A method, device and computing equipment for tree species identification based on remote sensing images

Abstract

The present application provides a method for tree species identification based on remote sensing images, comprising: obtaining long-term remote sensing forest images of a study area; the long-term remote sensing forest images include all remote sensing forest images within a set year; calculating NDFI values ​​for image bands based on the long-term remote sensing forest images to obtain NDFI value-time series curves; the NDFI value is a normalized degradation fraction index, which is used to indicate the degradation and health status of vegetation; reconstructing the NDFI value-time series curve of the tree species based on time harmonics to obtain the NDFI time series growth trajectory of the tree species; determining tree species identification features based on the NDFI time series growth trajectory of the tree species, the tree species identification features are used to describe the change pattern and error of the time series model; and stratifying and classifying the tree species in the study area through a classification model based on the tree species identification features.

Description

Technical Field

[0001] The present application relates to the technical field of environmental monitoring and resource management, and in particular to a method, device and computing equipment for tree species identification based on remote sensing images. Background Art

[0002] Trees are important reservoirs of carbon by absorbing carbon dioxide and fixing it in biomass. Different tree species have different growth rates, wood density and crown structure, which directly affect the biomass of trees. By accurately identifying tree species, different models and parameters are selected for different tree species to estimate biomass, thereby improving the accuracy of the estimation.

[0003] Traditional tree species surveys are mainly based on manual field surveys, supplemented by simple measurement and recording tools. This method is not only time-consuming and labor-intensive, but also fails to update forest resource information in a timely and effective manner. Remote sensing technology can obtain a wide range of various types of surface information in a relatively short period of time, providing more comprehensive and integrated geographic information for tree species identification, significantly improving the efficiency and accuracy of tree species identification.

[0004] At present, there have been many achievements in the research of tree species identification based on single-period remote sensing images. There have been inventions that use drone hyperspectral images to obtain characteristic bands and vegetation index and texture index as independent variables to identify dominant tree species. However, due to the existence of the "different objects with the same spectrum" phenomenon and the complex and changeable forest structure and composition, tree species identification based only on single-period images will result in low classification accuracy for some tree species.

[0005] The growth and development status and periodic changes of plants in different seasons are affected by phenological characteristics. At some key phenological time points, tree species will show significantly different seasonal characteristics and exhibit significantly different image characteristics in remote sensing images. With the improvement of the temporal resolution of remote sensing images, many results have emerged in tree species identification based on long-term remote sensing images.

[0006] The existing tree species methods mainly have the following problems:

[0007] (1) Research on tree species identification based on single-period high-resolution commercial satellite images, on the one hand, ignores the key nodes of the vegetation growth process and cannot effectively overcome the phenomenon of "same species, different spectra" or "different species, same spectra" in the tree species identification process; on the other hand, high spatial resolution commercial satellite images are limited by data cost and coverage, and are difficult to apply to long-term and large-scale forest information monitoring;

[0008] (2) The features used in existing studies on tree species identification based on time series remote sensing images are mostly remote sensing indices of one or more single static time nodes, ignoring the overall dynamic change characteristics of the tree species time series trajectory curve;

[0009] (3) Existing tree species identification based on time-series remote sensing images usually selects NDVI as the characteristic variable. However, when the vegetation density is low or the vegetation coverage is uneven, the NDVI value may have large errors.

[0010] Therefore, how to provide a method for identifying dominant tree species in forests that fully considers the characteristics of tree species' phenological growth trajectories is a technical problem that needs to be solved urgently. Summary of the invention

[0011] In the first aspect, an embodiment of the present application provides a method for tree species identification based on remote sensing images, including: obtaining long-term remote sensing forest images of a study area; the long-term remote sensing forest images include all remote sensing forest images within a set year; calculating NDFI values ​​for image bands based on the long-term remote sensing forest images to obtain NDFI value-time series curves; NDFI values ​​are used to indicate the degradation and health status of vegetation; reconstructing the NDFI value-time series curves of tree species based on time series harmonics to obtain the NDFI time series growth trajectory of the tree species; determining tree species identification features based on the NDFI time series growth trajectory of the tree species, the tree species identification features are used to describe the change pattern and error of the time series model; and classifying the tree species in the study area through a classification model based on the tree species identification features.

[0012] In some possible implementations, obtaining long-term remote sensing forest images of a study area includes: preprocessing the long-term remote sensing forest images to obtain forest growth images, wherein the preprocessing includes time and cloud amount screening, cloud and shadow masking, and image cropping.

