Intelligent identification and extraction method and equipment for felling / burned area and medium

Abstract

The invention discloses a felling / burned area intelligent identification and extraction method and device and a medium, and relates to the technical field of felling / burned area intelligent identification, and the method comprises the steps: carrying out the NaN value processing of multispectral information through ENVI 5.3, obtaining a plurality of wave band features, and carrying out the fusion of the features, and obtaining a fusion image; carrying out wave band operation on the fused image, and extracting a plurality of vegetation indexes; constructing a digital surface model by utilizing unmanned aerial vehicle image processing software based on the multispectral information; based on the fused image and the digital surface model, constructing a plurality of dotted samples corresponding to different surface feature types; determining a waveband fusion feature corresponding to each point-shaped sample; training a support vector machine model by using the dotted samples; and determining an initial classification result of the to-be-classified image by using the classification model. On the basis of the multispectral information and the support vector machine model, the recognition accuracy is improved while the recognition real-time performance of the felling / burned area is ensured.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent identification of felled / burned areas, and in particular to a method, device and medium for intelligent identification and extraction of felled / burned areas. Background Art

[0002] According to statistics, the global forest cover rate is approximately 32%. Increasing forest cover is crucial for maintaining ecological balance, improving environmental quality, protecting biodiversity, and addressing climate change. However, the frequent occurrence of extreme weather events worldwide has led to a severe situation with forest fires increasing in both size and intensity. Forest cover loss due to forest fires is increasing by approximately 5.4% annually, and nearly 6 million more hectares of forest cover are now lost annually to fires compared to 2001. Climate change is one of the main drivers of increased fire activity. When forests burn, they release carbon stored in tree trunks, branches, and leaves, as well as in the soil. As forest fires become larger and more frequent, they release more carbon, further exacerbating climate change and leading to more fires through a fire-climate feedback loop. Another major factor contributing to the increase in fire activity is human activity. Unregulated or illegal deforestation and forest degradation have led to higher temperatures and drying vegetation, creating more fuel and enabling fires to spread faster. Forest damage caused by fire or logging not only results in loss of forest resources, but also causes air pollution, induces serious soil erosion, and changes regional climate.

[0003] Felled / burned areas are empty spaces where vegetation has not been replanted after deforestation or natural or artificial fires. Felled / burned areas can significantly change regional runoff and erosion conditions. Furthermore, some felled areas are burned, which reduces soil infiltration rates, water storage capacity, and soil resistance to erosion, further exacerbating soil and water loss in forests. Therefore, intelligent, rapid, and accurate identification and extraction of felled / burned areas, and strengthening supervision of unauthorized and illegal felled / burned areas and areas beyond the approved felling scope, are of great significance for preventing and controlling soil and water loss in forests. These are also important prerequisites for subsequent vegetation restoration and forest development and planting. Currently, most felled / burned area identification and extraction methods are based on satellite remote sensing imagery, making it difficult to obtain real-time, high-resolution images of the supervised areas, making it impossible to conduct real-time, dynamic supervision. UAV remote sensing can meet the requirements in terms of time and spatial resolution, but visible light UAVs mainly obtain visible light and terrain information of ground objects, and cannot obtain multispectral band information of ground objects. The main feature of felled / burned areas is a significant reduction in vegetation cover, which can be distinguished by multispectral bands. Therefore, it is necessary to combine multispectral UAVs and supervised classification technology to carry out rapid and accurate identification, extraction and evaluation of felled / burned areas.

[0004] Application Contents

[0005] The purpose of this application is to provide a method, device and medium for intelligent identification and extraction of felled / burned areas, which can improve the identification accuracy while ensuring the real-time identification of felled / burned areas.

