A forest monitoring system and method based on remote sensing data

Abstract

The invention discloses a forest monitoring system and method based on remote sensing data, and specifically relates to the technical field of remote sensing monitoring, comprising the following steps: obtaining optical and radar remote sensing data in a study area, and performing preprocessing and quality analysis to obtain optical data quality index and radar data quality index respectively, generating fusion data quality coefficient through comprehensive analysis, dividing data fusion quality into high quality and low quality states, selecting regional division method 2 for high quality data, selecting regional division method 1 for low quality data, and determining a single data source, and performing alternative optimization processing when method 2 is selected; the invention can flexibly adjust the regional division method according to the dynamic evaluation of data quality, ensure the classification effect under different quality states, provide strong support for forest resource management and protection through accurate forest monitoring data, and help decision makers formulate scientific and reasonable management strategies.

Description

Technical Field

[0001] The present invention relates to the field of remote sensing monitoring technology, and more specifically, to a forest monitoring system and method based on remote sensing data. Background Art

[0002] High-resolution forest maps can more accurately reflect information such as forest coverage, tree species distribution, and tree density, which is conducive to the government and management departments to better understand the status of forest resources, formulate targeted protection and utilization measures, and provide detailed spatial information references for forestry production activities such as stand management, tree felling, afforestation and replanting, so as to improve the accuracy and effectiveness of production activities. By regularly updating high-resolution forest maps, changes in forest coverage can be dynamically monitored, and problems such as forest degradation can be discovered and responded to in a timely manner. Therefore, high-resolution forest maps can provide strong support for the scientific management and sustainable development of forest resources, and are an important part of the current forest informatization construction.

[0003] Methods for forest monitoring can be roughly divided into two categories: the first category generates forest classification maps based on single optical remote sensing data or radar data. Sentinel-2 images are widely used to extract forest information due to their high spatial resolution, frequent repeated visits, and rich spectral bands; Li et al. calculated the classification phenological index based on Sentinel-1 / 2 data by extracting the differences in phenological characteristics between evergreen forests and deciduous forests, and extracted the 2019 evergreen forest map from the annual forest map based on the constructed phenological indicators; Jeong et al. proposed a method to use Sentinel remote sensing images and Aerial photographs were used to map urban forest coverage in 2021, and city-scale urban forest cover maps were drawn for three pilot study areas in the Korean capital region, marking deciduous forests, coniferous forests, grasslands, wetlands, bare land, farmland, urban areas and water bodies; Wang et al. used Sentinel-2 images on the Google Earth Engine (GEE) platform to map forest distribution, and compared and analyzed the classification results, using different classification algorithms and feature variable combinations, selecting the optimal image seasonality, feature combination and classification algorithm, and compiling the Jiangxi Province forest map from 2019 to 2021.

[0004] The second method is the fusion of optical data and radar data. Zhang et al. used a forest classification method based on a decision tree to fuse PALSAR / PALSAR-2 and MODIS NDVI data to generate 25-meter forest maps for Paraguay, the United States, and Russia from 2007 to 2010 and 2015 to 2016; Qing et al. used images from phased array L-band synthetic aperture radar (PALSAR) and time series images from Moderate Resolution Imaging Spectroradiometer (MODIS) to generate annual forest cover maps for the Brazilian Amazon from 2007 to 2010 and evergreen forest cover maps from 2000 to 2021; Qin et al. used L-band synthetic aperture radar data (PALSAR) and vegetation index products from Moderate Resolution Imaging Spectroradiometer (MODIS) and extracted forest and non-forest classification maps for China in 2010 using a decision tree classification method; Zhang et al. used PALSAR / PALSAR2 and MODIS from Paraguay, the United States, and Russia from 2007 to 2010 and 2015 to 2016 to generate annual forest cover maps for the Brazilian Amazon from 2007 to 2010 and evergreen forest cover maps from 2000 to 2021. NDVI data is used to generate more accurate 25-meter forest maps through a decision tree approach.

