An automated multi-threshold discrimination and extraction method and device for a burned area

Abstract

The present invention discloses an automated multi-threshold discrimination and extraction method and device for burned areas. The method includes: preprocessing the pre-disaster and post-disaster remote sensing images of the burned area to generate preprocessed images; constructing an NDSI index based on the preprocessed images; processing the remote sensing images based on the NDSI index to enhance the spectral differences between cloud, water, shadow interference factors and the burned area, and generating raster images; automatically threshold-classifying the burned area and cloud, water, shadow interference factors based on the OTSU algorithm; performing interference factor elimination processing on the remote sensing images based on threshold segmentation to generate area elimination images; extracting the NDVI-D index for two periods of pre-disaster and post-disaster for the area elimination images, and performing adaptive threshold segmentation on the extraction results to obtain a binary raster map of the burned area, and obtaining a vector map of the burned area as needed. The present invention can achieve the purpose of accurately extracting the scope of the burned area.

Description

Technical Field

[0001] The present invention relates to the technical field of processed images of burned areas, and in particular to a method and device for automatically discriminating and extracting multiple thresholds of burned areas. Background Art

[0002] As one of the main natural disasters in China, forest and grassland fires have a very fast development trend and unpredictable combustion directions, making it difficult to be quickly controlled. However, with the continuous development of remote sensing satellites, reliable guarantee for real-time global monitoring of fires has been provided. Based on multi-source multi-spectral satellites, information such as the area and scope of burned areas can be accurately obtained, which is of great significance for real-time monitoring of the development trend of the fire situation during the disaster, and provides efficient assistance for the timely control of the fire situation and the disaster assessment after the disaster.

[0003] Remote sensing technology has been widely used in the monitoring of forest and grassland fires due to its characteristics such as large detection range, fast information acquisition, and less ground restriction. Before and after a fire occurs, the vegetation traits in the burned area change, and its characteristic spectra also change greatly, especially in the near-infrared band range (0.75~2.5μm). Currently, the detection of burned areas mainly relies on the analysis of spectral characteristics. By constructing spectral index operations within the area, the accurate range of the burned area is inversely calculated. However, due to the rapid development of forests and grasslands, and the influence of weather factors after a fire, a large number of clouds may appear around the burning area. Since the values of shadows in the near-infrared band are very close to those of clouds, it will seriously interfere with the extraction of the burned area. In addition, buildings around the burned area will also bring shadows. Therefore, removing cloud shadows is particularly important for automatically and quickly extracting the accurate range of the burned area. Summary of the Invention

[0004] The technical problem solved by the present invention is: overcoming the deficiencies of the prior art, and providing a method and device for automatically discriminating and extracting multiple thresholds of burned areas.

[0005] In order to solve the above technical problem, a method for automatically discriminating and extracting multiple thresholds of burned areas provided by an embodiment of the present invention includes:

[0006] Preprocessing the remotely sensed images of the burned area before and after the disaster to generate preprocessed images;

[0007] Based on the preprocessed images, constructing the NDSI index;

[0008] Based on the NDSI index, processing the remotely sensed images to enhance the spectral differences between cloud, water, shadow interference factors and the burned area, and generating raster images;

[0009] Automatically classifying the burned area and cloud, water, shadow interference factors based on the OTSU algorithm;

[0010] Based on threshold segmentation, the interference factors in the remote sensing image are removed to generate a regional removal image;

[0011] Extract the NDVI-D index for the two periods before and after the disaster for the regional removal image, and perform adaptive threshold segmentation on the extraction results to obtain a binary raster map of the burned area, and obtain a vector map of the burned area as needed.

[0012] Optionally, the preprocessing of the pre-disaster and post-disaster remote sensing images of the burned area is performed to generate a preprocessed image, including:

[0013] Perform radiometric calibration and atmospheric correction on the remote sensing image to obtain a corrected image;

[0014] Resample the corrected image to obtain a resampled image;

[0015] Perform bit-depth reduction on the resampled image to obtain a bit-depth reduced image;

[0016] Perform band composition on the bit-depth reduced image to generate the preprocessed image.