[0013] In some possible implementations, the NDFI value of a tree species is obtained by performing image band calculation based on the forest growth image, including: the NDFI value calculation formula is as follows:

[0014] NDFI=(GV-(NPV+Soil)) / (GV shade +(NPV+Soil))

[0015] GV shade =GV / (1-Shade)

[0016] Among them, GV is the green vegetation component, which indicates the proportion of vegetation that can photosynthesize in the forest growth image; Shade is the shadow component, which indicates the shadow component of the forest canopy in the forest growth image; NPV is the non-photosynthetic vegetation component, which indicates the proportion of vegetation that cannot photosynthesize or aging vegetation in the forest growth image; Soil represents the soil component; on the NDFI value-time curve, the NDFI value of a certain time period is close to 1, indicating that the vegetation is lush in that period; the NDFI value of a certain time period decreases, indicating that the vegetation is deciduous or disturbed in that period; the NDFI value of a certain time period is close to -1, indicating that the forest has completely disappeared in that period.

[0017] In some possible implementations, the NDFI value-time curve of the tree species is reconstructed based on time harmonics to obtain the NDFI time-series growth trajectory of the tree species, including: obtaining a discrete wave signal of the surface NDFI value corresponding to each pixel in the forest growth image that changes with time; removing the points that are greatly affected by cloud pollution and deviate from the maximum threshold in the discrete wave signal based on the minimum multiplication method to obtain a denoised waveform; using Fourier transform to transform the denoised waveform in the time domain to the frequency domain, and reconstructing based on time harmonics to obtain the fitting value of the NDFI time-series growth trajectory of the tree species as follows:

[0018]

[0019] In the formula, y i is the fitting value of the NDFI time series growth trajectory; i is the time mark of the fitting value y; A0 is the harmonic remainder; A j is the harmonic amplitude; ω j is the harmonic frequency, N is the length of the time series; θ j is the initial phase of harmonic; m is the number of harmonics, and j is the harmonic label.

[0020] In some possible embodiments, tree species identification characteristics are determined according to the NDFI time-series growth trajectory of the tree species, including: determining the tree species identification characteristics according to the NDFI time-series growth trajectory of the tree species, the tree species identification characteristics indicating the time-series change trend and periodic change law of different tree species; the tree species identification characteristics include intercept (INTP), slope (SLP), COS, SIN, COS2, SIN2, COS3, SIN3; wherein, INTP (Intercept) and SLP (Slope) are linear regression parameters for fitting the time series model; COS, SIN, COS2, SIN2, COS3, SIN3 and RMSE are parameters for describing the change pattern and error of the time series model; determining the time series model according to the identification characteristics of the tree species; the time series model is used to fit the growth process of different tree species, including leaf fall and growth.

[0021] In some possible implementations, tree species in the study area are stratified and classified through a classification model based on tree species identification characteristics, including: distinguishing deciduous forests and evergreen forests based on a long-term NDFI value curve of vegetation; determining characteristic parameter values ​​of different tree species based on long-term change trends of INTP, SLP, and cos2 characteristics of different deciduous forest vegetation; inputting the characteristic parameter values ​​of different tree species into a classification model, and outputting the classification results of the tree species.

[0022] In some possible implementations, the method further includes: verifying the accuracy of the tree species identification results, including: evaluating the tree species identification results from two aspects: overall accuracy and kappa coefficient; wherein the overall accuracy evaluation is to compare the pixel-level tree species identification results of the long-term remote sensing image data with the ground truth to evaluate the accuracy of the classification results; the calculation formula of the overall accuracy is as follows: overall accuracy = (TP+TN) / (TP+TN+FP+FN); TP represents the number of samples that are truly positive and correctly predicted as positive, TN represents the number of samples that are truly negative and correctly predicted as negative, FP represents the number of samples that are truly negative but incorrectly predicted as positive, and FN represents the number of samples that are truly positive but incorrectly predicted as negative; the Kappa coefficient is a statistical indicator for evaluating the consistency between classifiers or evaluators, and the calculation formula of the Kappa coefficient is as follows:

[0023]

[0024] Among them, Po is the sum of the number of correctly classified samples in each category divided by the total number of samples, that is, the overall classification accuracy; assuming that the number of real samples in each category is a1, a2, ..., a C , and the predicted number of samples of each category are b1, b2, ..., b C The total number of samples is n, then:

[0025]

[0026] The calculated result of kappa is -1 to 1, but usually kappa falls between 0 and 1. It can be divided into five groups to represent different levels of consistency: 0.0 to 0.20, very low consistency (slight), 0.21 to 0.40, general consistency (fair), 0.41 to 0.60, moderate consistency (moderate), 0.61 to 0.80, high consistency (substantial), and 0.81 to 1, almost perfect.

[0027] The tree species identification method based on remote sensing images provided in the embodiment of the present application can accurately identify different tree species according to the differences in their growth rates and biomass accumulation capabilities, and can more accurately estimate the biomass in the forest; the accurate identification of different tree species helps to quantify the carbon density of each tree species, and provide more detailed data for the calculation of forest carbon reserves; the accurate identification of different tree species can help forestry managers accurately understand the composition and distribution of forest resources, improve the efficiency of the timber industry, and formulate economical and effective forestry management strategies.