[0005] How to choose the appropriate method when drawing forest maps is a problem that needs to be solved. Therefore, a forest monitoring system and method based on remote sensing data is proposed. Summary of the invention

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A forest monitoring method based on remote sensing data comprises the following steps:

[0008] Obtain optical remote sensing data and radar remote sensing data within the study area;

[0009] The acquired data is preprocessed and then quality analyzed to obtain the optical data quality index and radar data quality index respectively;

[0010] Based on the optical data quality index and the radar data quality index, a comprehensive analysis is performed to generate a fusion data quality coefficient. According to the fusion data quality coefficient, the fusion quality of the optical remote sensing data and the radar remote sensing data is divided into a high-quality fusion state and a low-quality fusion state.

[0011] For optical remote sensing data and radar remote sensing data in high-quality fusion state, select regional division method 2; for optical remote sensing data and radar remote sensing data in low-quality fusion state, select regional division method 1 and determine the single data source used in division method 1;

[0012] When the second area division method is selected, alternative optimization processing is performed according to the preset alternative strategy.

[0013] In a preferred embodiment, preprocessing the acquired data includes performing cloud removal, radiation correction and geometric correction on the acquired optical data, and performing radiation correction and geometric correction on the acquired radar data.

[0014] In a preferred embodiment, the logic for obtaining the optical data quality index is:

[0015] Obtain the number of cloud-covered pixels and the total number of pixels in each optical data sample, and then calculate the cloud cover ratio of each optical data sample, where the cloud cover ratio is obtained by dividing the number of cloud-covered pixels by the total number of pixels;

[0016] Obtain the radiation value before and after correction for each optical data sample in the radiation correction, and then calculate the radiation correction residual of each optical data sample. The calculation formula is:

[0017] RCRI represents the radiation correction residual of sample number i, F1i represents the radiation value of sample number i before correction, F2i represents the radiation value of sample number i after correction, and N is the total number of pixels;

[0018] The actual position and expected position of each optical data sample in the geometric correction are obtained, and then the geometric correction residual of each optical data sample is calculated. The calculation formula is:

[0019] GCRi represents the geometric correction residual of sample number i, J1i represents the actual position of sample number i, J2i represents the expected position of sample number i, and M is the number of comparison points;

[0020] Using principal component analysis, the characteristic matrix X is constructed based on the cloud cover ratio, geometric correction residual, and radiation correction residual of each optical data sample after standardization, and then the covariance matrix C is calculated. XX T What is calculated is the covariance product between the variables, X T is the transpose of the characteristic matrix, n is the total number of samples, and the covariance matrix is ​​decomposed into eigenvalues ​​to obtain the eigenvalues ​​λ and eigenvectors v: Cv = λv; using the "inflection point" method, the eigenvalues ​​of each principal component are plotted, and the "inflection point" in the figure is found, that is, the point where the eigenvalue drops significantly, and the principal components before the inflection point are selected for retention, and then the mean of the retained principal components is calculated and marked as μPt, where Pt is the principal component numbered t, and then weighted calculation is performed to obtain the optical data quality index, ODQI represents the optical data quality index, T represents the total number of principal components retained, and wt represents the preset proportional coefficient corresponding to the retained principal components.

[0021] In a preferred embodiment, the acquisition logic of the radar data quality index is:

[0022] Obtain the signal strength and noise strength in each radar data sample, and then calculate the signal-to-noise ratio of each radar data sample, where the signal-to-noise ratio is obtained by dividing the signal strength by the noise strength;

[0023] Obtaining the radiation value before and after correction of each radar data sample in the radiation correction, and then calculating the radiation correction residual of each radar data sample;

[0024] Obtaining the actual position and expected position of each radar data sample in geometric correction, and then calculating the geometric correction residual of each radar data sample;

[0025] Based on the signal-to-noise ratio, radiation correction residual, and geometric correction residual of each radar data sample after standardization, the information entropy is calculated using the following formula: Ek represents the information entropy of the kth indicator, Pik represents the proportion of the standardized value of the i-th sample on the kth indicator, and the calculation formula is: Z represents the index value, N represents the total number of samples, and then the weight is determined: m represents the total number of indicators, Wk represents the weight value corresponding to indicator k, and then weighted calculation is performed to obtain the radar data quality index: μPk represents the average value corresponding to indicator k, μBk represents the preset standard value corresponding to indicator k, and RDQI represents the radar data quality index.

