A method and apparatus for constructing a remote sensing base map of glacial or periglacial landforms

Abstract

The application discloses a glacier or ice edge landform remote sensing base map construction method and equipment, and the method comprises the steps of taking an image location map as a basic processing unit, screening remote sensing images according to preset conditions for each location map of a glacier or ice edge landform distribution area, and labeling an interpretability score for each image based on snow cover, cloud cover and cloud shadow cover; determining a main image for each location map according to the interpretability score, if the location map has only one image or the main image has met the direct interpretation requirement, no auxiliary image needs to be specified, otherwise at least one selected auxiliary image is determined; for the area affected by cloud and cloud shadow in the main image, the pixels corresponding to the cloud-free area in the auxiliary image are selected for inlay replacement to generate a cloud-removed main image of the location map; and the cloud-removed main images generated for all the location maps are inlaid and spliced to form an image base map for remote sensing information extraction covering the entire glacier or ice edge landform distribution area.

Description

Technical Field

[0001] This invention relates to the field of remote sensing image processing technology, and in particular to a method and apparatus for constructing remote sensing base maps of glaciers or periglacial landforms. Background Technology

[0002] Remote sensing information extraction from glaciers or periglacial landforms is easily affected by snow cover, clouds, and cloud shadows, leading to errors in the extraction range. The impact of snow cover is particularly significant—because its color tone is similar to that of glaciers in images, if the glacier surface and surrounding area are covered by snow, it becomes difficult to accurately determine the boundaries of the bare ice area; simultaneously, snow cover can also obscure periglacial landforms, causing missing image information. Furthermore, the presence of clouds and cloud shadows in the image can result in missing ground feature information, making it impossible to correctly identify the target area.

[0003] In existing practices for extracting remote sensing information about glaciers or periglacial landforms, if manual interpretation is used, operators often need to simultaneously interpret and screen images to minimize the impact of snow, clouds, and cloud shadows on the interpretation results. This process is often time-consuming and labor-intensive, significantly reducing interpretation efficiency. If traditional supervised classification, unsupervised classification, or other machine learning methods, or methods based on band ratios and threshold segmentation are used for interpretation, extensive post-processing is usually required to eliminate interference from snow, clouds, and cloud shadows, with snow interference removal being particularly complex.

[0004] Therefore, there is an urgent need for an efficient method for constructing image base maps for extracting remote sensing information of glaciers or periglacial landforms.

[0005] The information disclosed in this background section is intended only to enhance the understanding of the overall background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0006] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention provides a method and apparatus for constructing remote sensing base maps of glaciers or periglacial landforms. The purpose is to provide an efficient method for constructing image base maps that can pre- and automatically eliminate snow and cloud interference, thereby solving the technical problems of low efficiency caused by the need for image screening during traditional manual interpretation methods, and the need for complex post-processing to eliminate snow and cloud interference when using traditional automated methods.

[0007] To achieve the above objectives, the present invention provides a method for constructing a remote sensing base map of glaciers or periglacial landforms, the method comprising:

[0008] Using image location maps as the basic processing unit, remote sensing images are selected for each location map of the glacier or periglacial landform distribution area according to preset conditions, and interpretability scores are assigned to each image based on snow cover, cloud cover and cloud shadow coverage.

[0009] Based on the interpretability score, a primary image is determined for each location map. If the location map has only one image or the primary image already meets the direct interpretation requirements, there is no need to specify an auxiliary image; otherwise, at least one alternative auxiliary image is determined.

[0010] For the areas in the main image affected by clouds and cloud shadows, pixels from the corresponding cloudless areas in the auxiliary image are selected for mosaic replacement to generate a cloudless main image of the location map.

[0011] All the cloud-free master images generated from the location maps are mosaicked to form an image base map covering the entire distribution area of ​​the glacier or periglacial landform, which is used for remote sensing information extraction.

[0012] In one embodiment of the present invention, filtering remote sensing images according to preset conditions includes:

[0013] The remote sensing images were initially screened based on time and cloud cover conditions.

[0014] The time condition is based on the base year set by the work requirements, extending forward and backward by no more than 2 years each.

[0015] The cloud cover conditions are determined by setting appropriate cloud cover thresholds based on the cloud cover characteristics of glaciers or geomorphological areas.

