Forest weak disturbance area identification method, device, equipment, medium and product

Abstract

The invention discloses a forest weak disturbance area identification method and device, equipment, a medium and a product, and relates to the technical field of forest detection, and the method comprises the steps: employing a change detection algorithm of a long-time sequence image to construct a prediction model according to a normalized red wave and short wave infrared distance index and a tasseled cap transformation humidity index; constructing a training set according to the two prediction models; each sample in the training set comprises input data and label data, the input data comprises waveform coefficients and environment data of the two prediction models, and the label data comprises a forest type and a non-forest type; the forest type comprises existence of forest weak disturbance and absence of forest weak disturbance; the detection threshold value for distinguishing the existence of the forest weak disturbance and the absence of the forest weak disturbance is obtained by performing adaptive calculation on a prediction model change probability threshold value by utilizing a gradient lifting decision tree algorithm; and training the classification model by adopting the training set to obtain a forest weak disturbance recognition model. The forest weak disturbance detection accuracy is improved.

Description

Technical Field

[0001] The present application relates to the field of forest monitoring technology, and in particular to a method, device, equipment, medium and product for identifying weakly disturbed areas in forests. Background Art

[0002] With the continuous development of satellite remote sensing technology, remote sensing data has gradually become a powerful data support for forest change detection due to its fast update speed and rich information. According to the amount of remote sensing data used, disturbance change detection methods can be divided into sparse and dense time series methods. (1) Change detection methods based on sparse remote sensing data. In the past, forest change detection algorithms mainly relied on the comparison of band values ​​or vegetation index differences between two or a few remote sensing images, and judged whether changes occurred based on the size of the difference, thereby identifying forest change information. Or, the land cover type of images from different periods was first classified, and then it was judged whether the forest was converted to other land cover types, thereby determining the forest disturbance area. (2) Change detection methods based on long-term remote sensing data series. This type of algorithm arranges remote sensing feature values ​​into a data stack in chronological order, thereby constructing a pixel-by-pixel time change curve, and extracting forest disturbance event information based on the change characteristics of the curve. Typical long-term change detection algorithms include Breaks for Additive Season and Trend Monitor (BFAST), Vegetation Change Tracker (VCT), Landsat-based detection of Trends in Disturbance and Recovery (LandTrendr), Continuous Change Detection and Classification (CCDC), and Continuous monitoring of Land Disturbance (COLD).

[0003] Long-term multi-continuous change detection algorithms have significant advantages in the field of forest resource monitoring. Their dynamic analysis models based on time series spectral characteristics can effectively identify strong disturbance events in forest ecosystems (such as clear-cutting and forest fires). However, for weak disturbance events with low disturbance amplitude and small spatial scale (such as forest thinning and gradual erosion by pests and diseases), traditional algorithms still have technical bottlenecks, including the lack of sensitivity of existing feature extraction models to sub-pixel changes, limited ability to separate multi-source noise from true weak disturbance signals, and the difficulty of fixed change detection thresholds to adapt to differences in signal characteristics in different scenarios. In other words, the accuracy of weak forest disturbance detection still needs to be improved. Summary of the Invention

[0004] The purpose of this application is to provide a method, device, equipment, medium and product for identifying weak disturbance areas in forests, which can improve the accuracy of weak disturbance detection in forests.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a method for identifying weakly disturbed areas in a forest, comprising:

[0007] The first prediction model is constructed based on the normalized red wave and short wave infrared distance index using a change detection algorithm for long time series images;

[0008] The second prediction model is constructed based on the Tasseled Cap Transform Wetness Index using a change detection algorithm for long time series images.

[0009] A training set is constructed based on the first prediction model and the second prediction model; each sample in the training set includes input data and label data, the input data includes waveform coefficients and environmental data of the first prediction model and the second prediction model at a time point, and the label data includes forest type and non-forest type; the forest type includes the presence of weak forest disturbance and the absence of weak forest disturbance; wherein the detection threshold for distinguishing the presence of weak forest disturbance from the absence of weak forest disturbance is obtained by adaptively calculating the change probability threshold of the first prediction model and the second prediction model using a gradient boosting decision tree algorithm;

[0010] The training set is used to train a classification model to obtain a forest weak disturbance recognition model;

[0011] The waveform coefficients and environmental data of the first prediction model and the second prediction model corresponding to the target area at the time point to be detected are input into the forest weak disturbance recognition model to obtain the forest weak disturbance recognition result.