[0026] In a preferred embodiment, the acquisition logic of the fusion data quality coefficient is:

[0027] The radar data quality index and the optical data quality index are obtained, and the radar data quality index and the optical data quality index are substituted into the pre-trained machine learning model to generate a fusion data quality coefficient.

[0028] In a preferred embodiment, dividing the fusion quality of optical remote sensing data and radar remote sensing data into a high-quality fusion state and a low-quality fusion state according to the fusion data quality coefficient means:

[0029] The fusion data quality coefficient is compared with the preset standard parameter threshold. If the fusion data quality coefficient is greater than or equal to the preset standard parameter threshold, the fusion quality of the optical remote sensing data and the radar remote sensing data is classified as a low-quality fusion state. If the fusion data quality coefficient is less than the preset standard parameter threshold, the fusion quality of the optical remote sensing data and the radar remote sensing data is classified as a high-quality fusion state.

[0030] In a preferred embodiment, for optical remote sensing data and radar remote sensing data in a low-quality fusion state, regional division method one is selected, and based on the radar data quality index and the optical data quality index, fuzzy reasoning is used to determine the single data source used in division method one.

[0031] In a preferred embodiment, a forest monitoring system based on remote sensing data includes a data acquisition module, a quality analysis module, a comprehensive analysis module, a state division module, a division mode selection module, and an alternative optimization module;

[0032] The data acquisition module is used to obtain optical remote sensing data and radar remote sensing data in the study area;

[0033] The quality analysis module is used to pre-process the acquired data and then perform quality analysis to obtain the optical data quality index and radar data quality index respectively;

[0034] The comprehensive analysis module is used to perform comprehensive analysis based on the optical data quality index and the radar data quality index to generate a fusion data quality coefficient;

[0035] The state division module is used to divide the fusion quality of optical remote sensing data and radar remote sensing data into a high-quality fusion state and a low-quality fusion state according to the fusion data quality coefficient;

[0036] The division method selection module is used to select the second regional division method for optical remote sensing data and radar remote sensing data in a high-quality fusion state, and to select the first regional division method for optical remote sensing data and radar remote sensing data in a low-quality fusion state and determine the single data source used by the first division method;

[0037] The alternative optimization module is used to perform alternative optimization processing according to a preset alternative strategy when the second area division method is selected.

[0038] Technical effects and advantages of the present invention:

[0039] The present invention realizes efficient and accurate forest monitoring through the coordinated work of a data acquisition module, a quality analysis module, a comprehensive analysis module, a state division module, a division mode selection module and an alternative optimization module; through the quality analysis module, optical and radar remote sensing data are preprocessed and quality analyzed to obtain an optical data quality index (ODQI) and a radar data quality index (RDQI), respectively, thereby ensuring the high quality of input data and providing a reliable basis for subsequent analysis and processing; the comprehensive analysis module generates a fusion data quality coefficient based on the ODQI and the RDQI; the state division module divides the data into a high-quality fusion state and a low-quality fusion state according to the coefficient; according to the level of the fusion data quality coefficient, the data processing and region division strategies are flexibly adjusted to ensure classification accuracy and reliability.

[0040] For data in a high-quality fusion state, the present invention selects the second regional division method to perform refined regional division. For data in a low-quality fusion state, the present invention selects the first regional division method and determines a single data source through fuzzy reasoning to perform simple and reliable regional division. The regional division method is dynamically adjusted according to data quality, thereby improving the adaptability and flexibility of the system. When the alternative optimization module selects the second regional division method, it performs alternative optimization processing according to a preset alternative strategy, and performs comparative optimization based on the methods in existing research.

[0041] Through high-quality data fusion and optimization processing, the overall classification accuracy of the system in the present invention reaches 97.62%, which is significantly higher than the classification accuracy of the existing fusion method (85%-88%). It can utilize the complementarity of multi-source data to process areas with complex terrain and diverse vegetation types, and provide more detailed and accurate forest information. It can flexibly adjust the regional division method according to the dynamic evaluation of data quality to ensure the classification effect under different quality conditions. Through accurate forest monitoring data, it provides strong support for forest resource management and protection, and helps decision makers to formulate scientific and reasonable management strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to facilitate understanding by those skilled in the art, the present invention is further described below in conjunction with the accompanying drawings;

[0043] Figure 1 This is a schematic diagram of a forest monitoring method based on remote sensing data in the present invention.