[0016] In one embodiment of the present invention, an interpretability score is assigned to each image based on snow cover, clouds, and cloud shadow coverage. The specific method is as follows:

[0017] The images are categorized into multiple levels based on their interpretability scores.

[0018] Images that simultaneously meet the criteria of having a base year in time phase and no significant snow cover or cloud and cloud shadow interference in glacier or periglacial landform areas are rated as having the highest interpretability level.

[0019] For images containing clouds and cloud shadows but without significant snow accumulation, a comprehensive assessment is made by combining the temporal phase and the degree of cloud interference.

[0020] For images with snow cover, the assessment is based on the degree of snow cover.

[0021] In one embodiment of the present invention, the specific method for determining the main image and the auxiliary image is as follows:

[0022] If there is only one image after screening in a location map, it is directly determined as the main image.

[0023] If there are multiple images, the image with the highest interpretability score is selected as the main image;

[0024] When the highest score indicates that the quality of the primary image has met the requirements for direct interpretation, there is no need to designate a secondary image; otherwise, the image with the second highest score will be designated as the secondary image.

[0025] In one embodiment of the present invention, a method for generating a cloud-free main image of the location map by selecting pixels from the auxiliary image corresponding to cloudless areas in the main image that are affected by clouds and cloud shadows for mosaic replacement includes:

[0026] Using a pre-trained deep learning model, clouds and cloud shadows in glacier or periglacial landform areas in the main image and the auxiliary image are identified, and the identified boundaries are buffered to generate corresponding cloud mask vectors; based on the cloud mask vectors, the corresponding areas of the image are assigned null values.

[0027] Determine whether there is spatial overlap between the cloud mask vectors of the main image and the auxiliary image; if not, directly use the pixel mosaic of the auxiliary image to fill the null area of ​​the main image; if so, use the non-null pixel mosaic of the auxiliary image to fill the null area of ​​the main image, and update the intersection area of ​​the two mask vectors to the new cloud mask area of ​​the main image.

[0028] In one embodiment of the present invention, the method for generating the de-clouded master image of the location map further includes:

[0029] If there are still empty regions in the main image, the image with the highest interpretability score is selected from the remaining unused images corresponding to the currently processed location map as a new auxiliary image. The cloud area recognition, masking, and pixel mosaic operations are repeated until the empty regions of the main image are completely filled or there are no remaining images available.

[0030] In one embodiment of the present invention, the specific grading rules for assigning interpretability scores to each image include:

[0031] The first category is based on the base year for the image time phase, and there is no obvious snow cover, cloud cover or cloud shadow in the glacier or periglacial landform area;

[0032] The second category is images taken in the base year, and there is no obvious snow accumulation in the glacier or periglacial landform area, but there are clouds and cloud shadows covering it.

[0033] The third category consists of images taken in a time period other than the baseline year, and where there is no significant snow cover, cloud cover, or cloud shadow in the glacier or periglacial landform area.

[0034] The fourth category consists of images taken in a time period other than the base year, and where there is no significant snow cover in glaciers or periglacial landforms, but there are clouds and cloud shadows covering the area.

[0035] The fifth category is areas with snow cover, such as glaciers or periglacial landforms.

[0036] In one embodiment of the present invention, the score for the first tier is 10 points, and the score for the third tier is 8 points.

[0037] In one embodiment of the present invention, the corresponding area of ​​the image is assigned a null value based on the cloud mask vector, and the operation is performed only on the cloud and cloud shadow corresponding areas in the distribution area of ​​glaciers or periglacial landforms.

[0038] This invention also provides a device for constructing remote sensing base maps of glaciers or periglacial landforms, the device comprising:

[0039] One or more processors;

[0040] A storage device for storing one or more programs that, when executed by one or more processors, enable the one or more processors to implement the method for constructing remote sensing base maps of glaciers or periglacial landforms as described in the first aspect.

[0041] In summary, the technical solutions conceived by this invention have the following beneficial effects compared with the prior art:

[0042] This invention uses image location maps as the basic processing unit. Through standardized image screening and interpretability scoring, it accurately selects primary and secondary images. Combined with a deep learning model, it accurately identifies clouds and cloud shadows and uses iterative mosaicking to fill gaps. This effectively eliminates the interference of snow, clouds, and cloud shadows on the remote sensing interpretation of glaciers or periglacial landforms, significantly improving the integrity and interpretability of the image base map. At the same time, it simplifies the base map construction process and improves construction efficiency. The generated base map can be directly used for remote sensing information extraction of glaciers or periglacial landforms without additional post-processing, further improving the efficiency and accuracy of subsequent remote sensing interpretation work. Attached Figure Description

[0043] Figure 1 This is a flowchart of a method for constructing a remote sensing base map of glaciers or periglacial landforms provided in this embodiment.

