Regional man-made disturbance supervision method based on multi-source remote sensing

Abstract

The invention discloses a regional man-made disturbance supervision method based on multi-source remote sensing, and belongs to the technical field of water and soil conservation supervision. The method comprises the following steps: acquiring at least two periods of remote sensing images, and performing segmentation and alignment; performing change detection on the segmented remote sensing image to obtain a change pattern spot; performing business processing on the change pattern spots, and deleting non-business pattern spots to obtain business pattern spots; supplementing the missed extraction pattern spots and forming interpretation result pattern spots together with the service pattern spots; acquiring prevention and control responsibility range data, forming prevention and control responsibility range pattern spots, comparing the interpretation result pattern spots with the prevention and control responsibility range data, and judging supervision categories; the interpretation result pattern spots which do not meet the requirements are listed as checking pattern spots to be rechecked; and the re-checking result and the result meeting the requirement supervision category jointly form artificial disturbance supervision data which is used as a basis for regional supervision. According to the method, the automation level of regional man-made disturbance supervision can be effectively improved, and man-made disturbance region supervision with short duration, high efficiency and high precision is realized.

Description

Technical Field

[0001] The invention belongs to the technical field of soil and water conservation supervision, and specifically relates to a regional artificial disturbance supervision method based on multi-source remote sensing. Background Art

[0002] With the advancement of urbanization in my country, production and construction projects and production and construction activities have caused severe disturbances to the surface, leading to serious soil and water erosion problems. Soil and water erosion supervision with a focus on soil and water conservation supervision and management of production and construction projects is becoming increasingly important.

[0003] At present, the national-scale remote sensing supervision of soil and water loss caused by human disturbance is mainly based on manual interpretation, and intelligent interpretation is rarely used in national business links or is only used in local areas, resulting in the research level of intelligent interpretation still being at a low level. Summary of the invention

[0004] The present invention aims to solve one of the technical problems in the above-mentioned related art at least to a certain extent.

[0005] To this end, the purpose of the present invention is to provide a method for regional human disturbance supervision based on multi-source remote sensing, which can effectively improve the automation level of regional human disturbance supervision and realize short-duration, high-efficiency and high-precision regional supervision of human disturbance.

[0006] In order to solve the above-mentioned technical problems, the present invention is achieved as follows: An embodiment of the present invention provides a method for monitoring regional artificial disturbances based on multi-source remote sensing, the method comprising: Acquire at least two phases of remote sensing images, segment the remote sensing images, and align the upper and lower phases of images; Perform change detection on the segmented two-phase remote sensing images to obtain change spots; Perform business processing on the changed spots and delete the non-business spots to obtain business spots; Supplement the missing patches, and the supplemented missing patches and business patches together constitute the interpretation result patches; Obtain the data on the scope of responsibility for prevention and control and form a map of the scope of responsibility for prevention and control, compare the interpreted map with the map of the scope of responsibility for prevention and control, and determine the supervision category of each interpreted map; The interpretation result blocks corresponding to the regulatory categories that do not meet the requirements shall be listed as verification blocks and reviewed; the review results and the judgment results of the regulatory categories that meet the requirements shall be combined to form the artificial disturbance supervision data, which will be used as the basis for regional supervision.

[0007] In addition, the regional artificial disturbance supervision method based on multi-source remote sensing according to the present invention may also have the following additional technical features: In some of the embodiments, the regulatory category of the interpretation result patch is determined based on the established spatial topological relationship overlap rules; the principle for establishing the spatial topological relationship overlap rules is to focus on the patch area and to take the patch spatial relationship as a supplement.

[0008] In some implementations, the content of the spatial topological relationship overlap rule includes: The interpreted result spots that do not overlap with any prevention and control responsibility area spots are judged as suspected unapproved construction or suspected location changes; The interpretation result spots included in the prevention and control responsibility area spots are judged as compliant; Interpretation result spots that overlap with the prevention and control responsibility scope and whose proportion of the overlapping area to the entire interpretation result spot area exceeds the judgment threshold will be judged as suspected of exceeding the prevention and control responsibility scope; interpretation result spots whose proportion of the overlapping area to the entire interpretation result spot area does not exceed the judgment threshold will be judged as having no obvious violations.

[0009] In some embodiments, the determination threshold is 30%.

[0010] In some of the embodiments, data that is not in the current human disturbance supervision area and is obviously erroneous are eliminated when forming the prevention and control responsibility area map.

[0011] In some of the embodiments, change detection of remote sensing images is achieved through the ResNet18+Transformer deep learning model and the existing collaborative interpretation platform.

[0012] In some implementations, when segmenting the remote sensing image, the remote sensing image is segmented into image grids of 256*256 pixels.