[0012] Optionally, the normalized red wave and short wave infrared distance index is expressed as:

[0013]

[0014] Among them, NDRS is the normalized red wave and short wave infrared distance index, DRS is the red wave and short wave infrared distance index, max For the 95% percentile of DRS, DRS min is the 5% percentile DRS, RED is the surface reflectance of the red band of the remote sensing image, and SWIR1 is the surface reflectance of the shortwave infrared 1 band.

[0015] Optionally, the Tasseled Cap Transformed Wetness Index is expressed as:

[0016] TCW=0.1511ρ1+0.1973ρ2+0.3283ρ3+0.3407ρ4-0.7117ρ5-0.4559ρ6;

[0017] Among them, TCW is the Tasseled Cap Transform Wetness Index, and ρ1, ρ2, ρ3, ρ4, ρ5, and ρ6 represent the surface reflectance of bands 2, 3, 4, 5, 6, and 7 of the Landsat-8 remote sensing image, respectively.

[0018] Optionally, the first prediction model or the second prediction model is expressed as:

[0019]

[0020] in, represents the predicted value of the i-th target index at time point x, where the target index is the normalized red wave and shortwave infrared distance index or the tasseled cap transform humidity index, a 0,i represents the overall coefficient of the target index, a s,i and b s,i represents the annual change coefficient of the target index, c 1,i represents the interannual coefficient of the target index, s represents the harmonic component of the time frequency, and T is a constant.

[0021] Optionally, the classification model adopts a random forest module.

[0022] Optionally, the environmental data includes topography, temperature and precipitation.

[0023] In a second aspect, the present application provides a device for identifying a forest weakly disturbed area, wherein the device applies any of the above-mentioned methods for identifying a forest weakly disturbed area, and the device comprises:

[0024] A first prediction model building module is used to build a first prediction model using a change detection algorithm of long time series images according to normalized red wave and short wave infrared distance indices;

[0025] A second prediction model building module is used to build a second prediction model based on the Tasseled Cap Transform Wetness Index using a change detection algorithm for long time series images;

[0026] A training set construction module is configured to construct a training set based on the first prediction model and the second prediction model; each sample in the training set includes input data and label data, the input data including waveform coefficients and environmental data of the first prediction model and the second prediction model at a time point, the label data including forest type and non-forest type; the forest type includes the presence of weak forest disturbance and the absence of weak forest disturbance; wherein the detection threshold for distinguishing the presence of weak forest disturbance from the absence of weak forest disturbance is obtained by adaptively calculating the change probability threshold of the first prediction model and the second prediction model using a gradient boosting decision tree algorithm;

[0027] A model training module is used to train a classification model using the training set to obtain a forest weak disturbance recognition model;

[0028] The forest weak disturbance identification module is used to input the waveform coefficients and environmental data of the first prediction model and the second prediction model corresponding to the target area to be detected into the forest weak disturbance identification model to obtain the forest weak disturbance identification result.

[0029] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any one of the above-described methods for identifying weakly disturbed forest areas.

[0030] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of the above-mentioned methods for identifying weakly disturbed areas in forests.

[0031] In a fifth aspect, the present application provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of any one of the above-mentioned methods for identifying weakly disturbed areas in forests.

[0032] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0033] The present application provides a method, device, equipment, medium and product for identifying forest weak disturbance areas. According to the Normalized Distance Red and SWIR (NDRS) and Tasselled Cap Wetness (TCW), a prediction model is constructed using a continuous monitoring of land disturbance (COLD) algorithm based on a long time series image. NDRS can effectively reflect the health of vegetation, and TCW can reflect the moisture content in vegetation, thereby providing effective basic data for improving forest weak disturbance detection. During the model training process, the input data includes the waveform coefficients and environmental data of the first prediction model and the second prediction model at a time point, and the label data includes forest type and non-forest type; the forest type includes the presence of forest weak disturbance and the absence of forest weak disturbance; and the detection threshold for distinguishing the presence of forest weak disturbance and the absence of forest weak disturbance is a gradient boosting decision tree algorithm (Gradient Boosting Decision Tree Algorithm). DecisionTree, GBDT) performs adaptive calculation of the change probability threshold of the first prediction model and the second prediction model, thereby obtaining the forest weak disturbance detection threshold that is most suitable for the target area, realizing effective detection of forest weak disturbances, and the present application can improve the accuracy of forest weak disturbance detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0035] Figure 1 A flowchart of a method for identifying weakly disturbed areas in a forest provided in one embodiment of the present application;