[0044] Figure 2 This is a glacier or periglacial landform map with an interpretability score of 10 provided in this embodiment;

[0045] Figure 3 This is a glacier or periglacial landform map with an interpretability score of 9 provided in this embodiment;

[0046] Figure 4 This is a glacier or periglacial landform map with an interpretability score of 8.9 provided in this embodiment;

[0047] Figure 5 This is a glacier or periglacial landform map with an interpretability score of 8.8 provided in this embodiment;

[0048] Figure 6This is a glacier or periglacial landform map with an interpretability score of 7 provided in this embodiment;

[0049] Figure 7 This is a glacier or periglacial landform map with an interpretability score of 6.9 provided in this embodiment;

[0050] Figure 8 This is a glacier or periglacial landform map with an interpretability score of 6.8 provided in this embodiment;

[0051] Figure 9 This is a glacier or periglacial landform map with an interpretability score of 6 provided in this embodiment;

[0052] Figure 10 This is a glacier or periglacial landform map with an interpretability score of 5.9 provided in this embodiment;

[0053] Figure 11 This is a glacier or periglacial landform map with an interpretability score of 5.8 provided in this embodiment;

[0054] Figure 12 This is a schematic diagram of the main image of a location map provided in this embodiment 1;

[0055] Figure 13 This is a supplementary image diagram of a location map provided in this embodiment 1;

[0056] Figure 14 This is a schematic diagram of the mosaic replacement of the main and auxiliary images of a location map provided in this embodiment 1;

[0057] Figure 15 This is a schematic diagram of a remote sensing base map construction device for glaciers or periglacial landforms provided in this embodiment. Detailed Implementation

[0058] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings, but it should be understood that the scope of protection of the present invention is not limited to the specific embodiments.

[0059] Unless otherwise expressly stated, throughout the specification and claims, the term "comprising" or its variations such as "including" or "comprises" shall be understood to include the stated elements or components without excluding other elements or other components.

[0060] Example 1:

[0061] This embodiment provides a method for constructing a remote sensing base map of glaciers or periglacial landforms, such as... Figure 1 As shown, it includes the following steps:

[0062] S101: Using image location maps as the basic processing unit, for each location map of a glacier or periglacial landform distribution area, remote sensing images are selected according to preset conditions, and an interpretability score is assigned to each image based on snow cover, cloud cover, and cloud shadow coverage.

[0063] Filtering remote sensing images based on preset conditions involves preliminary screening of remote sensing images based on time and cloud cover conditions.

[0064] The time frame is based on the base year set by the work requirements, extending forward and backward by no more than 2 years each, and prioritizing images from the local summer months;

[0065] In this first embodiment, the cloud cover condition is based on the cloud cover characteristics of glaciers or periglacial landforms, setting an appropriate cloud cover threshold. For most glaciers or periglacial landforms, such as the Tianshan Mountains, Qilian Mountains, West Kunlun Mountains, and the Alps, the cloud cover threshold ranges from 10% to 30%. For glaciers or periglacial landforms with perennial high cloud cover, such as the southeastern section of the Nyainqêntanglha Mountains in southeastern Tibet and the Hengduan Mountains, the cloud cover threshold is relaxed to no more than 50%. To avoid the lack of a unified standard for image screening, which could easily lead to the selection of images severely affected by snow cover and cloud shadows, thus affecting the quality of subsequent base map construction, this first embodiment provides quantitative initial screening standards in terms of time and cloud cover, achieving efficient and targeted data reduction.

[0066] In this first embodiment, a human-computer interaction method is used to further filter images and assign interpretability scores to glaciers or periglacial landforms. Based on the quick view of the initial screened images, an interpretability score is assigned to each image according to the following rules:

[0067] Score 10: The image time is the base year; there is no significant snow cover, cloud cover, or shadow coverage in the glacier or periglacial landform area. Figure 2 As shown. Figure 2 Sentinel-2 image (RGB true color display) of glaciers and periglacial landforms in the eastern section of the main peak area of ​​the West Kunlun Mountains on September 18, 2020 (base year).

