Intelligent delineation method for red soil type bauxite ore body boundary based on remote sensing image

Abstract

The present application relates to the technical field of image detection, in particular to a method for intelligent delineation of laterite type bauxite ore body boundary based on remote sensing images, comprising the following steps: obtaining multi-band remote sensing image data of a target mining area and performing registration, radiation correction and occlusion removal, dividing the image detection unit according to the surface cover difference, extracting the hue response, texture connectivity and terrain transition constraint related to the laterite type bauxite, and generating a mineralization response intermediate layer; performing adjacent unit consistency check, non-mineralization patch exclusion and boundary closure correction on the mineralization response intermediate layer to obtain the ore body boundary delineation result. The method makes the boundary formation process jointly constrained by pixel level response, patch level continuity and edge closure condition, reduces the mis-delineation caused by bare land, shadow and vegetation cover, maintains the continuity of the ore body boundary at the transition zone and broken patch, and can provide reviewable image basis for subsequent exploration point layout.

Description

Technical Field

[0001] This invention relates to the field of image detection technology, and in particular to a method for intelligent delineation of the boundary of lateritic bauxite ore bodies based on remote sensing images. Background Technology

[0002] Image detection technology involves techniques for identifying, extracting, and labeling target regions, edge morphology, and spatial relationships in remote sensing images, industrial images, or scene images. Among these, the traditional intelligent delineation method for lateritic bauxite orebody boundaries based on remote sensing images refers to a processing flow that uses satellite or aerial remote sensing images as data carriers, involving manual registration, visual interpretation, tone delineation, patch merging, and vectorization to form suspected orebody boundaries.

[0003] The current delineation of the boundaries of lateritic bauxite deposits relies heavily on manual interpretation and visual analysis of single-period images. The color tone of outcrops is affected by vegetation cover, slope shadows, residual clouds and fog, and surface humidity. Image detection results can easily include non-mineralized laterite, bare land, and gully erosion zones within the ore body area. When the ore body boundary is fragmented or the transition zone changes continuously, there is a lack of stable closure evidence. Subsequent exploration point layout needs to be repeatedly revised. Local suspected mineralized areas may also be missed due to image noise, resulting in inconsistencies between the boundary results and the actual spatial response in remote sensing images, which affects the boundary verification. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a method for intelligent delineation of the boundary of lateritic bauxite ore bodies based on remote sensing images.

[0005] To address the technical problem of unstable delineation of lateritic bauxite orebody boundaries due to the influence of bare land, shadows, vegetation cover, and continuous changes in the transition zone, this invention provides an intelligent delineation method for lateritic bauxite orebody boundaries based on remote sensing imagery, comprising: Acquire multi-band remote sensing image data of the target mining area, and perform geometric registration, radiometric correction and occlusion removal on the multi-band remote sensing image data to obtain a basic image raster. Image detection units are divided according to the differences in land cover, pixel adjacency, and local tone changes in the basic image raster, and a pixel set, an adjacency unit set, and an edge pixel set are established for each image detection unit. Extract the red soil tone response, texture connectivity response, terrain transition response and bare land interference response of the image detection unit, and generate a mineralization response intermediate layer based on the consistency between the red soil tone response, the texture connectivity response and the terrain transition response; In the intermediate layer of the mineralization response, non-mineralized patches with bare ground interference responses higher than the mineralization response are screened out. Adjacent cell connectivity checks and edge closure corrections are performed on the remaining patches, and the boundary delineation results of lateritic bauxite ore bodies are output.

[0006] Furthermore, when acquiring multi-band remote sensing image data of the target mining area, the image scene covering the target mining area is read from the local remote sensing image storage space, and control points are extracted from the overlapping area of ​​adjacent image scenes. When the control point offset exceeds the preset pixel deviation, geometric registration is re-executed and the image scene is updated. When cloud pixels, dark shadow pixels, and water pixels continuously cover the edge of the mining area, the corresponding coverage edge is marked as the area to be compensated.

[0007] This ensures that the underlying image raster used for image detection has a clear spatial correspondence and occlusion record.

[0008] As a progressive scheme, the image detection unit is formed step by step from the initial segmented patch, continuous patch and the splitting results of crossing valley lines, slope transition lines or tone change lines, and the adjacency relationship and edge direction of each image detection unit are recorded.

[0009] This constraint ensures that subsequent red earth tone responses, texture connectivity responses, and terrain transition responses all treat the same spatial unit as the processing object, avoiding object drift in boundary judgment between pixels, patches, and vector boundaries.