[0017] Optionally, based on the preprocessed image, construct the NDSI index, including:

[0018] Perform band operation on the preprocessed image to obtain a band operation result;

[0019] Based on the band operation result, construct the NDSI index.

[0020] Optionally, based on the OTSU algorithm, automatically threshold classify the burned area and interference factors such as clouds, water, and shadows, including:

[0021] Based on the OTSU algorithm, divide the raster image into two parts;

[0022] Convert the images of interference factor regions such as clouds, water, and shadows obtained by threshold segmentation of the two periods of pre-disaster and post-disaster images by the Otsu method into vectors, and take the union of the two periods of vectors to obtain the maximum removal vector of the two periods of images, which is used as the extraction mask;

[0023] Based on the removal of clouds, water, and shadows from the two pre-disaster and post-disaster images, extract the preprocessed image to remove invalid information that interferes with the extraction of the burned area;

[0024] Judge whether there are still regions below the set empirical threshold after the image is processed;

[0025] If the minimum pixel value of the post-disaster image is greater than the set empirical threshold, the operation ends, indicating that there is no burned area in the region;

[0026] If the minimum pixel value of the post-disaster image is less than the set empirical threshold, it is set to N to indicate the presence of a burned area within the region.

[0027] To solve the above technical problems, an embodiment of the present invention provides an automated multi-threshold discrimination and extraction device for burned areas, including:

[0028] A preprocessing image generation module for preprocessing the pre-disaster and post-disaster remote sensing images of the burned area obtained to generate a preprocessed image;

[0029] An NDSI index construction module for constructing an NDSI index based on the preprocessed image;

[0030] A raster image generation module for processing the remote sensing image based on the NDSI index, enhancing the spectral differences between strong cloud, water, shadow interference factors and the burned area, and generating a raster image;

[0031] A threshold segmentation module for automatically threshold-classifying the burned area and cloud, water, shadow area interference factors based on the OTSU algorithm;

[0032] An area elimination image generation module for performing interference factor elimination processing on the remote sensing image based on threshold segmentation to generate an area elimination image;

[0033] A burned area vector map acquisition module for extracting the NDVI-D index for two periods of pre-disaster and post-disaster of the area elimination image, and performing adaptive threshold segmentation on the extraction result to obtain a binary raster map of the burned area, and obtaining a vector map of the burned area as needed.

[0034] Optionally, the preprocessing image generation module includes:

[0035] A corrected image acquisition unit for performing radiometric calibration and atmospheric correction processing on the remote sensing image to obtain a corrected image;

[0036] A resampled image acquisition unit for resampling the corrected image to obtain a resampled image;

[0037] A downsampled image acquisition unit for performing downsampling processing on the resampled image to obtain a downsampled image;

[0038] A preprocessing image generation unit for performing band composition processing on the downsampled image to generate the preprocessed image.

[0039] Optionally, the NDSI index construction module includes:

[0040] A band operation result acquisition unit for performing band operations on the preprocessed image to obtain a band operation result;

[0041] The NDSI index construction unit is used to construct the NDSI index based on the result of the band operation.

[0042] Optionally, the burned area extraction module includes:

[0043] The raster image segmentation unit is used to segment the raster image into two parts based on the OTSU algorithm;

[0044] The mask extraction unit converts the images of interference factors such as clouds, water, and shadows obtained by threshold segmentation of the pre-disaster and post-disaster images by the Otsu method into vectors, takes the union of the two vectors, and obtains the maximum rejection vector of the two images as the extraction mask;

[0045] The invalid information removal unit is used to remove clouds, water, and shadows from the pre-disaster and post-disaster images, extract the preprocessed images, and remove the invalid information that interferes with the extraction of the burned area;

[0046] The area judgment unit is used to judge whether there is still an area below the set empirical threshold after the image is processed;

[0047] The non-burned area determination unit is used to end the operation if the minimum pixel value of the post-disaster image is greater than the set empirical threshold, indicating that there is no burned area in the area;

[0048] The burned area determination unit is used to set it to N if the minimum pixel value of the post-disaster image is less than the set empirical threshold, indicating that there is a burned area in the area.