[0068] Score 9: The image time is the base year. The image shows no significant snow cover in the glacier or periglacial landform area, but with cloud and shadow coverage. Cloud cover is ranked 1st in ascending order. Figure 3 As shown. Figure 3 Sentinel-2 image (RGB true color display) of glaciers and periglacial landforms in the perennially cloudy area of ​​southeastern Tibet on October 17, 2020 (base year).

[0069] Score 8.9: The image time is the base year. There is no significant snow cover in the glacier or periglacial landform area, but there is cloud and shadow coverage. Cloud cover ranks second in ascending order. Figure 4 As shown. Figure 4Sentinel-2 image (RGB true color display) of glaciers and periglacial landforms in the perennially cloudy area of ​​southeastern Tibet on August 18, 2020 (base year).

[0070] Score 8.8: The image time is the base year. There is no significant snow cover in the glacier or periglacial landform area, but there is cloud and shadow coverage. Cloud cover ranks 3rd in ascending order. Figure 5 As shown. Figure 5 Sentinel-2 image (RGB true color display) of glaciers and periglacial landforms in the perennially cloudy area of ​​southeastern Tibet on August 13, 2020 (base year).

[0071] Score 8.1: The image time is the base year. There is no obvious snow cover in the glacier or periglacial landform area, but there is cloud and shadow coverage. The cloud cover is ranked 10th in ascending order.

[0072] Score 8: The image time is not the base year, and there is no obvious snow cover, cloud cover, or shadow cover in the glacier or periglacial landform area.

[0073] Score 7: The image timeframe is not the base year; the image shows no significant snow cover in the glacier or periglacial landform area, but is covered by clouds and shadows; cloud cover ranks first in ascending order. Figure 6 As shown. Figure 6 Sentinel-2 image (RGB true color display) of glaciers and periglacial landforms in the perennially cloudy area of ​​southeastern Tibet on August 19, 2019.

[0074] Score 6.9: The image timeframe is not the base year; there is no significant snow cover in the glacier or periglacial landform area, but cloud and shadow coverage; cloud cover ranks second in ascending order. Figure 7 As shown. Figure 7 Sentinel-2 image (RGB true color display) of glaciers and periglacial landforms in the perennially cloudy area of ​​southeastern Tibet on August 8, 2022.

[0075] Score 6.8: The image timeframe is not the base year; there is no significant snow cover in the glacier or periglacial landform area, but cloud and shadow coverage; cloud cover ranks 3rd in ascending order. Figure 8 As shown. Figure 8 Sentinel-2 image (RGB true color display) of glaciers and periglacial landforms in the perennially cloudy area of ​​southeastern Tibet on July 24, 2022.

[0076] Score 6.1: The image time is not the base year. There is no obvious snow cover in the glacier or periglacial landform area in the image. There is cloud and shadow coverage. The cloud cover is ranked 10th in ascending order.

[0077] Score 6: The image shows snow cover in the glacier or periglacial landform area, with snow cover rate ranking first in ascending order, such as... Figure 9 As shown, Figure 9Sentinel-2 image (RGB true color display) of glaciers and periglacial landforms in the perennially cloudy area of ​​southeastern Tibet, taken on October 17, 2022.

[0078] Score 5.9: The image shows snow cover in the glacier or periglacial landform area, with snow cover rate ranking second in ascending order. Figure 10 As shown. Figure 10 Sentinel-2 image (RGB true color display) of glaciers and periglacial landforms in the perennially cloudy area of ​​southeastern Tibet on November 16, 2020.

[0079] Score 5.8: The image shows snow cover in the glacier or periglacial landform area, ranking 3rd in ascending order of snow cover. Figure 11 As shown, Figure 11 Sentinel-2 image (RGB true color display) of glaciers and periglacial landforms in the perennially cloudy area of ​​southeastern Tibet, taken on June 5, 2018.

[0080] Score 5.1: The image shows snow cover in the glacier or periglacial landform area, ranking 10th in snow cover rate.

[0081] S102: Based on the interpretability score, determine a main image for each location map. If the location map has only one image or the main image already meets the direct interpretation requirements, there is no need to specify an auxiliary image; otherwise, determine at least one alternative auxiliary image.