[0036] Figure 2 Schematic diagram of forest disturbance before and after according to one embodiment of the present application;

[0037] Figure 3 A schematic diagram comparing forest disturbance detection using different identification methods provided in one embodiment of the present application;

[0038] Figure 4 Schematic diagram of a forest before and after a strong disturbance provided in one embodiment of the present application;

[0039] Figure 5 for Figure 4 Corresponding waveform diagram obtained using the COLD algorithm;

[0040] Figure 6 Schematic diagram of a forest before and after weak disturbance provided in one embodiment of the present application;

[0041] Figure 7 for Figure 6 Corresponding waveform diagram obtained using the COLD algorithm;

[0042] Figure 8 A schematic diagram comparing detection accuracy for different change probabilities p provided in one embodiment of the present application;

[0043] Figure 9 Schematic diagram of forest weak disturbance (insect plague) patches detected with different change probabilities p provided in one embodiment of the present application;

[0044] Figure 10 A schematic diagram of the functional modules of a device for identifying weakly disturbed areas in a forest provided in one embodiment of the present application;

[0045] Figure 11 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0047] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0048] In an exemplary embodiment, the present application provides a method for identifying weakly disturbed areas in a forest, such as Figure 1 As shown, the method for identifying weakly disturbed forest areas includes:

[0049] Step 101: Construct a first prediction model using a change detection algorithm of long time series images according to the normalized red wave and short wave infrared distance index.

[0050] Step 102: Construct a second prediction model based on the Tasseled Cap Transform Wetness Index using a change detection algorithm for long time series images.

[0051] Step 103: Construct a training set based on the first prediction model and the second prediction model; each sample in the training set includes input data and label data, the input data includes waveform coefficients and environmental data of the first prediction model and the second prediction model at a time point, and the label data includes forest type and non-forest type; the forest type includes the presence of weak forest disturbance and the absence of weak forest disturbance; wherein, the detection threshold for distinguishing the presence of weak forest disturbance and the absence of weak forest disturbance is obtained by adaptively calculating the change probability threshold of the first prediction model and the second prediction model using the gradient boosting decision tree algorithm.

[0052] Step 104: Use the training set to train a classification model to obtain a forest weak disturbance recognition model.

[0053] Step 105: Input the waveform coefficients and environmental data of the first prediction model and the second prediction model corresponding to the target area at the time point to be detected into the forest weak disturbance recognition model to obtain a forest weak disturbance recognition result.

[0054] Since forest weak disturbances have low amplitude changes and small patch areas, forest weak disturbance detection algorithms based on long time series are usually difficult to identify, or difficult to distinguish them from noise in the results. Therefore, this application discloses a precise identification method for forest weak disturbance areas, the main contents of which include: (1) First, the normalized red wave and short wave infrared distance index and the tassel cap change humidity index are selected as the input data for long time series continuous change detection, and the change detection algorithm of long time series images is used to carry out change detection. (2) Then, the random forest algorithm (RF) is used to classify and train the waveform coefficients in COLD to obtain the forest and non-forest classification of each waveform segment, thereby removing strong forest disturbances and noise and retaining the detection results of weak forest disturbances. (3) Finally, the gradient boosting decision tree algorithm (GBDT) is used to adaptively calculate the COLD disturbance intensity threshold to obtain the most suitable forest weak disturbance detection threshold, ensuring the accurate identification of forest weak disturbance events.

[0055] In an exemplary embodiment, the present application is different from the existing COLD which uses the band reflectance of remote sensing images as input data for long-term continuous change detection. The present application uses normalized red wave and shortwave infrared distance index and tasseled cap transform humidity index as input data to carry out change detection.

[0056] The normalized red wave and shortwave infrared distance indices have better detection effects on vegetation health.