[0049] The advantages of the present invention compared with the prior art are as follows:

[0050] The embodiments of the present invention can effectively eliminate the interference factors around the burned area, and the obtained burned area effect is better. The OSTU Otsu method adaptive threshold extraction method is used, and the extraction threshold does not need to be set manually, which improves the extraction efficiency of large-scale burned areas. Moreover, the overall process of burned area extraction is constructed, and it can automatically process from image processing, influence factor elimination, burned area extraction to removing small patches and smoothing vectors. This automated processing method has a good application effect in practical applications. Description of the Drawings

[0051] Figure 1 It is a step flow chart of an automated multi-threshold discrimination extraction method for burned areas provided by an embodiment of the present invention;

[0052] Figure 2 It is a structural schematic diagram of an automated multi-threshold discrimination extraction device for burned areas provided by an embodiment of the present invention;

[0053] Figure 3 It is a burned area extraction diagram obtained by implementing the present invention. Detailed implementation manners

[0054] Example 1

[0055] Referring to Figure 1 , a step flowchart of an automated multi-threshold discrimination and extraction method for a burned area provided by an embodiment of the present invention is shown. As Figure 1 shown, the method may include the following steps:

[0056] Step 101: Preprocess the pre-disaster and post-disaster remote sensing images of the burned area obtained to generate preprocessed images;

[0057] Step 102: Based on the preprocessed images, construct the NDSI index;

[0058] Step 103: Based on the NDSI index, process the remote sensing images to enhance the spectral differences between cloud, water, shadow interference factors and the burned area, and generate raster images;

[0059] Step 104: Automatically classify the burned area and cloud, water, shadow interference factors based on the OTSU algorithm;

[0060] Step 105: Based on the threshold segmentation, perform interference factor removal processing on the remote sensing images to generate area removal images;

[0061] Step 106: Extract the NDVI-D index for two periods of pre-disaster and post-disaster of the area removal images, and perform adaptive threshold segmentation on the extraction results to obtain a binary raster map of the burned area, and obtain a vector map of the burned area as needed.

[0062] The technical solution adopted in this embodiment is as follows: First, process the pre-disaster and post-disaster Sentinel-2 images obtained, resample them to improve the resolution, then reduce the dimension to 8 bits, construct the NDSI index. The NDSI index has a good recognition effect on clouds, water, and shadows. Based on the OTSU adaptive threshold, extract the cloud, water, and shadow areas from the raster images calculated by the index and perform reclassification. Set the cloud, water, and shadow areas to 0, and other areas to 1, and create a mask to remove the original images. Extract the NDVI-D index for two periods of pre-disaster and post-disaster of the obtained removed images, and perform adaptive threshold segmentation on the extraction results to obtain a binary raster map of the burned area, and obtain a vector map of the burned area as needed. The overall process is automated, and only by inputting raster images can the accurate burned area range be obtained.

[0063] First, the indexes and algorithms involved in the automated burned area extraction process can be briefly described.

[0064] 1. NDSI index

[0065] Since the burned area is prone to be confused with surrounding water bodies, buildings, or cloud shadows during the extraction process, it is crucial to remove water bodies and shadows in advance to improve the accuracy of burned area extraction. Water bodies have obvious spectral characteristics in the near-infrared band, and shadow areas can be well distinguished from the burned area in the short-wave infrared and mid-infrared ranges. Therefore, the present invention constructs the NDSI index to eliminate the influence of clouds, water bodies, and shadows.

[0066] The constructed NDSI index is:

[0067]

[0068] 2. NDVI-D index

[0069] One of the conditions for a fire to occur is a large amount of combustibles, and the combustibles in forest fires or grassland fires are mostly vegetation. Therefore, there will be obvious changes in the vegetation cover on the ground surface before and after a fire in the burned area. So, the burned area can be identified based on the judgment of the ground surface vegetation. Healthy vegetation reflects more near-infrared (NIR) and green light but absorbs more red and blue light. Therefore, the NDVI index can be used as a standardized method to measure the vegetation state. Due to the obvious changes in vegetation before and after the fire, subtracting the NDVI before and after the fire can obtain a better result of the burned area range because there are no obvious spectral information changes in other ground objects.