[0044] Figure 2 This is a schematic diagram of a forest monitoring system based on remote sensing data in the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] Reference Figure 1 - Figure 2 The following embodiments are obtained:

[0047] Embodiment 1: A forest monitoring method based on remote sensing data comprises the following steps:

[0048] Obtain optical remote sensing data and radar remote sensing data within the study area;

[0049] The acquired data is preprocessed and then quality analyzed to obtain the optical data quality index and radar data quality index respectively;

[0050] Based on the optical data quality index and the radar data quality index, a comprehensive analysis is performed to generate a fusion data quality coefficient. According to the fusion data quality coefficient, the fusion quality of the optical remote sensing data and the radar remote sensing data is divided into a high-quality fusion state and a low-quality fusion state.

[0051] For optical remote sensing data and radar remote sensing data in high-quality fusion state, select regional division method 2; for optical remote sensing data and radar remote sensing data in low-quality fusion state, select regional division method 1 and determine the single data source used in division method 1;

[0052] When the second area division method is selected, alternative optimization processing is performed according to the preset alternative strategy.

[0053] The preprocessing of the acquired data includes cloud removal, radiation correction and geometric correction for the acquired optical data, and radiation correction and geometric correction for the acquired radar data.

[0054] The logic for obtaining the optical data quality index is:

[0055] Obtain the number of cloud-covered pixels and the total number of pixels in each optical data sample. The number of cloud-covered pixels and the total number of pixels are obtained from optical remote sensing data. Optical remote sensing data usually contains pixel values ​​of multiple bands, and each pixel value represents the reflectivity of a small area on the ground. In the image processing process, the number of pixels covered by clouds can be identified by the cloud detection algorithm; the number of cloud-covered pixels refers to the number of pixels detected as cloud-covered in the image; these pixels are blocked by clouds and cannot provide real information about the surface; the total number of pixels refers to the total number of all pixels in the image, including pixels covered by clouds and pixels not covered by clouds; then calculate the cloud coverage ratio of each optical data sample, which is obtained by dividing the number of cloud-covered pixels by the total number of pixels; obtain the radiation value before and after correction for each optical data sample in radiation correction, and then calculate the radiation correction residual of each optical data sample. The calculation formula is: RCRi represents the radiation correction residual of sample number i, F1i represents the radiation value of sample number i before correction, F2i represents the radiation value of sample number i after correction, and N is the total number of pixels; the actual position and expected position of each optical data sample in the geometric correction are obtained, and then the geometric correction residual of each optical data sample is calculated. The calculation formula is: GCRi represents the geometric correction residual of sample number i, J1i represents the actual position of sample number i, J2i represents the expected position of sample number i, and M is the number of comparison points;

[0056] Using principal component analysis, the characteristic matrix X is constructed based on the cloud cover ratio, geometric correction residual, and radiation correction residual of each optical data sample after standardization, and then the covariance matrix C is calculated. The covariance matrix C is used to describe the correlation between various variables. XX T What is calculated is the covariance product between the variables, X T is the transpose of the feature matrix, n is the total number of samples, and the covariance matrix is ​​decomposed into eigenvalues ​​to obtain eigenvalues ​​λ and eigenvectors v: Cv = λv; the eigenvalue represents the importance of each principal component, that is, the proportion of the total variance, and the eigenvector represents the direction of the principal component, that is, the weight of the linear combination;

[0057] Use the "inflection point" method to plot the eigenvalue of each principal component, and look for the "inflection point" in the figure, that is, the point where the eigenvalue drops significantly, select the principal component before the inflection point to retain, plot the eigenvalue of each principal component, and look for the "inflection point" in the figure, that is, the point where the eigenvalue drops significantly, and select the principal component before the inflection point. The specific steps are as follows: mark the principal component number on the horizontal axis, mark the corresponding eigenvalue on the vertical axis, and plot the eigenvalue versus the principal component number. Observe the "inflection point" in the figure, that is, the point where the eigenvalue drops significantly, and select the principal component before the inflection point. Assume that the eigenvalue drops significantly after P2 is observed, so the first two principal components (P1 and P2) are selected, and then the mean of the retained principal components is calculated and marked as μPt, where Pt is the principal component numbered t, and then a weighted calculation is performed to obtain the optical data quality index. ODQI represents the optical data quality index, T represents the total number of principal components retained, and wt represents the preset proportional coefficient corresponding to the retained principal components.