[0010] To solve the same sub-technical problem, one option is to limit it as follows: The distribution positions of red-toned pixels, yellow-toned pixels, and dark shadow pixels within the image detection unit are statistically analyzed. When red-toned pixels are continuously distributed and dark shadow pixels do not form closed occlusions along the edges, the image detection unit is determined as a spectral candidate unit, and a red earth tone response is generated based on the contact boundary length between the spectral candidate unit and adjacent spectral candidate units. Another optional limitation is: The slope extension direction, valley cutting direction, and patch texture direction are extracted from the base image raster and the local elevation raster registered with the base image raster. When the patch texture direction is continuous with the slope extension direction and the patch edge is not segmented into isolated fragments by the valley cutting direction, the corresponding image detection unit is identified as a terrain candidate unit, and texture connectivity response and terrain transition response are generated accordingly.

[0011] As a preferred embodiment, the red soil tone response, texture connectivity response, terrain transition response and bare land disturbance response are read for each candidate patch in the intermediate layer of mineralization response. When the red soil tone response is true and both the texture connectivity response and the terrain transition response are true, the candidate patch is written into the set of retained patches. When the bare ground interference response covers the main edge of the candidate patch and the texture connectivity response is not valid, the candidate patch is written into the set of unmineralized patches.

[0012] This scheme allows mineralization response judgment to first form intermediate results, and then form boundary results through interference elimination and connectivity verification.

[0013] As an additional limitation, the edge closure correction generates a sequence of boundary points according to the shared boundary of adjacent preserved patches in the preserved patch set. If the distance between boundary points exceeds the preset connection distance and the middle region is not written into the non-mineralized patch set, connection boundary points are added. When internal holes belong to shaded pixels or vegetation-covered pixels, the internal holes are marked as areas to be verified, and the outer closed boundary is preserved.

[0014] When the area to be compensated is adjacent to the set of preserved patches and the red soil hue response is continuous on both sides of the boundary, the area to be compensated participates in the edge closure correction as a boundary extension area. Boundary extension is stopped when the area to be compensated is adjacent to a set of unmineralized patches.

[0015] Finally, the boundary point sequence that has completed edge closure correction is converted into a vector boundary line, and the set of retained patches, the set of non-mineralized patches, the area to be compensated, and the area to be verified are written into the boundary attribute table to generate traceable boundary delineation results for lateritic bauxite ore bodies.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by registering, correcting, removing occlusions, and dividing image detection units into multi-band remote sensing image data, the boundary delineation is transformed from judging the tone of a single pixel to being constrained by both unit response and adjacency relationship. Mis-delineation caused by bare land, shadows, and vegetation cover is further eliminated by boundary closure rules. By forming a mineralization response intermediate layer by combining the red soil tone response, texture connectivity response, terrain transition response, and bare land interference response, the transition zone and broken patches are first confirmed in the intermediate layer before the ore body boundary delineation result is output. Cloud residue and local image loss trigger compensation processing, which can reduce repeated manual corrections and maintain the verifiability of the boundary results. Attached Figure Description

[0017] Figure 1 This is the main flowchart for intelligent delineation of the boundary of lateritic bauxite ore bodies according to the present invention; Figure 2 This is a schematic diagram of the image detection unit division in this invention; Figure 3 This is a schematic diagram illustrating the generation of red earth tone response in this invention; Figure 4 This is a schematic diagram of the terrain candidate unit constraints of the present invention; Figure 5 This is a schematic diagram of the boundary closure correction and verification marking of the present invention. Detailed Implementation

[0018] The following embodiments illustrate the feasible process of the technical solution described in the application documents in the remote sensing image processing of lateritic bauxite deposits. In subsequent embodiments, multi-band remote sensing image data refers to image data covering the target mining area and having band identification, imaging time, and spatial resolution records; the basic image raster refers to a unified spatial raster formed after geometric registration, radiometric correction, and removal of occluded areas from the multi-band remote sensing image data; the image detection unit refers to a processing unit formed in the basic image raster according to differences in land cover, pixel adjacency relationships, and local tonal variations; lateritic tone response, texture connectivity response, topographic transition response, and bare land interference response are used to characterize the image status in the image detection unit related to mineralization signs, patch continuity, topographic transition relationships, and non-mineralized bare land areas in lateritic bauxite deposits. The figure numbers and S-type step numbers in the embodiments are only used to illustrate the correspondence between the process and the figures, and do not constitute a limitation on the quantity, parameters, sequence range, or technical scope. Since the subject matter described in this case is a method, the following embodiments focus on data acquisition, image processing, judgment rules, intermediate layer generation, anomaly handling, and boundary result output, without introducing the subject matter of systems, devices, electronic devices, or storage media that do not appear in the independent protection project.

[0019] Example 1

[0020] Please see Figures 1 to 3 This embodiment provides an intelligent delineation method for the boundary of lateritic bauxite ore bodies based on remote sensing imagery. It is applicable to image detection of multi-band remote sensing image data of target mining areas and delineation of lateritic bauxite ore body boundaries within the images. This embodiment uses the lateritic hue response as the primary identification entry point for generating the intermediate layer of mineralization response, while simultaneously combining texture connectivity response, terrain transition response, and bare land interference response to filter and correct the boundary.