[0070] The NDVI index is:

[0071]

[0072]

[0073] 3. OSTU algorithm

[0074] The assumption of the OTSU algorithm is that there exists a threshold TH that divides all pixels of the image into two classes C1

[0075] (less than TH) and C2 (greater than TH). Then the means of these two classes of pixels are m1 and m2 respectively, and the global mean of the image is mG. At the same time, the probabilities that pixels are divided into classes C1 and C2 are p1 and p2 respectively. Therefore, there is:

[0076]

[0077]

[0078] The between-class variance expression is:

[0079]

[0080] Among them, , ,

[0081] The gray level k that maximizes σ2 is the OTSU threshold.

[0082] OSTU is considered the best algorithm for threshold selection in image segmentation. It is simple to calculate, not affected by image brightness and contrast, and can find ideal thresholds for the extraction of dewatered, cloud, shadow, and burned areas in the present invention.

[0083] Next, the detailed process of the embodiments of the present invention will be described as follows.

[0084] The automated process of this embodiment is generally divided into four modules: 1. Image preprocessing module; 2. Image dewatering, cloud, and shadow masking module; 3. Burned area judgment and calculation extraction module; 4. Burned area result processing and output module. Each step of the overall process can generate raster or vector data of the required intermediate products according to requirements and can be used as the input of other related algorithms.

[0085] The detailed process is as follows:

[0086] 1.1 Obtain the pre-disaster and post-disaster images of the burned area of Sentinel-2, and use the Sen2cor plugin released by ESA to perform radiometric calibration and atmospheric correction on the L1C data, and output the processed raster image.

[0087] 1.2 Resample the processed pre-disaster and post-disaster raster images, and resample all bands to a resolution of 10m.

[0088] 1.3 Reduce the bit depth of the obtained images of each band to 8bit, which not only reduces the data volume but also facilitates subsequent processing.

[0089] 1.4 Synthesize the required bands, and the synthesized raster image can be selected for output according to requirements. The quality of the processed image can be checked and other processing can be performed on the original image data according to the situation.

[0090] 2.1 Perform band operations on the preprocessed Sentinel-2 images before and after the disaster respectively. Select Band 4 of the Sentinel-2 image for the red band, Band 8 of the Sentinel-2 image for the near-infrared band, and Band 12 of the Sentinel-2 image for the short-wave infrared band. Calculate NDSI using these three bands. The result of NDSI is in the range of [-1,1], and NDSI can be selected for output as the intermediate process result.

[0091] 2.2 Automatically obtain the threshold using the OTSU method, divide the obtained raster image into two parts, set the part greater than the adaptive threshold to 0, and the part less than the adaptive threshold to 1.

[0092] 2.3 Convert the pre-disaster and post-disaster images obtained by Otsu threshold segmentation into vectors, take the union of the two-phase vectors, and obtain the maximum cloud, water, and shadow vectors of the two-phase images, which are used as the extraction mask.

[0093] 2.4 Use the maximum mask vector extracted from the pre-disaster and post-disaster images to remove clouds, water, and shadows from the pre-disaster and post-disaster images. Extract the pre-processed pre-disaster and post-disaster Sentinel-2 images and remove the invalid information extracted from the interfering burned areas.

[0094] 2.5 Perform threshold judgment on the images after removing clouds, water, and shadows. Set the empirical threshold and judge whether there are still areas below the set empirical threshold after the images are processed. If the minimum pixel value of Band 4 in the post-disaster image is greater than the set empirical threshold, the operation ends, indicating that there is no burned area in the region. If the minimum pixel value of Band 4 in the post-disaster image is less than the set empirical threshold, set it to N, indicating that there is a burned area in the region.