[0057] The normalized red wave and short wave infrared distance index is expressed as:

[0058]

[0059] Among them, NDRS is the normalized red wave and short wave infrared distance index, DRS is the red wave and short wave infrared distance index, max For the 95% percentile of DRS, DRS min is the 5% percentile DRS, RED is the surface reflectance of the red band of the remote sensing image, and SWIR1 is the surface reflectance of the shortwave infrared 1 band.

[0060] The tasseled cap wetness index is a multispectral linear transformation. The transformed wetness index can reflect the moisture content in vegetation.

[0061] The Tasseled Cap Transformed Wetness Index is expressed as:

[0062] TCW=0.1511ρ1+0.1973ρ2+0.3283ρ3+0.3407ρ4-0.7117ρ5-0.4559ρ6 (3)

[0063] Among them, TCW is the Tasseled Cap Transform Wetness Index, and ρ1, ρ2, ρ3, ρ4, ρ5, and ρ6 represent the surface reflectance of bands 2, 3, 4, 5, 6, and 7 of the Landsat-8 remote sensing image, respectively.

[0064] The principle of the change detection algorithm based on long time series images is as follows: First, a time series prediction model is constructed using existing stable observations (Equation (4)). Newly collected clear observations (NDRS or TCW) are used to iteratively update the current prediction model. Second, for each newly collected clear observation, the model prediction value is compared with the actual observation value. If the change range is within the expected range, the next observation value is considered stable and used to iteratively update the model. Otherwise, the observation value is considered to have changed. A stable observation sequence generates a time series prediction model, and a new time series prediction model is generated after a disturbance occurs.

[0065] The first prediction model or the second prediction model is expressed as:

[0066]

[0067] in, represents the predicted value of the i-th target index at time point x, where the target index is the normalized red wave and shortwave infrared distance index or the tasseled cap transform humidity index, a 0,i represents the overall coefficient of the target index, a s,i and b s,i represents the annual change coefficient of the target index, c 1,i represents the interannual coefficient of the target index, s represents the harmonic component of the time frequency (s = 1, 2, 3), and T is a constant.

[0068] In an exemplary embodiment, x represents a date, specifically the Julian day, and T is the number of days in a year (which is 365.25).

[0069] The waveform coefficients of the first prediction model or the second prediction model are fitted using the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm. The LASSO algorithm can reduce overfitting by minimizing the sum of squared residuals and the sum of absolute values ​​of coefficients (Equation (5)). The first term is the squared loss function, which measures the model fitting error. The second term is the L1 regularization term, and λ is the regularization parameter. A higher λ value means a greater penalty for overfitting.

[0070]

[0071] Among them, β is the coefficient vector composed of waveform coefficients, β j is the jth waveform coefficient, β∈R p Indicates that the coefficient vector β is a p-dimensional real vector, i is the target index, X i is the design matrix of the i-th target index, Y i is the response variable vector of the i-th target index, and λ is used to control the strength of regularization and is usually set to 20.

[0072] The change threshold is the most critical variable in change detection. The COLD algorithm uses the normalized change vector amplitude as the change detection threshold. By using the sum of squared deviations, changes in any band will show a large deviation. Since the standardized change deviations follow a standard normal distribution, the sum of squared deviations follows a chi-square distribution with degrees of freedom equal to the number of bands used for change detection. Therefore, a change threshold can be created based on the change probability derived from the chi-square distribution. The COLD algorithm finds that the best detection results are achieved when the change probability is 0.99 and five bands are input. The change probability calculation is shown in Equation (6).

[0073]

[0074] Where i is the i-th target index, k is the number of target indices used for change detection, and ρ i is the observed value of the i-th target index, is the predicted value of the i-th target index, χ 2 (k) represents the chi-square distribution with k degrees of freedom, χ 0.99 is the probability of change, and RMSE represents the root mean square error.

[0075] In an exemplary embodiment, the classification model uses a Random Forest (RF) module. The RF algorithm is used to classify and train the waveform coefficients in COLD to obtain forest and non-forest classifications for each waveform segment, removing strong forest disturbances and noise while retaining weak forest disturbance detection results.