[0058] Impact of cloud cover: Optical data is greatly affected by weather conditions, and cloud cover will significantly reduce data quality. ODQI can effectively evaluate the cloud coverage of the data, thereby judging the availability of the data. Radiometric correction error: Radiometric correction affects the true reflection ability of the data, and data with large errors will reduce classification accuracy. ODQI evaluates the effect of radiometric correction through RCE to ensure the reliability of the data. Geometric accuracy: Geometric correction affects the spatial positioning accuracy of the data, and large errors will cause data misalignment. ODQI evaluates geometric accuracy through GPE to ensure the spatial consistency of the data. The larger the value of the optical data quality index, the worse the data quality. A high ODQI value means more cloud coverage, poor radiometric correction and geometric correction effects, and low data accuracy and reliability.

[0059] The logic for obtaining the radar data quality index is:

[0060] Obtain the signal strength and noise strength in each radar data sample, and then calculate the signal-to-noise ratio of each radar data sample, where the signal-to-noise ratio is obtained by dividing the signal strength by the noise strength;

[0061] Obtain the radiation value before and after correction of each radar data sample in radiation correction, then calculate the radiation correction residual of each radar data sample, and calculate the process of obtaining the optical data quality index of the formula; obtain the actual position and expected position of each radar data sample in geometric correction, then calculate the geometric correction residual of each radar data sample, and calculate the process of obtaining the optical data quality index of the formula;

[0062] Based on the signal-to-noise ratio, radiation correction residual, and geometric correction residual of each radar data sample after standardization, the information entropy is calculated using the following formula: Ek represents the information entropy of the kth indicator, reflecting the amount of information and uncertainty of the indicator. The larger the information entropy, the greater the uncertainty of the indicator and the less information; the smaller the information entropy, the greater the amount of information of the indicator. When calculating the information entropy, the base of the logarithm is usually taken as the natural logarithm. Pik represents the proportion of the standardized value of the i-th sample on the k-th indicator. The calculation formula is: Z represents the index value, and N represents the total number of samples. For example, if there are 10 sample data sets used to calculate the entropy value, then N = 10, and then the weight is determined: m represents the total number of indicators, that is, the number of different indicators used to evaluate data quality. For example, information entropy, radiation correction residual and geometric correction residual are used to evaluate optical data quality. m = 3, Wk represents the weight value corresponding to indicator k. The smaller the information entropy, the greater the weight, indicating that it is more important in the comprehensive evaluation. Then, weighted calculation is performed to obtain the radar data quality index: μPk represents the average value corresponding to indicator k, μBk represents the preset standard value corresponding to indicator k, and RDQI represents the radar data quality index.

[0063] Signal-to-noise ratio (SNR): Radar data will be affected by various noise interferences during the acquisition process, and the signal-to-noise ratio directly affects the data quality. RDQI evaluates the signal-to-noise ratio of the data through SNR to ensure the clarity and reliability of the data. Radiation correction error: Similarly, the radiation correction of radar data also affects its true reflection ability. RDQI evaluates the radiation correction effect of radar data through RCE to ensure the reliability of the data. Geometric accuracy: The geometric correction of radar data affects its spatial positioning accuracy, especially in areas with complex terrain. RDQI evaluates the geometric accuracy of radar data through GPE to ensure the spatial consistency of the data. The larger the value of the radar data quality index, the worse the data quality. A high RDQI value means a low signal-to-noise ratio, poor radiation correction and geometric correction effects, and low data accuracy and reliability.