[0028] In the second aspect, the embodiment of the present application provides a device for tree species identification based on remote sensing images, the device is used to execute any method as described in the first aspect, including: an image acquisition module, used to obtain long-term remote sensing forest images of the study area; the long-term remote sensing forest images include all remote sensing forest images within a set year; an NDFI curve construction module, used to calculate the NDFI value of the image band according to the long-term remote sensing forest images, and obtain the NDFI value-time series curve; the NDFI value is used to indicate the degradation and health of vegetation; a growth trajectory reconstruction module, used to reconstruct the NDFI value-time series curve of the tree species based on time harmonics to obtain the NDFI time series growth trajectory of the tree species; a feature extraction module, used to determine the tree species identification features according to the NDFI time series growth trajectory of the tree species; a tree species classification module, used to stratify and classify the tree species in the study area through a classification model based on the tree species identification features. Its beneficial effects are as described in the first aspect and will not be repeated here.

[0029] In a third aspect, an embodiment of the present application provides a computing device, comprising: at least one memory for storing a program; at least one processor for executing the program stored in the memory, and when the program stored in the memory is executed, the processor is used to execute the method provided in the first aspect. Its beneficial effects are as described in the first aspect and will not be repeated here.

[0030] In a fourth aspect, an embodiment of the present application provides a computer storage medium, in which instructions are stored, and when the instructions are executed on a computer, the computer executes the method provided in the first aspect. The beneficial effects are as described in the first aspect and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions of the multiple embodiments disclosed in this specification, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only the multiple embodiments disclosed in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] The following is a brief introduction to the drawings required for describing the embodiments or prior art.

[0033] Figure 1 A flow chart of a method for tree species identification based on remote sensing images provided in an embodiment of the present application;

[0034] Figure 2 A flowchart of a method for tree species identification based on remote sensing images is provided for an embodiment of the present application 1;

[0035] Figure 3A schematic diagram of the long-term trajectory curve of NDFI of deciduous forest and evergreen forest tree species provided in Example 1 of the present application;

[0036] Figure 4a-4c A schematic diagram of the INTP, SLP, and cos2 tree species identification characteristics of deciduous forest vegetation provided in Example 1 of the present application;

[0037] Figure 5a-5c INTP, cos2, sin2, and tree species identification characteristics of evergreen forest vegetation provided in Example 1 of the present application;

[0038] Figure 6 A device for tree species identification based on remote sensing images provided in an embodiment of the present application;

[0039] Figure 7 A computing device is provided in accordance with an embodiment of the present application. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0041] In the description of the embodiments of the present application, words such as "exemplary", "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary", "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "for example" or "for example" is intended to present related concepts in a concrete way.

[0042] In the description of the embodiments of the present application, the term "and / or" is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, B exists alone, and A and B exist at the same time. In addition, unless otherwise specified, the term "multiple" means two or more. For example, multiple systems refers to two or more systems, and multiple terminals refers to two or more terminals.

[0043] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "include", "comprises", "has" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0044] In the description of the embodiments of the present application, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it can be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0045] In the description of the embodiments of the present application, the terms "first\second\third, etc." or module A, module B, module C, etc. are only used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that the specific order or sequence can be interchanged where permitted so that the embodiments of the present application described here can be implemented in an order other than that illustrated or described here.

[0046] In the description of the embodiments of the present application, the numbers representing the steps, such as S110, S120, etc., do not necessarily mean that the steps will be executed in this manner. If permitted, the order of the previous and next steps can be interchanged, or they can be executed simultaneously.

[0047] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0048] Figure 1 This is a flow chart of a method for tree species identification based on remote sensing images provided in an embodiment of the present application. Figure 1 As shown, the specific steps are as follows:

[0049] S11, obtain long-term remote sensing forest images of the study area.

[0050] Among them, the long-term remote sensing images include all bands of the original images, which can be obtained by downloading all available remote sensing satellite images in the corresponding year of the study area from the remote sensing satellite image download platform.

[0051] For example, the remote sensing satellite image download platform can be selected from: Google Earth Engine platform (https: / / code.earthengine.google.com / ), United States Geological Survey (https: / / www.usgs.gov / ), European Space Agency (https: / / scihub.copernicus.eu / ), Geospatial Data Cloud (http: / / www.gscloud.cn / ), etc., and the remote sensing satellite image can be selected from: Landsat TM / ETM + / OLI, Sentinel-2MSI, etc.

[0052] In some embodiments, forest resource data are obtained from other sources in the study area, including forest inventory data and field survey data.

[0053] Forest inventory data mainly use forest plots as survey units, including tree species composition, age structure, tree density, canopy density, height, diameter at breast height, etc. Forest inventory data is used to gain a preliminary understanding of the vegetation types and distribution conditions in the study area, and to provide a more accurate survey scope for subsequent field surveys. Field survey data is data obtained through field surveys under the guidance of technical personnel from the local forestry bureau in areas where tree species are unclear or boundaries are fuzzy in existing forest inventory data.