[0076] By utilizing all waveform coefficients in the COLD algorithm (mean amplitude of vegetation index, slope, third harmonic coefficient and root mean square error) and combining environmental data such as terrain, temperature and precipitation, forest and non-forest labels are extracted based on visual discrimination, and the RF algorithm is used for classification training to obtain the forest / non-forest type of each waveform in the COLD.

[0077] Strong forest disturbances are characterized by being forest before the change and non-forest after the change, with a decreased vegetation index. Weak forest disturbances are characterized by being forest both before and after the change, with a decreased vegetation index. Noise is characterized by uncertainty about the type of land cover before and after the change, and uncertainty about the change in vegetation index. Therefore, based on the land cover type before and after the disturbance, we selected points where both the forest type and the vegetation index decreased before and after the disturbance as weak forest disturbances. The RF algorithm parameter settings are shown in Table 1.

[0078] Table 1 RF algorithm parameter settings

[0079] parameter This application sets the value Number of decision trees 150 Number of variables per split Square root of the number of variables Fraction of the bag that each tree enters 0.5 The maximum number of leaf nodes in each tree No restrictions

[0080] This application can eliminate noise characteristics based on vegetation index changes and obtain preliminary weak disturbance results. However, the inherent change probability p has different application effects in different test areas, and the p value needs to be adaptively set according to the actual situation. Therefore, based on training data consisting of forest weak disturbance and healthy vegetation sample points, a gradient boosting decision tree algorithm is used to adaptively calculate the COLD disturbance intensity threshold, and the most suitable forest weak disturbance detection threshold, namely the change probability p, is obtained.

[0081] The existing COLD algorithm uses an empirically set threshold (the probability of change, p, is typically set to 0.99). Consequently, its application performance varies across different test areas, resulting in weak generalization. To improve the COLD algorithm's performance in different application scenarios, we visually identified weakly disturbed forests and healthy vegetation sample points based on the initial weakly disturbed identification zones. We then tested the COLD algorithm's ability to detect weakly disturbed forests when the probability of change, p, varied between 0.05 and 0.95. We then gradually modified the p-value in steps of 0.05, calculating the weakly disturbed detection accuracy using the GBDT algorithm. The p-value corresponding to the highest weakly disturbed detection accuracy was used as the final probability of change. Adaptive calculation of the probability of change using the GBDT algorithm improves the algorithm's generalization across different test areas. The GBDT algorithm parameter settings are shown in Table 2.

[0082] Table 2 GBDT algorithm parameter settings

[0083] parameter This application sets the value Number of decision trees 100 Learning rate 0.005 Random numbers increase sampling rate 0.7 The maximum number of leaf nodes in each tree No restrictions Regression loss function Least Absolute Deviation Function

[0084] The above improved algorithm for identifying weak forest disturbances based on long-term remote sensing images takes NDRS and TCW as input, uses the COLD algorithm to perform change detection, removes strong forest disturbances and noise patches from the detected change patch results, and further uses GBDT to adaptively select the optimal change probability p. This p value is used in COLD to perform global forest weak disturbance detection, which can obtain high-precision forest weak disturbance detection results.

[0085] In an exemplary embodiment, Figure 2 (a) and (b) are the effect diagrams before and after the disturbance respectively. Figure 3 For Figure 2 The effect diagram of different algorithms for detecting weak disturbance in the forest. Figure 3 (a) shows the effect of detection using BFAST, (b) shows the effect of detection using LandTrendr, and (c) shows the effect of detection using the COLD algorithm based on this application. By using NDRS and TCW as inputs to the COLD algorithm, this application achieves higher accuracy in identifying weakly disturbed areas of forest compared to other change detection algorithms such as BFAST and LandTrendr. The overall classification accuracy of the improved COLD algorithm reaches 85.57%, while the overall classification accuracy of BFAST and LandTrendr is 76.83% and 58.54%, respectively. The NDRS index can reflect the health of vegetation, and the TCW index reflects the water content of vegetation. Using this combination as input data for long-term continuous change detection can identify vegetation with weak disturbance events.