[0082] If there is only one image selected from the screening results for a location map, it is directly determined as the primary image; if there are multiple images, the image with the highest interpretability score is selected as the primary image. If the highest score is 10 or 8, it indicates that the quality of the primary image meets the requirements for direct interpretation, and there is no need to designate a secondary image; otherwise, the image with the second highest score is determined as the secondary image.

[0083] S103: For the areas in the main image affected by clouds and cloud shadows, select the corresponding cloudless area pixels from the auxiliary image for mosaic replacement to generate the cloudless main image of the location map.

[0084] Taking a location map with a main image (denoted as image M) and a secondary image (denoted as image S) as an example, the process is as follows:

[0085] In this embodiment, using a deep learning model pre-trained with working area samples, the attention-based DeepLabV3+ model can be selected to identify clouds and cloud shadows in glacier or periglacial landform areas in images M and S. The identified boundaries are buffered outwards by 100 meters (to eliminate identification errors), generating corresponding vector masks, referred to as vector M and vector S, respectively. Based on the vector masks, the corresponding regions in the images are assigned null values, such as... Figure 12 and Figure 13 As shown, Figure 12The main image M (RGB true color display) and vector M (the pixels are null after masking, and are displayed as black) of the perennially cloudy area in southeastern Tibet. Figure 13 The auxiliary image S (RGB true color display) and vector S (the pixels are null after masking and are displayed as black) of the perennially cloudy area in southeastern Tibet.

[0086] Next, determine whether vector M and vector S have spatial overlap.

[0087] If there is no overlay, the pixel mosaic of image S is used directly to fill all the empty value areas of image M, and the cloud removal of the location map is completed.

[0088] If there is overlap, the non-nullable pixel mosaic of image S is used to fill the nullable regions of image M. At this point, some cloud residue remains on the main image (i.e., the overlapping area of ​​the two masks). To address this, the intersection of vector M and vector S is calculated, and this intersection is updated to a new vector M; the corresponding region in the main image remains nullable. For example... Figure 14 As shown, Figure 14 The updated master image M (RGB true color display) and vector M (the pixels are null after masking, and are displayed as black) for the perennially cloudy area in southeastern Tibet.

[0089] If, after completing one mosaicking operation, there are still null regions in the main image (image M) (i.e., the new vector M is not empty), then select the image with the highest interpretability score from the remaining unused images in the location map as the new auxiliary image (i.e., the new image S), and repeat the above cloud recognition, masking, judgment and mosaicking operations until the null regions in the main image are completely filled or there are no remaining images available.

[0090] S104: Mosaic and stitch together the cloud-free master images generated from all location maps to form an image base map covering the entire distribution area of ​​glaciers or periglacial landforms for remote sensing information extraction.

[0091] Since the distribution area of ​​glaciers or periglacial landforms usually requires multiple image location maps to be fully covered, after completing the construction of the cloud-free master image of each location map according to the S101-S103 process, the master images of all location maps are mosaicked to finally obtain a complete remote sensing image base map that can be used for the extraction of glacier or periglacial landform information.

[0092] To address the issues of subjective scoring and inconsistent standards caused by unclear scoring criteria and classification principles, which prevent accurate assessment of the image's suitability for interpreting glaciers or periglacial landforms.

[0093] This first embodiment assigns an interpretability score to each image based on snow cover, cloud cover, and cloud shadow coverage. The specific method is as follows:

[0094] The images are categorized into multiple levels based on their interpretability scores.

[0095] The highest level of explanation is defined as the time phase being the base year and the glacier or periglacial landform area having no significant snow cover and cloud and cloud shadow interference.

[0096] For images containing clouds and cloud shadows but without significant snow accumulation, a comprehensive assessment is made by combining the temporal phase and the degree of cloud interference.

[0097] For images with snow cover, the assessment is based on the degree of snow cover.

[0098] This first embodiment establishes a tiered scoring system based on image temporal phase, snow cover, cloud cover, and cloud shadow cover as core dimensions. It clarifies the specific scoring rules under different coverage conditions, achieving an objective and standardized assessment of interpretability scores. This not only accurately selects images suitable for interpretation but also provides a reliable quantitative basis for determining subsequent primary and secondary images.

[0099] To address the issue of improper or redundant image selection due to ambiguity in the rules for determining primary and secondary images, this embodiment 1 describes the specific method for determining primary and secondary images as follows:

[0100] If there is only one image after screening in a location map, it is directly determined as the main image.