[0054] For example, for complex areas where field surveys are difficult to conduct and areas where network signal sampling is difficult, aerial photography-assisted sampling can be carried out using drones equipped with multispectral sensors under the guidance of local forestry bureau technicians to obtain field survey data.

[0055] S12, calculate the NDFI value of the image band based on the long-term remote sensing forest image, and obtain the NDFI value-time series curve; the NDFI value is the normalized degradation fraction index, which is used to indicate the degradation and health status of vegetation.

[0056] In some implementations, step S12 may be implemented by the following steps:

[0057] S121, preprocessing long-term remote sensing forest images.

[0058] Among them, the preprocessing of long-term remote sensing forest images includes: time and cloud coverage screening, cloud and shadow masks, and image cropping.

[0059] Exemplarily, the preprocessing platform for satellite remote sensing images may be a commercial satellite platform, ENVI software, Sen2Cor software, SNAP software, etc.

[0060] For example, the preprocessing of aerial photography sampling by drones equipped with multispectral sensors includes: image registration, radiation correction, geometric correction, image stitching and cropping, etc., which helps to eliminate the influence of different times and places and ensure the consistency of data.

[0061] If the processing level is Collection2Level-2, it means that the product has undergone basic preprocessing such as geometric correction, radiation correction and atmospheric correction, and no repeated processing is required.

[0062] S122, calculating the NDFI value of the image band according to the forest image, and obtaining the NDFI value-time series curve of the tree species; the NDFI value is the normalized degradation fraction index, which is used to indicate the degradation and health status of vegetation.

[0063] In some embodiments, deciduous forests and evergreen forests can be distinguished based on the NDFI value curve. The NDFI value calculation formula is as follows:

[0064] NDFI=(GV-(NPV+Soil)) / (GV shade +(NPV+Soil))

[0065] GV shade =GV / (1-Shade)

[0066] Among them, GV is the green vegetation component, which indicates the proportion of vegetation that can photosynthesize in the forest image; Shade is the shadow component, which indicates the shadow component of the forest canopy in the forest image; NPV is the non-photosynthetic vegetation component, which indicates the proportion of vegetation that cannot photosynthesize or senescent vegetation in the forest image; Soil represents the soil component. The NDFI index ranges from -1 to 1. The NDFI value of lush vegetation is close to 1. When the vegetation is deciduous or disturbed, the NDFI value decreases. When the forest disappears completely, the NDFI value is close to -1.

[0067] The process of remote sensing data acquisition by remote sensors is easily affected by a series of adverse factors such as clouds, aerosols, solar altitude angles, and bidirectional reflection of ground objects, resulting in a certain amount of noise data in the remote sensing data acquired by the remote sensors, which makes the calculated NDFI raw data curve of tree species rough, not smooth, and discontinuous, and there are abnormal values ​​of sudden decline or increase. Therefore, it is necessary to reconstruct the NDFI growth trajectory curve of tree species.

[0068] S13, reconstructing the NDFI value-time series curve of the tree species based on the time series harmonics to obtain the NDFI time series growth trajectory of the tree species.

[0069] Harmonic Analysis of Time Series (HATS) is a phenological analysis method for quantitatively monitoring dynamic changes in vegetation, which can accurately describe the changing patterns of time series data.

[0070] In some implementations, step S3 can achieve NDFI value-timing curve reconstruction by the following steps:

[0071] S131, obtaining a discrete wave signal of the surface NDFI corresponding to each pixel changing with time.

[0072] S132, based on the minimum multiplication method, the selected alternative fitting is used to eliminate the points in the NDFI original data that are greatly affected by cloud pollution and have the maximum deviation from the threshold.

[0073] S133, Fourier transform is used to transform the waveform in the time domain into the frequency domain, and a series of sine waves are used to represent the NDFI fluctuation trajectory. The calculation formula is as follows:

[0074]

[0075] Where: y i is the fitting value; i is the time mark of the fitting number y; A0 is the harmonic remainder; A j is the harmonic amplitude; ω j is the harmonic frequency, N is the length of the time series; θ j is the initial phase of harmonic; m is the number of harmonics, and j is the harmonic label.

[0076] In this way, the decomposition and reconstruction of the NDFI value-time series curve can be achieved, and more reasonable data can be obtained to accurately reflect the periodic changes of vegetation.

[0077] S14, determining the tree species identification features according to the NDFI time series growth trajectory of the tree species, and the tree species identification features are used to describe the change pattern and error of the time series model.

[0078] In order to fully explore the dynamic change characteristics of the growth trajectories of different tree species, in some embodiments, step S4 includes the following steps:

[0079] S141, based on the NDFI time series trajectory, the continuous change detection and classification (CCDC) algorithm is used to construct a time series model of the growth process (leaf fall, growth, etc.) of different tree species within a year to obtain the dynamic trajectory of different tree species within a year.