[0095] 2.6 Count the number of pixels in Band 4 of the post-disaster image that are less than the set threshold. Let the number be K. If it is less than the set threshold, it may be noise or the remaining points of clouds, water, and shadows after mask removal, and it is not considered, and it is considered that there is no burned area in the region. If it is greater than the set threshold, traverse each point with a Band 4 pixel value greater than the set empirical threshold. If there are no N values greater than the set ratio in its surrounding 8×8 matrix, the point is identified as a non-burned area point. If there are N values greater than the set ratio in its surrounding 8×8 matrix, it is regarded as a point in the burned area. After traversing, if there are such points, it is regarded as having a burned area in this range, otherwise it is regarded as having no burned area.

[0096] 3.1 The NDVI (Normalized Difference Vegetation Index) is used in the calculation of the burned area. The pre-disaster and post-disaster pre-processed Sentinel-2 images are used for band operations respectively. The red band selects Band 4 of the Sentinel image, and the near-infrared band selects Band 8 of the Sentinel image. Perform NDVI band operations on the pre-disaster and post-disaster Sentinel-2 images respectively, and subtract the NDVI processing results of the pre-disaster and post-disaster to obtain the range of the area with obvious spectral changes before and after the disaster, that is, the area of the burned area extracted. Construct the index NDVI-D based on the difference of NDVI before and after the disaster.

[0097] 3.2 Perform OSTU (Otsu) adaptive threshold optimization on the obtained NDVI-D raster image to select the best threshold. Set the part greater than the obtained threshold to 1, which is the range of the burned area, and set the part less than the obtained threshold to 0, which is the range of the non-burned area.

[0098] 4.1 Remove small patches from the obtained binary raster map of the burned area, remove the noise generated during the extraction process, convert the obtained processed binary map into a vector, smooth the edges of the vector, and eliminate the surface components of the vector according to requirements.

[0099] 4.2 Store all intermediate results and final achievements uniformly for easy calling and modification in subsequent processing.

[0100] Example of the automated extraction process for the burned area

[0101] Obtain the pre-disaster and post-disaster Sentinel-2 remote sensing images of the Muli County forest fire on March 30 in Sichuan, automatically preprocess the input images, obtain the images after radiometric calibration and atmospheric correction and downscale them to 8 bits, select the required images for band composition and output the image results.

[0102] Perform band operations on the preprocessed Sentinel-2 images, calculate the NDSI, and automatically select the optimal threshold according to the OSTU automatic threshold method to generate a binary map. Convert the two-phase raster to a vector and find the union, and use the merged vector to mask clouds, water, and shadows to remove the factors interfering with the extraction of the burned area. Judge the burned area of the image after removing clouds, water, and shadows. First, set an empirical threshold to judge whether there are points smaller than the threshold in the area. If so, calculate the number of points smaller than the threshold. If the number is too small, it is regarded as noise, that is, there is no burned area in this area. If the number of points is greater than the set threshold, traverse the suspected burned area points. If the number of suspected burned area points within the 8×8 matrix range around this point is greater than the set percentage, there is a burned area in this area and the next operation is performed.

[0103] Calculate the NDVI band for the pre-disaster and post-disaster processed images respectively, and find the difference between the results to obtain the NDVI-D. Use the OSTU method to find the optimal threshold for this image for segmentation to obtain the range of the burned area.

[0104] Remove small patches from the obtained burned area range and smooth the edges of the patches. Store the obtained burned area extraction results and intermediate processes, as specifically shown in Figure 3 Figures (a) and (b).