[0101] If there are multiple images, the image with the highest interpretability score is selected as the main image;

[0102] When the highest score indicates that the quality of the primary image has met the requirements for direct interpretation, there is no need to designate a secondary image; otherwise, the image with the second highest score will be designated as the secondary image.

[0103] This first embodiment clarifies the specific rules for determining primary and secondary images. Based on the number of images and the highest interpretability score, these rules establish an automated decision-making process that includes the selection and matching of secondary images. This enables the accurate and efficient selection of primary and secondary images, thereby ensuring the basic quality of base map construction while avoiding the ineffective use of redundant images and reducing time and data costs.

[0104] To address the issues of incomplete cloud and cloud shadow recognition, chaotic mosaicking logic leading to poor cloud removal results and insufficient base map integrity, a method is proposed to generate a cloud-removed main image of the location map by selecting pixels from corresponding cloud-free areas in the auxiliary image for mosaicking replacement of regions affected by clouds and cloud shadows in the main image. This method includes:

[0105] Using a pre-trained deep learning model, clouds and cloud shadows in glacier or periglacial landform areas in the main image and the auxiliary image are identified, and the identified boundaries are buffered to generate corresponding cloud mask vectors; based on the cloud mask vectors, the corresponding areas of the image are assigned null values.

[0106] Determine whether there is spatial overlap between the cloud mask vectors of the main image and the auxiliary image; if not, directly use the pixel mosaic of the auxiliary image to fill the null area of ​​the main image; if so, use the non-null pixel mosaic of the auxiliary image to fill the null area of ​​the main image, and update the intersection area of ​​the two mask vectors to the new cloud mask area of ​​the main image.

[0107] This embodiment constructs a standardized pixel-level replacement process that includes deep learning recognition, boundary buffering, mask assignment, and spatial overlay judgment. This process achieves accurate identification and targeted filling of cloud and cloud shadow regions, effectively eliminating cloud shadow interference in the main image; at the same time, its clear spatial overlay processing logic lays a clear program foundation for possible subsequent iterative gap filling operations.

[0108] To address the issue that single mosaicking may not completely eliminate null values ​​due to the overlay of cloud shadow vector spaces between the main and auxiliary images, the method for generating the de-clouded main image of the location map in this embodiment further includes:

[0109] If there are still empty regions in the main image, the image with the highest interpretability score is selected from the remaining unused images corresponding to the currently processed location map as a new auxiliary image. The cloud area recognition, masking, and pixel mosaic operations are repeated until the empty regions of the main image are completely filled or there are no remaining images available.

[0110] This embodiment introduces an iterative gap-filling mechanism. After a single processing iteration, this mechanism automatically selects the image with the highest interpretability score from the remaining unused images as a new auxiliary image, and repeats the cloud region identification, masking, and pixel mosaic operations until the gaps are completely filled or no images are available. This iterative mechanism based on the remaining high-resolution images can fill gap areas to the maximum extent, ensuring the integrity of the de-clouded master image of a single location map, thus providing a reliable unit foundation for the subsequent construction of a high-quality global mosaic base map.

[0111] In this first embodiment, the specific grading rules for assigning interpretability scores to each image include:

[0112] The first category is based on the base year for the image time phase, and there is no obvious snow cover, cloud cover or cloud shadow in the glacier or periglacial landform area;

[0113] The second category is images taken in the base year, and there is no obvious snow accumulation in the glacier or periglacial landform area, but there are clouds and cloud shadows covering it.

[0114] The third category consists of images taken in a time period other than the baseline year, and where there is no significant snow cover, cloud cover, or cloud shadow in the glacier or periglacial landform area.

[0115] The fourth category consists of images taken in a time period other than the base year, and where there is no significant snow cover in glaciers or periglacial landforms, but there are clouds and cloud shadows covering the area.

[0116] The fifth category is areas with snow cover, such as glaciers or periglacial landforms.

[0117] This first embodiment specifies the classification rules for interpretability scores into five clear judgment levels, clearly defining the core judgment conditions for different quality images, ranging from interference-free in the base year to those with snow cover. These specific classification rules eliminate ambiguity in the scoring process, making the scoring more consistent and objective, and thus making the selection of primary and secondary images based on scores more accurate and reliable.