[0080] The calculation formula of the CCDC time series model is as follows:

[0081]

[0082] In the formula, represents the predicted value of the i-th band of the fitted Julian date x; i represents the i-th satellite wave; T represents the number of days per year; a 0,i Represents the overall value of the reflectivity of the i-th band; a k,i 、b k,i The coefficient representing the annual variation of the reflectivity of band i; c 1,i The coefficient representing the annual variation of the reflectivity of band i, i.e., the slope (Slope, SLP).

[0083] S142, by analyzing the INTP (intercept), SLP (slope), COS, SIN, COS2, SIN2, COS3, SIN3 and other characteristics of the NDFI index of different tree species in the CCDC time series model, the temporal variation trends, characteristics and periodic changes of different tree species were determined.

[0084] Among them, intercept (INTP) and slope (SLP) are linear regression parameters used to fit the time series model, while COS, SIN, COS2, SIN2, COS3, and SIN3 are parameters used in the CCDC algorithm to describe the change pattern and error of the time series model. The meanings of each parameter are as follows:

[0085] INTP (Intercept): The intercept is a parameter in the time series model, which represents the initial state or reference state in the time series model. It reflects the starting value or reference value of the time series.

[0086] SLP (Slope): Slope is a parameter in the time series model, which indicates the rate or trend of change of the time series over time. It reflects the changing trend or growth rate of the time series.

[0087] COS, SIN, COS2, SIN2, COS3, SIN3: These parameters are used to describe the periodic changes of the time series model.

[0088] COS and SIN represent the annual variation coefficient of the curve.

[0089] COS2 and SIN2 represent the bimodal variation coefficients of the curve within the year.

[0090] COS3 and SIN3 represent the three-peak variation coefficients of the curve within the year.

[0091] The values ​​of these parameters can be used to analyze the periodic variation patterns in the time series as well as the frequency and amplitude of the variations.

[0092] S15, classifying the tree species in the study area through a classification model based on the tree species identification features.

[0093] In some implementations, a hierarchical classification strategy for tree species identification may be determined based on tree species identification features. The hierarchical classification strategy is to first classify vegetation into evergreen forests and deciduous forests, and then identify each tree species category in the evergreen forests and deciduous forests respectively.

[0094] For example, the vegetation is first divided into evergreen forest and deciduous forest, and then bamboo forest, pine, pine tree and fir are extracted from the evergreen forest to achieve hierarchical classification.

[0095] Determine the recognition threshold of tree species, and perform hierarchical classification through a classification model based on the recognition threshold and tree species recognition features.

[0096] In some embodiments, step S15 includes the following steps:

[0097] S151, determining the NDFI threshold according to the NDFI long-term trajectory curve reconstructed in S14, and distinguishing deciduous forests from evergreen forests according to the NDFI value threshold.

[0098] For example, there are significant differences in the values ​​of deciduous forest vegetation and evergreen forest vegetation on days 1 to 60, where the NDFI values ​​of deciduous forest vegetation are all less than 0, while the NDFI values ​​of evergreen forest vegetation are all greater than 0. Therefore, setting the threshold NDFI value greater than 0 on any day from 1 to 61 days can distinguish deciduous forests from evergreen forests.

[0099] S152, using the tree species identification features constructed in S15 as input variables of the classification model. The selectable classification models include random forest, support vector machine, etc. The output is the tree species identification result, including the identification of each tree species in the deciduous forest and each tree species in the evergreen forest.

[0100] In some implementations, the method for identifying tree species based on remote sensing images provided in the embodiments of the present application further includes S16, verifying the accuracy of the tree species identification results.

[0101] Accuracy verification can evaluate the tree species identification results from the overall accuracy (OA) and kappa coefficient (kappa coefficient), producer's accuracy (PA) and user's accuracy (UA). The calculation formulas for each parameter are as follows:

[0102]

[0103]

[0104]

[0105]

[0106] In the formula, h i is the number of correctly classified samples of the i-th category; X i is the total number of classified pixels of the i-th category; T i is the total number of true pixels of the i-th category; N is the total number of classified pixels.

[0107] The calculated result of kappa is -1 to 1, but usually kappa falls between 0 and 1. It can be divided into five groups to represent different levels of consistency: 0.0 to 0.20, very low consistency (slight), 0.21 to 0.40, general consistency (fair), 0.41 to 0.60, moderate consistency (moderate), 0.61 to 0.80, high consistency (substantial), and 0.81 to 1, almost perfect.

[0108] The method for tree species identification based on remote sensing images provided in the embodiment of the present application involves a method and device for tree species identification based on long-term remote sensing images, and its application scenarios include but are not limited to:

[0109] (1) Ecological and environmental monitoring: Different tree species play different roles and functions in the ecosystem. By identifying tree species, we can understand the impact of specific tree species on soil conservation, carbon cycle, water cycle, climate regulation and wildlife habitat. This helps to assess and monitor the health of the ecosystem and its contribution to human well-being.