[0105] Example Two

[0106] Refer to Figure 2 , which shows the structural schematic diagram of an automated multi-threshold discrimination extraction device for the burned area provided by an embodiment of the present invention. As shown in Figure 2 shown, the device may include the following modules:

[0107] The preprocessing image generation module 210 is configured to preprocess the obtained pre- and post-disaster remote sensing images of the burned area to generate a preprocessed image;

[0108] The NDSI index construction module 220 is configured to construct an NDSI index based on the preprocessed image;

[0109] The raster image generation module 230 is configured to process the remote sensing image based on the NDSI index to enhance the spectral difference between cloud, water, shadow interference factors and the burned area, and generate a raster image;

[0110] The threshold segmentation module 240 is configured to automatically classify the burned area and cloud, water, shadow area interference factors based on the OTSU algorithm;

[0111] The area rejection image generation module 250 is configured to perform interference factor rejection processing on the remote sensing image based on threshold segmentation to generate an area rejection image;

[0112] The burned area vector map acquisition module 260 is configured to extract the NDVI-D index for two periods of pre- and post-disaster of the area rejection image, perform adaptive threshold segmentation on the extraction result to obtain a binary raster map of the burned area, and obtain a vector map of the burned area as needed.

[0113] Optionally, the preprocessing image generation module includes:

[0114] The calibrated image acquisition unit is configured to perform radiometric calibration and atmospheric correction processing on the remote sensing image to obtain a calibrated image;

[0115] The resampled image acquisition unit is configured to resample the calibrated image to obtain a resampled image;

[0116] The bit-depth reduction image acquisition unit is configured to perform bit-depth reduction processing on the resampled image to obtain a bit-depth reduced image;

[0117] The preprocessing image generation unit is configured to perform band composition processing on the bit-depth reduced image to generate the preprocessed image.

[0118] Optionally, the NDSI index construction module includes:

[0119] The band operation result acquisition unit is configured to perform band operation on the preprocessed image to obtain a band operation result;

[0120] The NDSI index construction unit is configured to construct the NDSI index based on the band operation result.

[0121] Optionally, the burned area extraction module includes:

[0122] A raster image segmentation unit for segmenting the raster image into two parts based on the OTSU algorithm;

[0123] A mask extraction unit that converts the images of interference factor regions such as clouds, water, and shadows obtained by threshold segmentation of the pre-disaster and post-disaster images using the Otsu method into vectors, takes the union of the two vectors, and obtains the maximum exclusion vector of the two images as the extraction mask;

[0124] An invalid information removal unit for removing clouds, water, and shadows from the pre-disaster and post-disaster images, extracting the preprocessed images, and removing the invalid information that interferes with the extraction of the burned area;

[0125] A region judgment unit for judging whether there are still regions below the set empirical threshold after the image is processed;

[0126] A non-burned area determination unit for ending the operation if the minimum pixel value of the post-disaster image is greater than the set empirical threshold, indicating that there is no burned area in the region;

[0127] A burned area determination unit for setting it to N if the minimum pixel value of the post-disaster image is less than the set empirical threshold, indicating that there is a burned area in the region.

[0128] The specific embodiments described in this application can enable those skilled in the art to understand this application more comprehensively, but do not limit this application in any way. Therefore, those skilled in the art should understand that they still make modifications or equivalent replacements to this application; and all technical solutions and their improvements that do not depart from the spirit and technical essence of this application should be covered by the protection scope of this application's patent.

[0129] The content not described in detail in the specification of the present invention belongs to the well-known technology of those skilled in the art.