[0064] The logic for obtaining the fusion data quality coefficient is as follows: obtain the radar data quality index and the optical data quality index, substitute the radar data quality index and the optical data quality index into the pre-trained machine learning model to generate the fusion data quality coefficient, obtain the consistency anomaly index and the deviation anomaly index of the current target detection equipment, and then substitute the consistency anomaly index and the deviation anomaly index into the pre-trained machine learning model to obtain the comprehensive detection quality coefficient; the machine learning model is not specifically limited here, and any model that can comprehensively analyze the radar data quality index and the optical data quality index to generate the fusion data quality coefficient is acceptable. In order to implement the technical solution of the present invention, the present invention provides a specific implementation method: RH=r1*RDQI+r2*ODQI; RH is the fusion data quality coefficient, the larger the value of the fusion data quality coefficient, the worse the fusion quality, and r1 and r2 are both preset The proportional coefficient is used to adjust the contribution ratio of the radar data quality index and the optical data quality index, and can be adjusted according to actual conditions; dividing the fusion quality of optical remote sensing data and radar remote sensing data into high-quality fusion state and low-quality fusion state according to the fusion data quality coefficient means: comparing the fusion data quality coefficient with the preset standard parameter threshold. If the fusion data quality coefficient is greater than or equal to the preset standard parameter threshold, the fusion quality of the optical remote sensing data and the radar remote sensing data is divided into a low-quality fusion state. If the fusion data quality coefficient is less than the preset standard parameter threshold, the fusion quality of the optical remote sensing data and the radar remote sensing data is divided into a high-quality fusion state. The larger the value of the fusion data quality coefficient, the worse the fusion quality. A high fusion data quality coefficient means that the comprehensive quality of the optical data and the radar data is low, and there may be more errors and uncertainties, which is not suitable for accurate forest monitoring and analysis.

[0065] For optical remote sensing data and radar remote sensing data in low-quality fusion state, select regional division method 1, and use fuzzy reasoning to determine the single data source used in division method 1 based on radar data quality index and optical data quality index; define fuzzy sets and membership functions for optical data quality index (ODQI) and radar data quality index (RDQI), respectively. For example, define ODQI and RDQI as three fuzzy sets of "low", "medium" and "high", respectively. Construct fuzzy rules: According to different combinations of ODQI and RDQI, formulate different rules to decide which single data source to use. For example: if ODQI is low and RDQI is high, radar data is selected; if ODQI is high and RDQI is low, optical data is selected. Fuzzy reasoning: Use fuzzy reasoning mechanism to match fuzzy rules according to the input ODQI and RDQI values ​​to obtain the output membership. Fuzzy decision: According to the result of fuzzy reasoning, determine the single data source to be finally selected.

[0066] Embodiment 2: A forest monitoring system based on remote sensing data, comprising a data acquisition module, a quality analysis module, a comprehensive analysis module, a state division module, a division mode selection module, and an alternative optimization module;

[0067] The data acquisition module is used to obtain optical remote sensing data and radar remote sensing data in the study area;

[0068] The quality analysis module is used to pre-process the acquired data and then perform quality analysis to obtain the optical data quality index and radar data quality index respectively;

[0069] The comprehensive analysis module is used to perform comprehensive analysis based on the optical data quality index and the radar data quality index to generate a fusion data quality coefficient;

[0070] The state division module is used to divide the fusion quality of optical remote sensing data and radar remote sensing data into a high-quality fusion state and a low-quality fusion state according to the fusion data quality coefficient;

[0071] The division method selection module is used to select the second regional division method for optical remote sensing data and radar remote sensing data in a high-quality fusion state, and to select the first regional division method for optical remote sensing data and radar remote sensing data in a low-quality fusion state and determine the single data source used by the first division method;

[0072] The alternative optimization module is used to perform alternative optimization processing according to the preset alternative strategy when the second regional division method is selected. Although the MODIS data fused with Palsar data in the second regional division method improves the accuracy of forest monitoring, the spatial resolution of MODIS data is 250 meters, 500 meters or 1000 meters, compared with the spatial resolution of 10-30 meters of high-resolution satellite data such as Landsat, it has limitations in extracting detailed information and monitoring small-scale forest changes. Although the MODIS sensor also has 36 spectral bands, the band setting is relatively broad and cannot provide detailed spectral information like Landsat, which to a certain extent affects the recognition accuracy of forest tree species and forest phases. Landsat satellites have long-term data series of more than 50 years, which can well capture seasonal changes and multi-year forest succession processes, which is very important for improving the stability and reliability of forest classification. Substitution optimization processing refers to the fusion of Landsat and PALSAR-2 remote sensing data, replacing MODIS data with Landsat data to obtain the forest distribution map of Helan Mountain from 2015 to 2022. Taking the monitoring and analysis of forest distribution in Helan Mountain, Ningxia as an example, both data are free and open global coverage data with high spatial resolution, which can achieve fine monitoring of forest resources in a certain area. Comprehensively utilize the rich multispectral information of Landsat optical remote sensing data and the geometric information of PALSAR-2 radar remote sensing data to improve the overall accuracy of forest monitoring. Specifically, based on Landsat multispectral data and ALOS PALSAR-2 radar data from 2015 to 2022, the NDVI, HH / HV, HH-HV, HH, and HV values ​​of the Helan Mountain forest in Ningxia were extracted by the decision threshold method to generate a forest distribution map. On this basis, the changes in forest cover in Helan Mountain and its driving factors between 2015 and 2022 were analyzed, providing an important scientific basis for the rational protection and management of regional forest resources. The evaluation results are shown in Table 1 Evaluation Form:

[0073] Table 1 Evaluation form

[0074]

[0075]

[0076] Explanation of the improvement after alternative optimization: Taking the improvement of spatial resolution as an example, before the replacement: the spatial resolution of other comparison methods (such as PALSAR / PALSAR-2+MODIS NDVI, PALSAR+MODIS time series images, PALSAR+MODIS vegetation index products) was 250 meters.

[0077] After replacement: The spatial resolution of the PALSAR-2+Landsat method is 30 meters, which is significantly higher than other methods. Higher spatial resolution means that more detailed ground information can be captured, which improves the fineness and accuracy of classification. It integrates the data of PALSAR-2 and Landsat, combines the advantages of optical remote sensing and radar remote sensing, and provides higher spatial resolution and classification accuracy.

[0078] Through alternative optimization, the fusion of PALSAR-2+Landsat is significantly superior to other comparative methods in many aspects, including higher spatial resolution, overall accuracy, user accuracy, producer accuracy and Kappa coefficient. Alternative optimization not only enhances the effect of data fusion, but also further improves the accuracy and reliability of classification results. These improvements make the present invention have higher practical value and application prospects in forest monitoring and surface feature identification.

[0079] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0080] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0081] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0082] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0083] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A forest monitoring method based on remote sensing data, characterized in that: The following steps are involved: Obtain optical remote sensing data and radar remote sensing data within the study area; The acquired data is preprocessed and then quality analyzed to obtain the optical data quality index and radar data quality index respectively; Based on the optical data quality index and the radar data quality index, a comprehensive analysis is performed to generate a fusion data quality coefficient. According to the fusion data quality coefficient, the fusion quality of the optical remote sensing data and the radar remote sensing data is divided into a high-quality fusion state and a low-quality fusion state. For optical remote sensing data and radar remote sensing data in high-quality fusion state, select regional division method 2; for optical remote sensing data and radar remote sensing data in low-quality fusion state, select regional division method 1 and determine the single data source used in division method 1; When the second regional division method is selected, alternative optimization processing is performed according to the preset alternative strategy; The logic for obtaining the optical data quality index is: Obtain the number of cloud-covered pixels and the total number of pixels in each optical data sample, and then calculate the cloud cover ratio of each optical data sample, where the cloud cover ratio is obtained by dividing the number of cloud-covered pixels by the total number of pixels; Obtain the radiation value before and after correction for each optical data sample in the radiation correction, and then calculate the radiation correction residual of each optical data sample. The calculation formula is: RCRi represents the radiation correction residual of sample number i, F1i represents the radiation value of sample number i before correction, F2i represents the radiation value of sample number i after correction, and N is the total number of pixels; The actual position and expected position of each optical data sample in the geometric correction are obtained, and then the geometric correction residual of each optical data sample is calculated. The calculation formula is: GCRi represents the geometric correction residual of sample number i, J1i represents the actual position of sample number i, J2i represents the expected position of sample number i, and M is the number of comparison points; Using principal component analysis, the characteristic matrix X is constructed based on the cloud cover ratio, geometric correction residual, and radiation correction residual of each optical data sample after standardization, and then the covariance matrix C is calculated. XX T What is calculated is the covariance product between the variables, X T is the transpose of the characteristic matrix, n is the total number of samples, and the covariance matrix is ​​decomposed by eigenvalue to obtain the eigenvalue λ and eigenvector v: Cv = λv; using the "inflection point" method, the eigenvalue of each principal component is plotted, and the "inflection point" in the figure is found, that is, the point where the eigenvalue drops significantly, and the principal components before the inflection point are selected for retention, and then the mean of the retained principal components is calculated and marked as μPt, Pt is the principal component numbered t, and then weighted calculation is performed to obtain the optical data quality index, ODQI represents the optical data quality index, T represents the total number of principal components retained, and wt represents the preset proportional coefficient corresponding to the retained principal components; The logic for obtaining the radar data quality index is: Obtain the signal strength and noise strength in each radar data sample, and then calculate the signal-to-noise ratio of each radar data sample, where the signal-to-noise ratio is obtained by dividing the signal strength by the noise strength; Obtaining the radiation value before and after correction of each radar data sample in the radiation correction, and then calculating the radiation correction residual of each radar data sample; Obtain the actual position and expected position of each radar data sample in geometric correction, and then calculate the geometric correction residual of each radar data sample; Based on the signal-to-noise ratio, radiation correction residual, and geometric correction residual of each radar data sample after standardization, the information entropy is calculated using the following formula: Ek represents the information entropy of the kth indicator, Pik represents the proportion of the standardized value of the i-th sample on the kth indicator, and the calculation formula is: Z represents the index value, N represents the total number of samples, and then the weight is determined: m represents the total number of indicators, Wk represents the weight value corresponding to indicator k, and then weighted calculation is performed to obtain the radar data quality index: μPk represents the average value corresponding to indicator k, μBk represents the preset standard value corresponding to indicator k, and RDQI represents the radar data quality index.