[0086] Figure 4 This is a schematic diagram of strong forest disturbance. Figure 5 This is a schematic diagram of the waveform changes corresponding to strong forest disturbance. Figure 6 This is a schematic diagram of strong forest disturbance. Figure 7 This is a schematic diagram of the waveform change corresponding to the weak disturbance of the forest. Figure 4-Figure 7It can be seen that when the waveform is severely deformed, it indicates a strong forest disturbance, while weak forest disturbances cannot be observed between waveform changes. According to the COLD algorithm principle, when forest disturbances occur, breakpoints will appear in the long time series observation sequence. However, the occurrence of breakpoints includes several different situations, such as strong forest disturbances, weak forest disturbances, and other noise. Therefore, weak forest disturbances need to be screened and distinguished from strong forest disturbances and noise based on their characteristics. Strong forest disturbances are characterized by: forest before the change, non-forest after the change, and a decrease in vegetation index; weak forest disturbances are characterized by: forest before and after the change, and a decrease in vegetation index; noise is characterized by uncertain type before and after the change, and uncertain changes in vegetation index. By using RF to classify the COLD algorithm waveforms, the weak forest disturbances can be identified.

[0087] The change probability of the COLD algorithm is usually an empirical value. The GBDT algorithm can adaptively optimize the change probability and select the change probability p corresponding to the highest detection accuracy based on a small sample, such as Figure 8 As shown. GBDT is an ensemble learning algorithm that optimizes the value of the loss function by iteratively adding new decision trees. Each decision tree learns based on all previous trees, thereby reducing the accumulation of residuals. Through experiments, it can be seen that in the test area of ​​a certain forest farm, such as Figure 9 As shown, Figure 9 (a) is the forest area to be detected, and (b)-(g) are the forest weak disturbance (insect plague) patches corresponding to different change probabilities p. When the change probability p is 0.6, the detection accuracy of forest weak disturbance is the highest (84.58%), which is 15.45% higher than the accuracy when p is the empirical value of 0.95 (69.13%). The forest weak disturbance recognition result of this application includes the forest weak disturbance recognition result of each pixel point, and the forest weak disturbance patch is obtained based on the forest weak disturbance recognition result of each pixel point.

[0088] Based on the same inventive concept, embodiments of the present application also provide a device for identifying weakly disturbed forest areas, for implementing the aforementioned method for identifying weakly disturbed forest areas. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the device for identifying weakly disturbed forest areas provided below can be found in the limitations of the method for identifying weakly disturbed forest areas above, and will not be further elaborated here.

[0089] In an exemplary embodiment, Figure 10 As shown, a device for identifying a forest weak disturbance area is provided, comprising:

[0090] The first prediction model construction module is used to construct a first prediction model using a change detection algorithm of long time series images according to normalized red wave and short wave infrared distance index.

[0091] The second prediction model construction module is used to construct a second prediction model based on the Tasseled Cap Transform Wetness Index using a change detection algorithm for long time series images.

[0092] A training set construction module is used to construct a training set based on the first prediction model and the second prediction model; each sample in the training set includes input data and label data, the input data includes the waveform coefficients and environmental data of the first prediction model and the second prediction model at a time point, and the label data includes forest type and non-forest type; the forest type includes the presence of weak forest disturbance and the absence of weak forest disturbance; wherein, the detection threshold for distinguishing the presence of weak forest disturbance and the absence of weak forest disturbance is obtained by adaptively calculating the change probability threshold of the first prediction model and the second prediction model using the gradient boosting decision tree algorithm.

[0093] The model training module is used to train the classification model using the training set to obtain a forest weak disturbance recognition model.

[0094] The forest weak disturbance identification module is used to input the waveform coefficients and environmental data of the first prediction model and the second prediction model corresponding to the target area to be detected into the forest weak disturbance identification model to obtain the forest weak disturbance identification result.

[0095] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 11 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store forest weak disturbance area identification data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a forest weak disturbance area identification method is implemented.

[0096] Those skilled in the art will understand that Figure 11The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present application and does not constitute a limitation on the computer device to which the solution of the present application is applied. A specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps of the above-mentioned method embodiments when executing the computer program.

[0097] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0098] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0099] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0100] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0101] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, data processing logic of programmable logic devices, and the like.