Claims

1. An automated multi-threshold discrimination and extraction method for burned areas, characterized in that, Including: Preprocessing the pre-disaster and post-disaster remote sensing images of the burned area to obtain a preprocessed image; Constructing the NDSI index based on the preprocessed image; Processing the remote sensing image based on the NDSI index to enhance the spectral differences between cloud, water, shadow interference factors and the burned area, and generating a raster image; Automatically threshold-classifying the burned area and cloud, water, shadow interference factors based on the OTSU algorithm; Performing interference factor removal processing on the remote sensing image based on threshold segmentation to generate a region removal image; Extracting the NDVI-D index for two periods (pre-disaster and post-disaster) of the region removal image, and performing adaptive threshold segmentation on the extraction results to obtain a binary raster map of the burned area, and obtaining a vector map of the burned area as needed; The preprocessing of the pre-disaster and post-disaster remote sensing images of the burned area to obtain a preprocessed image includes: Performing radiometric calibration and atmospheric correction processing on the remote sensing image to obtain a corrected image; Resampling the corrected image to obtain a resampled image; Performing bit-depth reduction processing on the resampled image to obtain a bit-depth reduced image; Performing band composition processing on the bit-depth reduced image to generate the preprocessed image; The automatic threshold classification of the burned area and cloud, water, shadow interference factors based on the OTSU algorithm includes: Segmenting the raster image into two parts based on the OTSU algorithm; Converting the cloud, water, shadow interference factor region images obtained by Otsu threshold segmentation of the pre-disaster and post-disaster two-period images into vectors, and taking the union of the two-period vectors to obtain the maximum removal vector of the two-period images, which is used as an extraction mask; Removing cloud, water and shadow from the pre-disaster and post-disaster two scenes of images, and extracting the preprocessed image to remove invalid information interfering with the extraction of the burned area; Judging whether there are still areas below the set empirical threshold after the image is processed; If the minimum pixel value of the post-disaster image is greater than the set empirical threshold, the operation ends, indicating that there is no burned area in the region; If the minimum pixel value of the post-disaster image is less than the set empirical threshold, set it as N, indicating that there is a burned area in the region.

2. The method according to claim 1, characterized in that, The construction of the NDSI index based on the preprocessed image includes: Performing band operation on the preprocessed image to obtain a band operation result; Constructing the NDSI index based on the band operation result.

3. An automated multi-threshold discrimination and extraction device for a burned area, characterized in that, Including: A preprocessed image generation module for preprocessing the pre-disaster and post-disaster remote sensing images of the burned area to generate a preprocessed image; An NDSI index construction module for constructing the NDSI index based on the preprocessed image; A raster image generation module for processing the remote sensing image based on the NDSI index to enhance the spectral differences between cloud, water, shadow interference factors and the burned area, and generating a raster image; A threshold segmentation module for automatically threshold-classifying the burned area and cloud, water, shadow area interference factors based on the OTSU algorithm; A region removal image generation module for performing interference factor removal processing on the remote sensing image based on threshold segmentation to generate a region removal image; The burned area vector map acquisition module is used to extract the NDVI-D index for two periods before and after the disaster for the area-removed image, perform adaptive threshold segmentation on the extraction results to obtain a binary raster map of the burned area, and obtain a vector map of the burned area as needed; The preprocessing image generation module includes: The calibrated image acquisition unit is used to perform radiometric calibration and atmospheric correction processing on the remote sensing image to obtain a calibrated image; The resampled image acquisition unit is used to resample the calibrated image to obtain a resampled image; The bit-depth reduced image acquisition unit is used to perform bit-depth reduction processing on the resampled image to obtain a bit-depth reduced image; The preprocessing image generation unit is used to perform band composition processing on the bit-depth reduced image to generate the preprocessing image; The burned area extraction module includes: The raster image segmentation unit is used to segment the raster image into two parts based on the OTSU algorithm; The mask extraction unit converts the cloud, water, and shadow interference factor region images obtained by threshold segmentation of the two periods of pre-disaster and post-disaster images using the Otsu method into vectors, takes the union of the two-period vectors to obtain the maximum removal vector of the two-period images, and uses this as the extraction mask; The invalid information removal unit is used to remove clouds, water, and shadows from the two pre-disaster and post-disaster images, extract the preprocessing image, and remove invalid information that interferes with the extraction of the burned area; The area judgment unit is used to judge whether there are still areas below the set empirical threshold after the image is processed; The no-burned area determination unit is used to end the operation if the minimum pixel value of the post-disaster image is greater than the set empirical threshold, indicating that there is no burned area in the region; The burned area determination unit is used to set it as N if the minimum pixel value of the post-disaster image is less than the set empirical threshold, indicating that there is a burned area in the region.

4. The device according to claim 3, wherein The NDSI index construction module includes: The band operation result acquisition unit is used to perform band operations on the preprocessing image to obtain band operation results; The NDSI index construction unit is used to construct the NDSI index based on the band operation results.

Images