[0006] To achieve the above objectives, this application provides the following solutions:

[0007] In a first aspect, the present application provides a method for intelligently identifying and extracting felled / burned areas, comprising:

[0008] Obtain multispectral information of logged / burned areas;

[0009] Based on the NaN values corresponding to different bands, ENVI 5.3 was used to process the NaN values of the multispectral information to obtain multiple band features;

[0010] Fuse the features of multiple bands to obtain a fused image;

[0011] performing band operations on the fused image to extract multiple vegetation indices;

[0012] Based on the multispectral information, a digital surface model is constructed using drone image processing software;

[0013] constructing a plurality of point samples corresponding to different types of land features based on the fused image and the digital surface model;

[0014] The multiple band features and multiple vegetation indices corresponding to each point sample are band-fused to determine the band-fusion features corresponding to each point sample;

[0015] With the band fusion features as input and the ground feature types as output, the support vector machine model is trained to obtain the classification model;

[0016] Determine the band fusion features of each point in the image to be classified;

[0017] Based on the band fusion feature of each point in the image to be classified, the classification model is used to classify the ground objects in the image to be classified to obtain an initial classification result.

[0018] Optionally, the multispectral information is acquired using an unmanned aerial vehicle equipped with a multispectral imager.

[0019] Optionally, the wavelength bands include: a near infrared wavelength band, a red edge wavelength band, a red light wavelength band and a green light wavelength band.

[0020] Optionally, the vegetation index includes: green normalized vegetation index, leaf chlorophyll index, normalized difference red edge index, normalized vegetation index and optimized soil adjusted vegetation index.

[0021] Optionally, the green normalized difference vegetation index is:

[0022] GNDVI=(NIR-Green) / (NIR+Green);

[0023] The leaf chlorophyll index is:

[0024] LCI=(NIR-Red edge) / (NIR+Red);

[0025] The normalized difference red edge index is:

[0026] NDRE=(NIR-Red edge) / (NIR+Red edge);

[0027] The normalized difference vegetation index is:

[0028] NDVI=(NIR–Red) / (NIR+Red);

[0029] The optimized soil-adjusted vegetation index is:

[0030] OSAVI=(NIR-Red) / (NIR+Red+0.16);

[0031] Where, GNDVI is the green normalized vegetation index;

[0032] LCI is the leaf chlorophyll index; NDRE is the normalized difference red edge index; NDVI is the normalized vegetation index; OSAVI is the optimized soil adjusted vegetation index; NIR is the near-infrared band characteristics; Red is the red light band characteristics; Green is the green light band characteristics; Red edge is the red edge band characteristics.

[0033] Optionally, the drone image processing software is: Agisoft PhotoscanProfessional.

[0034] Optionally, the types of land features include: felled / burned land, roads and woodlands.

[0035] Optionally, after classifying the ground objects in the image to be classified using the classification model based on the band fusion feature of each point in the image to be classified and obtaining an initial classification result, the method further includes:

[0036] The MajorityAnalysis tool in ENVI 5.3 was used to post-process the initial classification results to obtain classification results.

[0037] In a second 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 above-mentioned method for intelligent identification and extraction of felled / burned areas.

[0038] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned method for intelligent identification and extraction of felled / burned areas.

[0039] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0040] The present application provides a method, device and medium for intelligent identification and extraction of felled / burned areas. The multispectral band information of the supervised area is obtained through drone images. Based on the main feature that felled / burned areas have less vegetation coverage than normal forest land, multiple vegetation indices of the supervised area are calculated and extracted. The supervised classification method of the support vector machine algorithm is used to achieve fast and accurate extraction of felled / burned areas.

[0041] This application makes a quantitative evaluation of the classification results, integrating ground data collection, feature extraction, classification extraction, and quality assessment in the felling / burnt site supervision area to form an intelligent method, realizing real-time dynamic supervision of the felling / burned site monitoring area. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] 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.

[0043] Figure 1 This is a flow chart of a method for intelligently identifying and extracting felled / burned areas in one embodiment of the present application;

[0044] Figure 2 This is a schematic diagram of a method for intelligently identifying and extracting felled / burned areas in one embodiment of the present application;

[0045] Figure 3 A schematic diagram of processing drone multispectral images to generate a digital surface model DSM in one embodiment of the present application;

[0046] Figure 4 A schematic diagram of extracting vegetation indices (e.g., NDVI) by performing band operations on multispectral images in one embodiment of the present application;

[0047] Figure 5A schematic diagram of creating a region of interest using ROI in one embodiment of the present application, including three types of point samples: tracks, roads, and woodlands;

[0048] Figure 6 Schematic diagram of support vector machine supervised classification according to land use types of ruins, woodlands and roads in one embodiment of the present application;

[0049] Figure 7 This is a schematic diagram of post-processing the initial classification results to remove small plaques in one embodiment of the present application;

[0050] Figure 8 A confusion matrix report for verifying the accuracy of classification results in one embodiment of the present application;

[0051] Figure 9 This is a schematic diagram of the deforestation site finally extracted in one embodiment of the present application. DETAILED DESCRIPTION

[0052] 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.