[0102] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0103] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for identifying weakly disturbed areas in a forest, characterized in that: The method for identifying forest weak disturbance areas includes: The first prediction model is constructed based on the normalized red wave and short wave infrared distance index using a change detection algorithm for long time series images; The second prediction model is constructed based on the Tasseled Cap Transform Wetness Index using a change detection algorithm for long time series images. A training set is constructed based on the first prediction model and the second prediction model; each sample in the training set includes input data and label data, the input data includes waveform coefficients and environmental data of the first prediction model and the second prediction model at a time point, and the label data includes forest type and non-forest type; the forest type includes the presence of weak forest disturbance and the absence of weak forest disturbance; wherein the detection threshold for distinguishing the presence of weak forest disturbance from the absence of weak forest disturbance is obtained by adaptively calculating the change probability threshold of the first prediction model and the second prediction model using a gradient boosting decision tree algorithm; The training set is used to train a classification model to obtain a forest weak disturbance recognition model; The waveform coefficients and environmental data of the first prediction model and the second prediction model corresponding to the target area at the time point to be detected are input into the forest weak disturbance recognition model to obtain the forest weak disturbance recognition result.

2. The method for identifying weakly disturbed forest areas according to claim 1, characterized in that: The normalized red wave and short wave infrared distance index is expressed as: Among them, NDRS is the normalized red wave and short wave infrared distance index, DRS is the red wave and short wave infrared distance index, max For the 95% percentile of DRS, DRS min is the 5% percentile DRS, RED is the surface reflectance of the red band of the remote sensing image, and SWIR1 is the surface reflectance of the shortwave infrared 1 band.

3. The method for identifying weakly disturbed forest areas according to claim 1, wherein: The Tasseled Cap Transformed Wetness Index is expressed as: TCW=0.1511ρ1+0.1973ρ2+0.3283ρ3+0.3407ρ4-0.7117ρ5-0.4559ρ6; Among them, TCW is the Tasseled Cap Transform Wetness Index, and ρ1, ρ2, ρ3, ρ4, ρ5, and ρ6 represent the surface reflectance of bands 2, 3, 4, 5, 6, and 7 of the Landsat-8 remote sensing image, respectively.

4. The method for identifying weakly disturbed forest areas according to claim 1, wherein: The first prediction model or the second prediction model is expressed as: in, represents the predicted value of the i-th target index at time point x, where the target index is the normalized red wave and shortwave infrared distance index or the tasseled cap transform humidity index, a 0,i represents the overall coefficient of the target index, a s,i and b s,i Indicates the annual change coefficient of the target index, c 1,i represents the interannual coefficient of the target index, s represents the harmonic component of the time frequency, and T is a constant.

5. The method for identifying weakly disturbed forest areas according to claim 1, wherein: The classification model adopts the random forest module.

6. The method for identifying weakly disturbed forest areas according to claim 1, wherein: The environmental data includes topography, temperature and precipitation.

7. A device for identifying weakly disturbed areas in a forest, characterized in that: The forest weakly disturbed area identification device applies the forest weakly disturbed area identification method according to any one of claims 1 to 6, and the forest weakly disturbed area identification device comprises: A first prediction model building module is used to build a first prediction model using a change detection algorithm of long time series images according to normalized red wave and short wave infrared distance indices; A second prediction model building module is used to build a second prediction model based on the Tasseled Cap Transform Wetness Index using a change detection algorithm for long time series images; A training set construction module is configured to construct a training set based on the first prediction model and the second prediction model; each sample in the training set includes input data and label data, the input data including waveform coefficients and environmental data of the first prediction model and the second prediction model at a time point, the label data including forest type and non-forest type; the forest type includes the presence of weak forest disturbance and the absence of weak forest disturbance; wherein the detection threshold for distinguishing the presence of weak forest disturbance from the absence of weak forest disturbance is obtained by adaptively calculating the change probability threshold of the first prediction model and the second prediction model using a gradient boosting decision tree algorithm; A model training module is used to train a classification model using the training set to obtain a forest weak disturbance recognition model; The forest weak disturbance identification module is used to input the waveform coefficients and environmental data of the first prediction model and the second prediction model corresponding to the target area to be detected into the forest weak disturbance identification model to obtain the forest weak disturbance identification result.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for identifying weakly disturbed areas in forests according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for identifying weakly disturbed areas in forests according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method for identifying weakly disturbed areas in forests according to any one of claims 1 to 6 is implemented.

Images