[0110] (2) Carbon storage estimation: Tree species identification can help determine the biomass of different tree species, which is an important indicator for estimating carbon storage. The biomass of different tree species varies greatly because of their different growth rates, trunk density, crown shape and other factors. By accurately identifying tree species, the carbon storage of forests or vegetation communities can be estimated more accurately, thus providing a basis for carbon sink assessment and climate change research.

[0111] (3) Ecosystem function assessment: Combining tree species identification with biomass estimation can help assess ecosystem functions and services. Different tree species contribute differently to ecosystem functions, such as water cycle regulation, soil conservation, climate regulation, and wildlife habitat. By identifying and assessing the biomass of each tree species, we can better understand the impact of different tree species on ecosystem functions and provide a scientific basis for ecosystem management and protection.

[0112] Example 1

[0113] Figure 2 The application scenario provided for Example 1 of this application is a flow chart. Figure 2 As shown in the figure, a certain forest is taken as the research object, and there are 8 vegetation types to be identified, including chestnut, oak, catalpa, paulownia, bamboo forest, foreign pine, fir, and pine. The specific classification steps are as follows:

[0114] S21, obtaining original remote sensing image data.

[0115] Exemplarily, the original remote sensing image data all come from the Space Agency, and a total of 258 original Sentinel-2 surface reflectance image data of the study area in 2023 can be obtained through the satellite platform; the sample data come from existing forest inventory data and field survey data, of which 70% are used as training samples for training classification models, and 30% are used as verification samples for classification model accuracy verification.

[0116] S22, pre-processing of raw remote sensing image data through satellite platform, including cloud removal, terrain correction and image cropping.

[0117] Exemplarily, the cloud removal process may adopt the Simple Cloud Score algorithm; the terrain correction may adopt the SCS+C correction method of the Lambertian reflectance model based on the DEM; and the image cropping may adopt the Quality Mosaic algorithm.

[0118] S23, calculate NDFI based on satellite platform and obtain NDFI-time series curves of deciduous forest and evergreen forest tree species.

[0119] S24, reconstruct the NDFI long-term trajectory curve, and use the time harmonic analysis method to reconstruct the NDFI long-term trajectory curves of deciduous forest and evergreen forest tree species, such as Figure 3 As shown, the horizontal axis of the graph is DOY.

[0120] S25, constructing tree species identification features based on long time series trajectory of NDFI:

[0121] (1) Identification features of deciduous forest vegetation include: INTP, SLP, cos2, such as Figure 4a-4c As shown;

[0122] (2) Identification characteristics of evergreen forest vegetation include: INTP, cos2, sin2, such as Figure 5a-5c As shown;

[0123] S26, hierarchical classification based on tree species identification features, threshold of NDFI index and classification model.

[0124] Stratification included: determining tree species as deciduous or evergreen forest based on the threshold of NDFI index.

[0125] By analyzing the NDFI change curve of the long-term vegetation, it can be found that there are significant differences in the values ​​of deciduous forest vegetation and evergreen forest vegetation on days 1 to 60. The NDFI values ​​of deciduous forest vegetation are all less than 0, while the NDFI values ​​of evergreen forest vegetation are all greater than 0. Therefore, setting the NDFI value of any day from 1 to 61 days to be greater than 0 can distinguish deciduous forest from evergreen forest.

[0126] The classification involves using tree species identification features as input variables to the random forest model to identify individual tree species in deciduous and evergreen forests.

[0127] For example, for deciduous forests, by analyzing the long-term temporal variation trends of deciduous forest vegetation in INTP, SLP, and cos2 characteristics, it can be found that the INTP value of Catalpa ovata from 1 to 61 days is significantly higher than that of other deciduous forest vegetation. Therefore, the INTP feature (any day from 1 to 61 days) is used as the input variable of the random forest model to extract Catalpa ovata; in the SLP feature from 1 to 61 days, the value of Quercus acutissima is significantly higher than that of chestnut and Paulownia. Therefore, the SLP feature (any day from 1 to 61 days) is used as the input variable of the random forest model to extract Quercus acutissima; analyzing the cos2 feature of chestnut and Paulownia from 1 to 61, the cos2 value of chestnut is positive while the cos2 value of Paulownia is negative, and the two have significant differences in this feature. Therefore, the cos2 feature (any day from 1 to 61 days) is used as the input variable of the random forest model to distinguish chestnut from Paulownia.