[0053] 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.

[0054] In an exemplary embodiment, Figure 1 and Figure 2 As shown, a method for intelligent identification and extraction of felled / burned areas is provided, comprising:

[0055] Step 101: Acquire multispectral information of a felled / burned area.

[0056] Use a drone equipped with a multispectral imaging device to capture aerial images of the logged / burned areas, obtaining multispectral information. Specifically, pre-set the drone's flight path and altitude based on the logging / burned area monitoring scope, required accuracy, and scale. Use the drone's multispectral imaging device to capture aerial images of the soil and water conservation monitoring area. Before takeoff, set flight parameters to ensure a heading overlap of greater than 60% and a lateral overlap of greater than 30%.

[0057] Step 102: Based on the NaN values corresponding to different bands, ENVI 5.3 is used to process the multispectral information and obtain multiple band features. The bands include: near infrared band, red edge band, red light band, and green light band.

[0058] Process the NaN values of the multispectral bands in the cleared area and change them to 0. The specific steps are as follows:

[0059] (1) Modify NaN to a specific value (-999): perform operations on the band using the formula "finite(b1, / nan)*(-999)or(~finite(b1, / nan))*b1".

[0060] (2) Further modify the band modified in (1) to 0: perform calculation on the band, and the calculation formula is "(b1 ne-999)*b1".

[0061] The code reference URL is https: / / www.cnblogs.com / enviidl / p / 16267337.html or https: / / www.cnblogs.com / enviidl / p / 16280290.html. finite(b1, / nan)*(-999) means that if the pixel value is a valid value (not NaN), it returns 0; if it is NaN, it returns -999. ~ represents negation, and or represents OR. (~finite(b1, / nan))*b1 means that if the pixel value is a valid value (not NaN), it returns b1; if it is NaN, it returns 0. ne represents inequality. This operation evaluates each pixel in the band b1 modified in the previous step: if the pixel value is not equal to -999, it returns 1; if it is equal to -999, it returns 0. Then, multiplying by the band b1, the NaN value in step 1 becomes 0 after being modified to a specific value (-999).

[0062] Step 103: Fuse the features of multiple bands to obtain a fused image.

[0063] Step 104: Perform band operations on the fused image to extract multiple vegetation indices; the extraction results are as follows: Figure 4 .

[0064] Band operations were performed on the fused data (fused image) to extract five vegetation indices: GNDVI (Green Normalized Difference Vegetation Index), LCI (Leaf Color Index), NDRE (Normalized Difference Red Edge), NDVI (Normalized Difference Vegetation Index), and OSAVI (Optimized Soil Adjusat Vegetation Index). The specific calculation formulas are shown in Table 1.

[0065] Table 1 Vegetation index

[0066]

[0067] Step 105: Based on the multispectral information, a digital surface model is constructed using drone image processing software.

[0068] The drone image processing software Agisoft Photoscan Professional was used to process the drone images, and the digital surface model (DSM) was generated by performing aerial triangulation, terrain modeling, and splicing on the drone images. Figure 3 shown).

[0069] Step 106: Based on the fused image and the digital surface model, multiple point samples corresponding to different types of land features are constructed. The land feature types include: felled / burned land, roads, and forest land.

[0070] ROI was applied to create regions of interest for DSM and fused images, and three types of point samples (such as tracks, roads, and woodlands) were drawn in the images through visual interpretation. Figure 5 xml file, and use the Compute ROI Separability function under the ROI Tool in ENVI 5.3 (if the separation degree is greater than 1.9, it means the separation effect is very good) to evaluate the training samples.

[0071] Step 107: performing band fusion on the multiple band features and the multiple vegetation indices corresponding to each point sample to determine the band fusion feature corresponding to each point sample.