[0021] S1: Acquire multi-band remote sensing image data of the target mining area.

[0022] In this step, image scenes covering the target mining area are read from the local remote sensing image storage space. Each image scene carries imaging time, spatial resolution, and band identifier. Imaging time is used to distinguish the acquisition order of different image scenes, spatial resolution is used to determine whether different image scenes can be processed at the same raster scale, and band identifier is used to determine the image channel source used in subsequent tone response extraction. If there are multiple image scenes for the same target mining area, the image scene covering the main area of ​​the target mining area is first used as a spatial reference, and then control points are extracted from the overlapping area of ​​adjacent image scenes. Control points are used to express the correspondence of the same ground feature location in adjacent image scenes. When the control point offset exceeds the preset pixel deviation, it indicates that the spatial position between adjacent image scenes does not meet the consistency condition for subsequent unit-level detection, and geometric registration needs to be re-performed and the image scenes updated. After re-registration, the ground feature location in the overlapping area is consistent with the spatial reference of the target mining area, and subsequent pixel sets, adjacent unit sets, and edge pixel sets can be established under the same spatial reference.

[0023] S2 performs geometric registration, radiometric correction, and occlusion removal on multi-band remote sensing image data to obtain the basic image raster.

[0024] Geometric registration is used to adjust the positions of ground features in different image scenes to a unified spatial position; radiometric correction is used to reduce the brightness differences of the same area under different imaging conditions, so that red-toned pixels, yellow-toned pixels, and dark shadow pixels have a consistent interpretation basis in subsequent statistics; occlusion removal is used to identify cloud and fog pixels, dark shadow pixels, and water body pixels, and to exclude occlusion positions that cannot reliably participate in mineralization response judgment from the direct judgment objects.

[0025] When cloud, dark shadow, and water pixels continuously cover the edge of a mining area within the overlapping area of ​​adjacent image scenes, the corresponding covered edges are marked as areas to be compensated. These areas are not directly used as the basis for establishing a mineralization response; instead, their participation in boundary extension is determined during the subsequent edge closure correction stage based on the continuity of the red hue response of adjacent preserved patches. Unoccluded pixels are written into the base image raster, which serves as input for subsequent image detection unit division and the generation of intermediate mineralization response layers.

[0026] S3 divides the image detection units according to the differences in land cover, pixel adjacency, and local tone changes in the basic image raster.

[0027] In this step, initial segmentation patches are first generated based on the category boundaries between bare surface pixels, vegetation cover pixels, and shading pixels. Bare surface pixels represent surface areas not significantly covered by vegetation, vegetation cover pixels represent locations where mineralization responses are easily obscured, and shading pixels represent localized dark areas caused by terrain or imaging conditions. Then, initial segmentation patches with areas smaller than a preset patch area and adjacent to similar patches are merged into continuous patches. This merging process does not change the category meaning of the pixels themselves; instead, it organizes scattered but similar and adjacent small patches in the image into continuous objects that can participate in adjacency determination.

[0028] Subsequently, continuous patches crossing gully lines, slope transition lines, or tonal abrupt change lines are divided into image detection units. Gully lines are used to characterize linear cuts caused by surface erosion or runoff, slope transition lines are used to characterize changes in slope aspect or slope extension, and tonal abrupt change lines are used to characterize the edge changes between lateritic and non-latrine tones. Image detection units formed through this division method have clearly defined sets of pixels, adjacent units, and edge pixels, allowing subsequent mineralization response assessment to establish a correspondence between the pixel states within a unit and the boundary states between units.

[0029] S4, extract the red earth tone response of the image detection unit.

[0030] In this embodiment, the distribution locations of red-toned pixels, yellow-toned pixels, and dark shading pixels within the image detection unit are statistically analyzed. Red-toned pixels are used to characterize candidate image features of surface responses related to lateritic bauxite deposits; yellow-toned pixels are used to help distinguish between lateritic tone transition areas and ordinary bare ground tone areas; and dark shading pixels are used to mark edge locations that may obscure the continuity of lateritic tones.

[0031] When red-toned pixels are continuously distributed within an image detection unit, and dark shadow pixels do not form closed occlusions along the edges, the image detection unit is identified as a spectral candidate unit. If dark shadow pixels form closed occlusions along the edges, the continuity of red tones in the image detection unit cannot be directly used as the basis for establishing a mineralization response, and further processing of the areas to be verified or compensated is required.