2. A forest monitoring method based on remote sensing data according to claim 1, characterized in that: The preprocessing of the acquired data includes cloud removal, radiation correction and geometric correction for the acquired optical data, and radiation correction and geometric correction for the acquired radar data.

3. A forest monitoring method based on remote sensing data according to claim 2, characterized in that: The logic for obtaining the fusion data quality coefficient is: The radar data quality index and the optical data quality index are obtained, and the radar data quality index and the optical data quality index are substituted into the pre-trained machine learning model to generate a fusion data quality coefficient.

4. A forest monitoring method based on remote sensing data according to claim 3, characterized in that: According to the fusion data quality coefficient, the fusion quality of optical remote sensing data and radar remote sensing data is divided into high-quality fusion state and low-quality fusion state, which means: The fusion data quality coefficient is compared with the preset standard parameter threshold. If the fusion data quality coefficient is greater than or equal to the preset standard parameter threshold, the fusion quality of the optical remote sensing data and the radar remote sensing data is classified as a low-quality fusion state. If the fusion data quality coefficient is less than the preset standard parameter threshold, the fusion quality of the optical remote sensing data and the radar remote sensing data is classified as a high-quality fusion state.

5. A forest monitoring method based on remote sensing data according to claim 4, characterized in that: For optical remote sensing data and radar remote sensing data in a low-quality fusion state, regional division method one is selected. Based on the radar data quality index and the optical data quality index, fuzzy reasoning is used to determine the single data source used in division method one.

6. A forest monitoring system based on remote sensing data, based on a forest monitoring method based on remote sensing data according to any one of claims 1 to 5, characterized in that: It includes data acquisition module, quality analysis module, comprehensive analysis module, state division module, division method selection module and alternative optimization module; The data acquisition module is used to obtain optical remote sensing data and radar remote sensing data in the study area; The quality analysis module is used to pre-process the acquired data and then perform quality analysis to obtain the optical data quality index and radar data quality index respectively; The comprehensive analysis module is used to perform comprehensive analysis based on the optical data quality index and the radar data quality index to generate a fusion data quality coefficient; The state division module is used to divide the fusion quality of optical remote sensing data and radar remote sensing data into a high-quality fusion state and a low-quality fusion state according to the fusion data quality coefficient; The division method selection module is used to select the second regional division method for optical remote sensing data and radar remote sensing data in a high-quality fusion state, and to select the first regional division method for optical remote sensing data and radar remote sensing data in a low-quality fusion state and determine the single data source used by the first division method; The alternative optimization module is used to perform alternative optimization processing according to a preset alternative strategy when the second area division method is selected.

Images