[0072] ENVI 5.3 was used to fuse the red, green, blue, and near-infrared bands and the five extracted vegetation indices, and the selected training samples were loaded into the fused image.

[0073] Step 108: Using the band fusion features as input and the ground feature types as output, the support vector machine model is trained to obtain a classification model.

[0074] Step 109: Determine the band fusion feature of each point in the image to be classified.

[0075] Step 1010: Based on the band fusion features of each point in the image to be classified, the ground objects in the image to be classified are classified using a classification model to obtain an initial classification result.

[0076] The support vector machine classification function of ENVI 5.3 is used to classify images and perform classification training based on all samples to obtain classification results (such as Figure 6 shown).

[0077] Step 1011: Use the MajorityAnalysis tool in ENVI 5.3 to post-process the initial classification results to obtain the classification results.

[0078] The results of support vector machine classification are post-classified and processed. The post-classification processing method used is Majority / Minority processing. In ENVI 5.3, it is located under Toolbox / Classification / PostClassification / Majority / MinorityAnalysis. Use MajorityAnalysis (main analysis) and set the kernel size of the area adjacent to the central pixel according to the actual situation. In this embodiment, the kernel size is set to 9*9. If the effect is not good and the small patch problem is still obvious, the operation of step 8 can be repeated multiple times to remove the small patches in the classified image after classification, and finally obtain the post-classification processing result. The post-processing result is as follows Figure 7 The final extraction results of the felled land are as follows: Figure 9 .

[0079] The classification results are evaluated for accuracy. Since there are no real samples in this embodiment, the process of creating regions of interest is repeated, and each type of ROI region of interest is selected again as the verification sample data for classification, and the selected verification sample data is evaluated. The confusion matrix is selected to evaluate the classification results. This function is located in Toolbox / Classification / Post Classification / Confusion MatrixUsing GroundTruthROIs in ENVI 5.3. In the pop-up matching category dialog box, match the training set and verification set sample attribute fields to output the accuracy verification confusion matrix report (such as Figure 8As shown in Figure 2, a kappa coefficient value greater than 0.80 generally indicates excellent classification quality. In this embodiment, the kappa coefficient reaches 0.9655.

[0080] Among them, the mixed matrix report includes the following evaluation indicators:

[0081] 1. Overall classification accuracy (OverallAccuracy).

[0082] Overall classification accuracy = sum of correctly classified pixels / total number of pixels

[0083] 2. Kappa Coefficient

[0084] The Kappa coefficient can be used to test spatial consistency in image classification and measure classification effectiveness. It is calculated by multiplying the total number of true reference pixels (N) by the sum of the diagonal lines of the confusion matrix (Xkk), subtracting the product of the number of true reference pixels in a class and the total number of classified pixels in that class, and then dividing the result by the square of the total number of pixels minus the product of the total number of true reference pixels in a class and the total number of classified pixels in that class.

[0085] The Kappa coefficient calculation formula is:

[0086]

[0087] 3. Commission error.

[0088] Misclassification error = the sum of pixels that are not classified as Class A / the total number of pixels in Class A

[0089] 4. Omission error.

[0090] Omission error = the sum of pixels belonging to class A but misclassified as other types / the total number of pixels in class A

[0091] 5. Cartographic accuracy (Prod.Acc).

[0092] It refers to the ratio of the number of pixels in the entire image that the classifier correctly classifies as class A (diagonal values) to the total number of true reference pixels in class A (the sum of the class A columns in the confusion matrix).

[0093] 6. User Accuracy (User.Acc).

[0094] It refers to the ratio of the total number of pixels correctly classified as class A (diagonal values) to the total number of pixels in the entire image that the classifier classifies as class A (the sum of the class A rows in the confusion matrix).

[0095] The six metrics in the confusion matrix report can be used to evaluate the accuracy of supervised classification of deforestation / burned land images.

[0096] This application uses multispectral drones to acquire information from multiple spectral bands in the monitoring area, calculates multiple vegetation indices in the monitoring area, and uses a supervised classification method based on support vector machines to quickly and accurately identify and extract felled / burned areas, and quantitatively evaluates the classification results. This system integrates ground data collection, feature extraction, classification and identification, and quality assessment in the felled / burned area monitoring area to form an intelligent system. This improves the efficiency, accuracy, and intelligence of felled / burned area identification and information extraction, enables real-time dynamic supervision of felled / burned areas, promotes the sustainable management of forest resources, and provides basic data for subsequent vegetation restoration and forest land development and planting.