[0128] For example, for evergreen forests, by analyzing the long-term temporal variation trends of evergreen forest vegetation in INTP, cos2, and sin2 characteristics, it can be found that the INTP value of fir from 1 to 61 days is more prominent, significantly higher than that of other evergreen forest vegetation. Therefore, the INTP feature (any day from 1 to 61 days) is used as the input variable of the random forest model to extract fir; by analyzing the variation trends of bamboo, pine, and pine COS2, it can be found that the COS2 of bamboo throughout the year is significantly higher than that of other evergreen forest tree species. Therefore, when the cos2 feature (any day from 1 to 358 days) is used as the input variable of the random forest model, the range of bamboo forest can be effectively extracted; through analysis, it is found that the difference in sin2 characteristics between pine and pine from 1 to 61 days is large, one positive and one negative, so the sin2 feature (any day from 1 to 61 days) is used as the input variable of the random forest model to distinguish pine from pine.

[0129] S27, verify the accuracy of the tree species classification results. Evaluate the tree species identification results from the perspective of overall accuracy, kappa coefficient, producer accuracy, and user accuracy.

[0130] Table 1 Tree species extraction accuracy

[0131]

[0132] The Kappa coefficient of the tree species extraction accuracy in Table 1 shows that the method of the present application has moderate consistency in the tree species classification results.

[0133] Figure 6 This is a functional block diagram of a device for tree species identification based on remote sensing images provided in an embodiment of the present application. Figure 6As shown, the device 60 is used to execute the method for tree species identification based on remote sensing images as described in any of the above embodiments, including: an image acquisition module 61, used to obtain long-term remote sensing forest images of the study area; the long-term remote sensing forest images include all remote sensing forest images within a set year; an NDFI curve construction module 62, used to calculate the NDFI value of the image band according to the long-term remote sensing forest images, and obtain the NDFI value-time series curve; the NDFI value is a normalized degradation fraction index, which is used to indicate the degradation and health status of vegetation; a growth trajectory reconstruction module 63, used to reconstruct the NDFI value-time series curve of the tree species based on time harmonics to obtain the NDFI time series growth trajectory of the tree species; a feature extraction module 64, used to determine the tree species identification features according to the NDFI time series growth trajectory of the tree species; a tree species classification module 65, used to stratify and classify the tree species in the study area through a classification model based on the tree species identification features.

[0134] Figure 7 A computing device provided in an embodiment of the present application, such as Figure 7 As shown, the computing device 70 includes: at least one memory 702 for storing programs; and at least one processor 701 for executing the programs stored in the memory. When the programs stored in the memory are executed, the processor is used to execute the method for tree species identification based on remote sensing images as described in any of the above embodiments.

[0135] An embodiment of the present application provides a computer storage medium 703, in which instructions are stored. When the instructions are executed on a computer, the computer executes the method for tree species identification based on remote sensing images as described in any of the above embodiments.

[0136] It is understood that the processor in the embodiments of the present application may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0137] The method steps in the embodiments of the present application can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, and the software modules can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an ASIC.

[0138] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions may be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid state disk (SSD)), etc.

[0139] It should be understood that the various numerical numbers involved in the embodiments of the present application are only used for the convenience of description and are not used to limit the scope of the embodiments of the present application.

Claims

1. A method for tree species identification based on remote sensing images, characterized in that: The method comprises: Acquire long-term remote sensing forest images of the study area; the long-term remote sensing forest images include all remote sensing forest images within a set year; Calculating NDFI values ​​for image bands based on the long-term remote sensing forest image to obtain an NDFI value-time series curve; the NDFI value is used to indicate the degradation and health status of vegetation; The NDFI value-time series curve is reconstructed based on the time series harmonic analysis method to obtain the NDFI time series growth trajectory, including: Obtain the discrete wave signal of the surface NDFI value corresponding to each pixel changing with time; Iterative fitting based on the minimum multiplication method eliminates the point that is most affected by cloud pollution and has the largest deviation from the threshold in the discrete wave signal to obtain a noise-reduced waveform; The Fourier transform is used to transform the noise reduction waveform in the time domain into the frequency domain, and the fitting value of the NDFI time series growth trajectory of the tree species is obtained as follows: In the formula, y i is the fitting value of the NDFI time series growth trajectory; i is the time mark of the fitting value y; A0 is the harmonic remainder; A j is the harmonic amplitude; ω j is the harmonic frequency, N is the length of the time series; θ j is the initial phase of harmonic; m is the number of harmonics, j is the harmonic mark; Determine tree species identification features according to the NDFI time series growth trajectory, wherein the tree species identification features are used to describe the change pattern and error of the time series model; The tree species in the study area are classified using a classification model based on the tree species identification features.

2. The method according to claim 1, characterized in that The acquisition of long-term remote sensing forest images of the study area includes: The long-term remote sensing forest image is preprocessed to obtain a forest growth image, wherein the preprocessing includes time and cloud amount screening, cloud and shadow masking, and image clipping.