[0013] In some of the embodiments, the verification spot is reviewed by taking an aerial photo of the verification spot area, re-examining the verification spot based on the aerial image and the collected basic information of the verification spot area, and determining the supervision category of the verification spot.

[0014] In some of the embodiments, the ResNet18+Transformer deep learning model is associated with the existing collaborative interpretation platform through the WebAPI service interface.

[0015] In some of the embodiments, the non-operational patches are patches that change due to seasonal changes or natural changes.

[0016] Compared with the prior art, the present invention has at least the following beneficial effects: In the embodiment of the present invention, the provided method for regional human disturbance supervision based on multi-source remote sensing can effectively improve the automation level of regional human disturbance supervision through the combination of image segmentation technology, automatic extraction technology and informatization, and realize short-duration, high-efficiency and high-precision human disturbance regional supervision; In the embodiments of the present invention, a regional human disturbance supervision method based on multi-source remote sensing is provided, which utilizes high-resolution remote sensing images better than 2 meters, cooperates with intelligent interpretation technology and a supervision platform system, and realizes full coverage of human disturbance soil and water loss supervision once a year across the country; the use of remote sensing images greatly shortens the time for supervising the same task of the national land area compared to no impact, and intelligent interpretation greatly shortens the time and improves the efficiency in the interpretation link compared to manual visual interpretation; at the same time, intelligent interpretation is based on an excellent sample library, and can more accurately identify disturbances and achieve high precision.

[0017] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A schematic diagram of the modular structure of a method for monitoring regional artificial disturbances based on multi-source remote sensing disclosed in one embodiment of the present invention; Figure 2 A flow chart of regional man-made disturbance supervision and identification disclosed in one embodiment of the present invention; Figure 3 A schematic diagram of the cooperation mode between the deep learning model and the collaborative interpretation platform disclosed in one embodiment of the present invention; Figure 4 A schematic diagram of overlapping rules of spatial topological relationships disclosed in one embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] The embodiments of the present invention are described in detail below through specific embodiments and application scenarios in conjunction with the accompanying drawings.

[0021] See also Figure 1As shown, in some embodiments of the present invention, a method for regional human disturbance supervision based on multi-source remote sensing is provided, high-resolution remote sensing images are segmented, and regional human disturbance supervision is achieved by relying on the existing soil and water conservation information service platform ("collaborative interpretation platform", "information supervision platform", "mobile field verification APP", "soil and water conservation mobile data platform"), with "intelligent interpretation", "business application", "visual discrimination" and "precise verification" as the technical route. The latest automatic extraction technology, the latest image segmentation technology, and a high degree of informatization are used to achieve short-duration, high-efficiency, and high-precision supervision on a national scale.

[0022] The following explains intelligent interpretation, business applications, visual identification, and precise verification respectively.

[0023] (1) Intelligent interpretation Image segmentation: The two phases of remote sensing images are divided into several square images of defined sizes, and the upper and lower images are aligned. In this embodiment, the provincial administrative region is used as the operation unit, and it is divided into image grids of 256*256 pixels. The upper and lower remote sensing images are images of the two years before and after the same quarter, and the latter year is compared with the previous year. The meaning of aligning the upper and lower images is: after image correction, there is no offset between the upper and lower periods, and the four boundaries are used as boundaries, and they are framed in a rectangular frame and cut into image grids. The computer uses squares as units to compare the same position of the two images to see if there are any changes; the role of alignment is to compare and identify the changes between the upper and lower images.

[0024] Change Detection: Change detection is performed through the ResNet18+Transformer deep learning model and collaborative interpretation platform.

[0025] Specifically, based on the existing "ResNet18+Transformer deep learning model" for change detection, in order to realize the business application of regional human disturbance supervision, this method is deeply integrated with the collaborative interpretation platform to realize intelligent change detection.

[0026] In this embodiment, the integration process of the deep learning model and the collaborative interpretation platform is as follows: The deep learning model and the collaborative interpretation platform are associated through the WebAPI service interface. The interpretation range of the collaborative interpretation platform's map is input into the deep learning model through the API, and the model then returns the results of automatic change detection of graphic elements to the collaborative interpretation platform. Figure 3 shown.

[0027] (2) Business Application Non-business image deletion: In the operationalization process, not all changed spots are the objects of human disturbance supervision, and the interpretation spots will first be deleted as non-operational spots.

[0028] The supervision objects of regional human disturbance business map are: artificial and disturbed map, including production and construction projects and production and construction activities, excluding natural landform changes. Because there is no unified and clear rule, the deletion of non-business map is reviewed by experienced auditors with the help of existing platform gadgets. The platform used can be a collaborative interpretation platform.