[0097] In an exemplary embodiment, a computer device is provided, which may be a server or a terminal. The computer device includes a processor, a memory, an input / output (I / O) interface, and a communication interface. The processor, memory, and I / O interface are connected via a system bus, and the communication interface is connected to the system bus via the I / O 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 computer program in the non-volatile storage medium. The I / O 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 via a network connection. When executed by the processor, the computer program implements a method for intelligently identifying and extracting felled / burned areas.

[0098] 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.

[0099] 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.

[0100] 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.

[0101] 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).

[0102] 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 include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0103] 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.

[0104] 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 intelligent identification and extraction of felled / burned areas, characterized in that: include: Obtain multispectral information of logged / burned areas; Based on the NaN values corresponding to different bands, ENVI 5.3 was used to process the NaN values of the multispectral information to obtain multiple band features; Fuse the features of multiple bands to obtain a fused image; performing band operations on the fused image to extract multiple vegetation indices; Based on the multispectral information, a digital surface model is constructed using drone image processing software; constructing a plurality of point samples corresponding to different types of land features based on the fused image and the digital surface model; The multiple band features and multiple vegetation indices corresponding to each point sample are band-fused to determine the band-fusion features corresponding to each point sample; The support vector machine model is trained with band fusion features as input and feature types as output to obtain a classification model. Determine the band fusion features of each point in the image to be classified; Based on the band fusion feature of each point in the image to be classified, the classification model is used to classify the ground objects in the image to be classified to obtain an initial classification result.

2. The method for intelligent identification and extraction of felled / burned areas according to claim 1, characterized in that: The multispectral information is obtained using an unmanned aerial vehicle equipped with a multispectral imager.

3. The method for intelligent identification and extraction of felled / burned areas according to claim 1, characterized in that: The wavebands include: near infrared waveband, red edge waveband, red light waveband and green light waveband.

4. The method for intelligent identification and extraction of felled / burned areas according to claim 3, characterized in that: The vegetation index includes: green normalized vegetation index, leaf chlorophyll index, normalized difference red edge index, normalized vegetation index and optimized soil adjusted vegetation index.

5. The method for intelligent identification and extraction of felled / burned areas according to claim 4, characterized in that: The green normalized vegetation index is: GNDVI=(NIR-Green) / (NIR+Green); The leaf chlorophyll index is: LCI=(NIR-Red edge) / (NIR+Red); The normalized difference red edge index is: NDRE=(NIR-Red edge) / (NIR+Red edge); The normalized difference vegetation index is: NDVI=(NIR–Red) / (NIR+Red); The optimized soil-adjusted vegetation index is: OSAVI=(NIR-Red) / (NIR+Red+0.16); Where, GNDVI is the green normalized vegetation index; LCI is the leaf chlorophyll index; NDRE is the normalized difference red edge index; NDVI is the normalized vegetation index; OSAVI is the optimized soil adjusted vegetation index; NIR is the near-infrared band characteristics; Red is the red light band characteristics; Green is the green light band characteristics; Red edge is the red edge band characteristics.

6. The method for intelligent identification and extraction of felled / burned areas according to claim 1, characterized in that: The drone image processing software is: Agisoft PhotoscanProfessional.

7. The method for intelligent identification and extraction of felled / burned areas according to claim 1, characterized in that: The types of land features include: clearcut / burned land, roads and woodlands.

8. The method for intelligent identification and extraction of felled / burned areas according to claim 1, characterized in that: After obtaining an initial classification result by classifying the ground objects in the image to be classified using the classification model based on the band fusion feature of each point in the image to be classified, the method further includes: The MajorityAnalysis tool in ENVI 5.3 was used to post-process the initial classification results to obtain classification results.

9. 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 intelligent identification and extraction of felled / burned areas according to any one of claims 1 to 8.

10. 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 intelligently identifying and extracting felled / burned areas according to any one of claims 1 to 8 is implemented.

Images