[0118] To address the issue of ambiguous judgment logic and interrupted automated decision-making caused by the lack of specific quantitative thresholds for the criterion of meeting direct interpretation requirements, this first embodiment further clarifies that the score for the first tier (no interference in the base year) is 10 points, and the score for the third tier (no interference in non-base year) is 8 points. This specific score setting provides clear and unambiguous decision thresholds for the automated process, ensuring the consistency and efficiency of the main image selection criteria.

[0119] To avoid wasting computing resources and reducing processing efficiency by performing cloud processing on the entire image, this embodiment strictly limits the cloud masking and null value assignment operations to the cloud and cloud shadow areas corresponding to the distribution areas of glaciers or periglacial landforms.

[0120] This first embodiment effectively eliminates the invalid identification and processing of clouds over non-target ground features, significantly reducing the amount of data required for deep learning model inference and pixel operations, thereby significantly improving the overall computational efficiency of the cloud removal process.

[0121] Example 2:

[0122] This second embodiment provides a device for constructing remote sensing base maps of glaciers or periglacial landforms, used to execute the method for constructing remote sensing base maps of glaciers or periglacial landforms provided in the first embodiment, such as... Figure 15 As shown, the device includes:

[0123] One or more processors;

[0124] A storage device for storing one or more programs that, when executed by one or more processors, enable the one or more processors to implement the method for constructing remote sensing base maps of glaciers or periglacial landforms as described in Embodiment 1.

[0125] Figure 15 This is a schematic diagram of the structure of the remote sensing base map construction device for glaciers or periglacial landforms provided in Embodiment 2 of the present invention. Figure 15 The glacier or periglacial landform remote sensing base map construction device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0126] like Figure 15 As shown, the glacier or periglacial landform remote sensing base map construction device is represented in the form of a general-purpose device. The components of the glacier or periglacial landform remote sensing base map construction device may include, but are not limited to: one or more processors or processing units, memory, and buses connecting different system components (including memory and processing units).

[0127] A bus refers to one or more of several bus architectures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus architectures. Examples of these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnect (PCI) bus.

[0128] Glacier or periglacial landform remote sensing base map building equipment typically includes a variety of computer-readable media. These media can be any available media that can be accessed by the glacier or periglacial landform remote sensing base map building equipment, including volatile and non-volatile media, and portable and immovable media.

[0129] The memory may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory. The glacier or periglacial landform remote sensing base map construction device may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system may be used to read and write non-removable, non-volatile magnetic media (…). Figure 15 Not shown; usually referred to as a "hard drive".

[0130] A program / utility having a set (at least one) of program modules can be stored, for example, in memory. Such program modules include, but are not limited to, an operating system, one or more applications, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. The program modules typically perform the functions and / or methods described in the embodiments of this invention.

[0131] The glacier or periglacial landform remote sensing base map construction device can also communicate with one or more external devices (e.g., keyboard, pointing device, monitor, etc.), and with one or more devices that enable users to interact with the glacier or periglacial landform remote sensing base map construction device, and / or with any device that enables the glacier or periglacial landform remote sensing base map construction device to communicate with one or more other devices (e.g., network card, modem, etc.). This communication can be performed through an input / output (I / O) interface. Furthermore, the glacier or periglacial landform remote sensing base map construction device can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via a network adapter. Figure 15 As shown, the network adapter communicates with other modules of the glacier or periglacial landform remote sensing base map construction device via a bus. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the glacier or periglacial landform remote sensing base map construction device, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0132] Those skilled in the art will understand that all or part of the steps in the various methods of the embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.