[0029] Using the platform's "programming selection function", first delete the non-operational spots with an area less than 1 that do not need to be reviewed, and then use the platform's "browse" function for the remaining spots, using the current image as the base map, and delete the "undisturbed" spots; "Undisturbed" spots refer to spots that change due to seasonal changes or color changes. The reasons for the interpretation of these spots may be caused by seasonal changes between the two images, or by natural factors (such as landslides, mudslides) or differences in image recording equipment.

[0030] During the audit process, the auditor found that there were disturbance spots that belonged to the supervision object and needed to be interpreted. There are two ways to interpret them. One is to directly draw the boundary of the spot manually, and the other is to use the platform tool "Assisted Interpretation Mode". This function uses the collaborative interpretation platform to call the "Segment Anything Algorithm" within the visible range of the screen to extract the disturbance boundary by inputting the given point or mask with the mouse. The given point or mask here can be a point clicked by the mouse or a circle drawn by the mouse, and the vector boundary of the disturbance is automatically extracted according to the point or circle.

[0031] Additional information on missing images: For missed patches found outside the existing patches, the boundaries are automatically extracted using the “SegmentAnything algorithm” that has been integrated into the collaborative interpretation platform. All patches constitute the first-stage interpretation result patches.

[0032] (3) Visual discrimination like Figure 2 shown.

[0033] The scope of responsibility for prevention and control of production and construction projects is entered into the platform: The National Soil and Water Conservation Information Management System is used to collect the prevention and control responsibility areas of the corresponding regions in the past five years and map them to the collaborative interpretation platform. The quality is checked and vectors that are not in the administrative area or are obviously erroneous are eliminated.

[0034] Since the main purpose of the prevention and control responsibility scope in the supervision of this area is to superimpose with the disturbance spots and establish spatial topological rules to judge the compliance of the spots. Therefore, the accuracy (quality) of the prevention and control responsibility scope has a direct impact on the spots; generally speaking, the prevention and control responsibility scope is reported by the soil and water conservation authorities and is accurate by default, but before establishing the topological rules, a rough inspection procedure is added. The specific quality inspection content of a project is: 1. Whether the spatial position of the prevention and control responsibility scope vector is obviously different from the administrative district to which the "construction site" in its attributes belongs; 2. Whether the vector area of ​​the prevention and control responsibility scope vector is obviously different from the approved "prevention and control responsibility scope area" in the attributes.

[0035] The specific implementation process is: 1. Use a computer (or ARCGIS software) to calculate the longitude and latitude, and vector range area of ​​the prevention and control responsibility area; 2. Read the construction site and approved area in the attributes; 3. Make a comparison. If the longitude and latitude fall within the administrative area to which the construction site belongs, the spatial location is judged to be accurate, otherwise it is judged to be inaccurate. If |(vector area-approved area) / approved area|*100%<30%, it is judged to be accurate, otherwise it is judged to be inaccurate.

[0036] Establishment of spatial topological relationship: Establish the topological spatial overlap rules between business spots and prevention and control responsibility spots, and remove business spots with an area of ​​less than 1 hectare. Directly assign the attribute of "suspected of being built without approval" to non-intersecting spots; perform visual identification of topologically overlapping business spots, automatically calculate the excess area ratio based on the principle of "area first, space second", and copy the excess area ratio to the business spot attributes. For those with an excess ratio greater than or equal to 30%, the attribute is assigned as "suspected of exceeding the boundary of the prevention and control responsibility range", and for those with an excess ratio less than 30%, the attribute is assigned as "no obvious violation".

[0037] The spatial topological relationship overlap rules are as follows Figure 4 As shown. The red circle in the figure represents the scope of prevention and control responsibilities, and the yellow circle represents the business map. When the business map is included in the map of the scope of prevention and control responsibilities, it is considered compliant. When there is only the business map, it is considered "suspected of construction without approval / suspected location change". When the map of the scope of prevention and control responsibilities is included in the business map or the business map intersects with the map of the scope of prevention and control responsibilities (that is, partially overlaps), if the proportion of the business map area outside the scope of prevention and control responsibilities to the entire business map area is greater than or equal to 30%, it is considered "suspected of exceeding the boundary of the scope of prevention and control responsibilities"; if it is less than 30%, it is considered "no obvious violation".

[0038] Assign ID to the patch: For business spots without topological overlap and spots with an over-range ratio of more than 30%, they will enter the next stage as "verification spots". A unique identity "ID" will be assigned to the "verification spots" in the "year-administrative district code-code" method.

