands corresponding to disturbed forests and the mean reflectance of a preset number of bands corresponding to undisturbed forests in the remote sensing feature set of the historical remote sensing images. The module for determining the difference band is used to calculate the difference between the average reflectance of a preset number of bands corresponding to the disturbed forest and the average reflectance of a preset number of bands corresponding to the undisturbed forest, and select 3 bands from the corresponding preset number of bands as the set difference bands based on the magnitude of the difference. The feature map determination module is used to obtain the spectral feature map corresponding to the current period's remote sensing image of the area to be monitored and the spectral feature map corresponding to the previous period's remote sensing image; the remote sensing image is GF-6 WFV satellite data. The disturbed forest area determination module is used to determine the current disturbed forest area of ​​the area to be monitored by using the image difference method based on the spectral feature map corresponding to the current period's remote sensing image and the spectral feature map corresponding to the previous period's remote sensing image, according to the set difference band.

5. An electronic device, characterized in that, The device includes a memory and a processor, the memory being used to store a computer program, and the processor running the computer program to enable the electronic device to perform the forest disturbance monitoring method based on image interpolation according to any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the forest disturbance monitoring method based on image interpolation as described in any one of claims 1 to 3.

Images