[0133] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for constructing a remote sensing base map of glaciers or periglacial landforms, characterized in that, The methods include: Using image location maps as the basic processing unit, remote sensing images are selected for each location map of the glacier or periglacial landform distribution area according to preset conditions, and interpretability scores are assigned to each image based on snow cover, cloud cover and cloud shadow coverage. The process of filtering remote sensing images based on preset conditions includes: The remote sensing images were initially screened based on time and cloud cover conditions. The time condition is based on the base year set by the work requirements, extending forward and backward by no more than 2 years each. The cloud cover conditions are based on the cloud cover characteristics of glaciers or periglacial landforms, and appropriate cloud cover thresholds are set accordingly. The interpretability score is assigned to each image based on snow cover, cloud cover, and cloud shadow coverage. The specific method is as follows: The images are categorized into multiple levels based on their interpretability scores. Images that simultaneously meet the criteria of having a base year in time phase and no significant snow cover or cloud and cloud shadow interference in glacier or periglacial landform areas are rated as having the highest interpretability level. For images containing clouds and cloud shadows but without significant snow accumulation, a comprehensive assessment is made by combining the temporal phase and the degree of cloud interference. For images with snow cover, the assessment is based on the degree of snow cover. Based on the interpretability score, a primary image is determined for each location map. If the location map has only one image or the primary image already meets the direct interpretation requirements, there is no need to specify an auxiliary image; otherwise, at least one alternative auxiliary image is determined. For the areas in the main image affected by clouds and cloud shadows, pixels from the corresponding cloudless areas in the auxiliary image are selected for mosaic replacement to generate a cloudless main image of the location map. All the cloud-free master images generated from the location maps are mosaicked to form an image base map covering the entire distribution area of ​​the glacier or periglacial landform, which is used for remote sensing information extraction.

2. The method for constructing a remote sensing base map of glaciers or periglacial landforms according to claim 1, characterized in that, The specific method for determining the main image and the auxiliary image is as follows: If there is only one image after screening in a location map, it is directly determined as the main image. If there are multiple images, the image with the highest interpretability score is selected as the main image; When the highest score indicates that the quality of the primary image has met the requirements for direct interpretation, there is no need to designate a secondary image; otherwise, the image with the second highest score will be designated as the secondary image.

3. The method for constructing remote sensing base maps of glaciers or periglacial landforms according to claim 2, characterized in that, The method for generating a cloud-free main image of the location map by selecting pixels from the auxiliary image corresponding to cloudless areas in the main image that are affected by clouds and cloud shadows, and then mosaicking and replacing them, includes: Using a pre-trained deep learning model, clouds and cloud shadows in glacier or periglacial landform areas in the main image and the auxiliary image are identified, and the identified boundaries are buffered to generate corresponding cloud mask vectors; based on the cloud mask vectors, the corresponding areas of the image are assigned null values. Determine whether there is spatial overlap between the cloud mask vectors of the main image and the auxiliary image; if not, directly use the pixel mosaic of the auxiliary image to fill the null area of ​​the main image; if so, use the non-null pixel mosaic of the auxiliary image to fill the null area of ​​the main image, and update the intersection area of ​​the two mask vectors to the new cloud mask area of ​​the main image.

4. The method for constructing remote sensing base maps of glaciers or periglacial landforms according to claim 3, characterized in that, The method for generating the de-clouded main image of the location map further includes: If there are still empty regions in the main image, the image with the highest interpretability score is selected from the remaining unused images corresponding to the currently processed location map as a new auxiliary image. The cloud area recognition, masking, and pixel mosaic operations are repeated until the empty regions of the main image are completely filled or there are no remaining images available.

5. The method for constructing a remote sensing base map of glaciers or periglacial landforms according to claim 1, characterized in that, The specific grading rules for assigning interpretability scores to each image include: The first category is based on the image time period as the base year, and there is no obvious snow cover, cloud cover or cloud shadow in the glacier or periglacial landform area; The second category is images taken in the base year, and there is no obvious snow accumulation in the glacier or periglacial landform area, but there are clouds and cloud shadows covering it. The third category consists of images taken in a time period other than the base year, and where there is no significant snow cover, cloud cover, or cloud shadow in the glacier or periglacial landform area. The fourth category consists of images taken in a time period other than the base year, and where there is no significant snow cover in glaciers or periglacial landforms, but there are clouds and cloud shadows covering the area. The fifth category is areas with snow cover, such as glaciers or periglacial landforms.

6. The method for constructing a remote sensing base map of glaciers or periglacial landforms according to claim 5, characterized in that, The first tier scores 10 points, and the third tier scores 8 points.

7. The method for constructing a remote sensing base map of glaciers or periglacial landforms according to claim 3, characterized in that, The method of assigning null values ​​to the corresponding areas of the image based on the cloud mask vector is only applied to the cloud and cloud shadow areas in glacier or periglacial landform distribution areas.

8. A device for constructing remote sensing base maps of glaciers or periglacial landforms, characterized in that the device... include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method for constructing remote sensing base maps of glaciers or periglacial landforms as described in any one of claims 1 to 7.

Images