[0039] (4) Accurate review The "verification spots" are synchronized to the "regional supervision platform" and "field verification platform", and the "field verification platform + drone" is used to collect basic information and aerial photography, and as verification information, the verification information is returned to the business spots of the "regional supervision platform" to obtain a list of violations and achieve the purpose of regional supervision of human disturbances. For spots with an area of ​​more than 1 hectare that are suspected of being built without approval, suspected of location changes, and suspected of exceeding the scope of responsibility for prevention and control, the purpose of the review is to determine whether it is indeed "built without approval". If the review finds that there is an approval for the scope of responsibility for prevention and control and the disturbance area does not exceed 30% compared with the approval, (A approval-A spot) / A spot), then it is compliant and there is an injustice, but the reason is that the approval scope is not entered into the database in time; if the review finds that there is indeed no plan and there is no document stating that the water conservation plan can be exempted from preparation, then it is undoubtedly illegal and irregular, and it is one of the following: construction without approval, exceeding the scope, and location change.

[0040] It should be noted that the national soil and water conservation information management system, collaborative interpretation platform, information-based supervision platform, mobile field verification APP, and soil and water conservation mobile data platform are all existing systems or platforms. This embodiment does not impose too many restrictions on their contents and will not be described in detail here.

[0041] Parts of the present invention that are not described in detail are well known to those skilled in the art.

[0042] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

Claims

1. A method for monitoring regional human disturbance based on multi-source remote sensing, characterized in that: The method comprises: Obtain at least two phases of remote sensing images, segment the remote sensing images, and align the upper and lower phases of images; Perform change detection on the segmented two-phase remote sensing images to obtain change spots; Perform business processing on the changed spots and delete the non-business spots to obtain business spots; Supplement the missing patches, and the supplemented missing patches and business patches together constitute the interpretation result patches; Obtain the data on the scope of responsibility for prevention and control and form a map of the scope of responsibility for prevention and control, compare the interpreted map with the map of the scope of responsibility for prevention and control, and determine the regulatory category of each interpreted map; The interpretation result blocks corresponding to the regulatory categories that do not meet the requirements shall be listed as verification blocks and reviewed; the review results and the judgment results of the regulatory categories that meet the requirements shall be combined to form the artificial disturbance supervision data, which will be used as the basis for regional supervision.

2. The method for regional artificial disturbance supervision based on multi-source remote sensing according to claim 1 is characterized in that: When judging the regulatory category of the interpretation result map, it is judged according to the established spatial topological relationship overlapping rules; the principle of establishing the spatial topological relationship overlapping rules is to focus on the map area and supplemented by the spatial relationship of the map.

3. The method for regional artificial disturbance supervision based on multi-source remote sensing according to claim 2 is characterized in that: The content of the spatial topological relationship overlap rule includes: The interpreted result spots that do not overlap with any prevention and control responsibility area spots are judged as suspected unapproved construction or suspected location changes; The interpretation result spots included in the prevention and control responsibility area spots are judged as compliant; Interpretation result spots that overlap with the prevention and control responsibility scope and whose proportion of the overlapping area to the entire interpretation result spot area exceeds the judgment threshold will be judged as suspected of exceeding the prevention and control responsibility scope; interpretation result spots whose proportion of the overlapping area to the entire interpretation result spot area does not exceed the judgment threshold will be judged as having no obvious violations.

4. The method for regional artificial disturbance supervision based on multi-source remote sensing according to claim 3 is characterized in that: The determination threshold is 30%.

5. The method for regional artificial disturbance supervision based on multi-source remote sensing according to claim 1 is characterized in that: When forming the prevention and control responsibility area map, data that is not in the current human disturbance supervision area and data that is obviously erroneous are eliminated.

6. The method for monitoring regional artificial disturbances based on multi-source remote sensing according to any one of claims 1 to 5, characterized in that: Change detection of remote sensing images is achieved through the ResNet18+Transformer deep learning model and the existing collaborative interpretation platform.

7. The method for monitoring regional artificial disturbances based on multi-source remote sensing according to any one of claims 1 to 5, characterized in that: When segmenting the remote sensing image, the remote sensing image is divided into image grids of 256*256 pixels.

8. The method for monitoring regional artificial disturbances based on multi-source remote sensing according to claim 1, characterized in that: The method for reviewing the verification map is to take aerial photos of the verification map area, re-examine it based on the aerial images and the collected basic information of the verification map area, and determine the supervision category of the verification map.

9. The method for monitoring regional artificial disturbances based on multi-source remote sensing according to claim 6, characterized in that: The ResNet18+Transformer deep learning model is associated with the existing collaborative interpretation platform through the WebAPI service interface.

10. The method for monitoring regional artificial disturbances based on multi-source remote sensing according to claim 1, characterized in that: The non-operational patches are patches that change due to seasonal changes or natural changes.

Images