3. The method according to claim 1 or 2, characterized in that: The NDFI value is calculated based on the image band of the long-term remote sensing forest image, including: the NDFI value calculation formula is as follows: NDFI=(GV-(NPV+Soil)) / (GV shade +(NPV+Soil)) GV shade =GV / (1-Shade) Wherein, GV is the green vegetation component, which indicates the proportion of vegetation that can photosynthesize in the forest image; Shade is the shadow component, which indicates the shadow component of the forest canopy in the forest image; NPV is the non-photosynthetic vegetation component, which indicates the proportion of vegetation that cannot photosynthesize or senescent vegetation in the forest image; Soil represents the soil component; the forest image includes a long-term remote sensing forest image or a forest growth image; On the NDFI value-time curve, the NDFI value in a certain time period is equal to 1, indicating that the vegetation in that time period is lush; the NDFI value in a certain time period decreases, indicating that the vegetation in that time period is deciduous or disturbed; the NDFI value in a certain time period is equal to -1, indicating that the forest has completely disappeared in that time period.

4. The method according to claim 1, characterized in that: Determining the tree species identification feature according to the NDFI time series growth trajectory includes: Determining tree species identification features according to the NDFI temporal growth trajectory, wherein the tree species identification features indicate temporal change trends and periodic change laws of different tree species; The tree species identification features include intercept (INTP), slope (SLP), COS, SIN, COS2, SIN2, COS3, SIN3; Wherein, the INTP (Intercept) and SLP (Slope) are linear regression parameters used to fit the time series model; the COS, SIN, COS2, SIN2, COS3, SIN3 and RMSE are parameters used to describe the change pattern and error of the time series model; The time series model is determined according to the identification characteristics of the tree species; the time series model is used to fit the growth process of different tree species, including leaf fall and growth.

5. The method according to claim 1, characterized in that The stratification and classification of the tree species in the study area by a classification model based on the tree species identification features includes: The deciduous forest and evergreen forest are distinguished based on the NDFI value curve of the long-term vegetation series; According to the long-term variation trend of INTP, SLP and cos2 characteristics of different deciduous forest vegetation, the characteristic parameter values ​​of different tree species are determined; The characteristic parameter values ​​of the different tree species are input into the classification model, and the classification results of the tree species are output.

6. The method according to claim 1, characterized in that The method further includes: verifying the accuracy of the tree species identification result, including: The tree species identification results were evaluated from the perspectives of overall accuracy and kappa coefficient; The overall accuracy evaluation is to compare the pixel-level tree species recognition results of the long-term remote sensing image data with the ground truth to evaluate the accuracy of the classification results; the calculation formula for the overall accuracy is as follows: Overall accuracy = (TP+TN) / (TP+TN+FP+FN) TP represents the number of samples that are truly positive and correctly predicted as positive, TN represents the number of samples that are truly negative and correctly predicted as negative, FP represents the number of samples that are truly negative but incorrectly predicted as positive, and FN represents the number of samples that are truly positive but incorrectly predicted as negative. The Kappa coefficient is a statistical indicator for evaluating the consistency between classifiers or evaluators. The calculation formula of Kappa coefficient is as follows: Among them, Po is the sum of the number of correctly classified samples in each category divided by the total number of samples; assuming that the number of true samples in each category is a1, a2, ..., a C , and the predicted number of samples of each category are b1, b2, ..., b C The total number of samples is n, then: The kappa calculation result is -1 to 1. When kappa falls between 0 and 1, it is divided into five groups to represent different levels of consistency: 0.0 to 0.20 very low consistency (slight), 0.21 to 0.40 general consistency (fair), 0.41 to 0.60 moderate consistency (moderate), 0.61 to 0.80 high consistency (substantial) and 0.81 to 1 almost perfect.

7. A device for identifying tree species based on remote sensing images, characterized in that: The device is used to perform the method according to any one of claims 1 to 6, comprising: An image acquisition module is used to acquire long-term remote sensing forest images of the study area; the long-term remote sensing forest images include all remote sensing forest images within a set year; An NDFI curve construction module is used to calculate the NDFI value of the image band according to the long-term remote sensing forest image to obtain an NDFI value-time series curve; the NDFI value is used to indicate the degradation and health status of vegetation; A growth trajectory reconstruction module, used to reconstruct the NDFI value-time series curve of the tree species based on a time series harmonic analysis method to obtain an NDFI time series growth trajectory; A feature extraction module, used to determine tree species identification features according to the NDFI time series growth trajectory; The tree species classification module is used to stratify and classify the tree species in the study area through a classification model based on the tree species identification characteristics.

8. A computing device, characterized in that include: at least one memory for storing a program; At least one processor is used to execute the program stored in the memory. When the program stored in the memory is executed, the processor is used to execute the method according to any one of claims 1 to 6.

9. A computer storage medium, wherein instructions are stored in the computer storage medium, and when the instructions are executed on a computer, the computer executes the method according to any one of claims 1 to 6.

Images