[0032] For identified spectral candidate units, a red soil tone response is generated based on the contact boundary length between the spectral candidate unit and its adjacent spectral candidate units. The more sufficient the adjacent contact, the more the red soil tone response tends to support a mineralization response; when adjacent contact is interrupted by bright pixels of bare land, road image morphology, residential area image morphology, or floodplain image morphology, the red soil tone response enters a state of pending verification.

[0033] S5 extracts the texture connectivity response and terrain transition response of the image detection unit.

[0034] In this embodiment, the texture connectivity response is used to indicate whether the texture of the patches inside the image detection unit remains continuous along the surface extension direction of the target mining area, and the terrain transition response is used to indicate whether the edge of the image detection unit conforms to the transition relationship formed by lateritic bauxite on the slope, terrace or valley edge.

[0035] During this step, the slope extension direction, valley cutting direction, and patch texture direction can be read from the base image raster and the local elevation raster registered with the base image raster. The slope extension direction is used to express the trend of surface topography, the valley cutting direction is used to express the effect of linear cutting on patch edges, and the patch texture direction is used to express the texture arrangement direction within the image detection unit.

[0036] When the patch texture direction and the slope extension direction remain continuous between adjacent image detection units, and the edges of the image detection units are not segmented into isolated fragments by the gully cutting direction, the texture connectivity response and terrain transition response tend to support the mineralization response. If an image detection unit has red-toned pixels, but the patch texture direction is discontinuous with the slope extension direction, or the edges are segmented into isolated fragments by the gully cutting direction, then the image detection unit cannot be included in the preserved patch set solely based on the red soil tone response.

[0037] S6, extract the bare ground interference response of the image detection unit.

[0038] The bare ground interference response is used to identify image conditions that pose a risk of intrusion into the delineation of ore body boundaries, such as non-mineralized laterite, bright areas of bare ground, roads, residential areas, and floodplains. This step involves performing adjacency comparisons between red-toned pixels within spectral candidate units and red-toned pixels in adjacent non-candidate units.

[0039] When a continuous region of red-toned pixels within a spectral candidate unit extends into adjacent spectral candidate units, the mineralization continuity level of that spectral candidate unit is increased. The mineralization continuity level is used to indicate the support status of the image detection unit for the intermediate layer of the mineralization response.

[0040] When a red-toned pixel within a spectral candidate unit is separated by a bright pixel in the bare land, and the separation location is consistent with the image morphology of a road, residential area, or floodplain, the red soil tone response is updated based on the separation location. After the bare land interference response is formed, subsequent intermediate layers of mineralization response will not directly use this separation location as the inner continuous region of the ore body boundary, but rather as a constraint condition for screening non-mineralized patches and edge closure correction.

[0041] S7 generates a mineralization response intermediate layer based on the consistent state between the red earth tone response, texture connectivity response, and terrain transition response.

[0042] In this step, the red hue response is used to provide candidate sources of mineralization-related hues, the texture connectivity response is used to verify whether the candidate sources have patch-level continuity, and the terrain transition response is used to verify whether the candidate sources conform to the terrain edge and slope extension relationship.

[0043] When the red soil tone response is valid, and both the texture connectivity response and the terrain transition response are valid, the corresponding image detection unit is written into the candidate patch of the mineralization response intermediate layer. This candidate patch is not the final ore body boundary, but rather an intermediate result for subsequent screening and correction.

[0044] When the red earth tone response is valid but the texture connectivity response or terrain transition response is invalid, the corresponding image detection unit is marked as a patch to be verified. The patch to be verified needs to be retained in the adjacent unit connectivity verification based on the red earth tone response, edge direction and occlusion status of the adjacent image detection units.

[0045] When the red earth tone response is not valid, but the image detection unit is located at the edge of the area to be compensated or the area to be verified, the image detection unit is not directly excluded. Instead, it is handed over to the anomaly processing step to determine whether it should be included in the final result as a boundary extension area or a verification mark.

[0046] S8, filter out non-mineralized patches in the intermediate layer of the mineralization response.

[0047] This step reads the red hue response, texture connectivity response, terrain transition response, and bare ground disturbance response for each candidate patch in the intermediate layer of the mineralization response.

[0048] When the red soil hue response, texture connectivity response, and terrain transition response are all valid, the candidate patch is added to the reserved patch set. The reserved patch set is used to store continuous patches that can be used as the basis for delineating the orebody boundary.

[0049] When the bare ground interference response covers the main edge of a candidate patch, and the outer side of this coverage edge or the direction of boundary extension does not form a continuous response consistent with the mineralization texture inside the candidate patch, the bare ground interference coverage area and its corresponding edge constraint unit are written into the non-mineralized patch set. The non-mineralized patch set is used to constrain subsequent boundary closure corrections to prevent vector boundary lines from extending incorrectly in areas of high-brightness bare ground, roads, residential areas, or floodplain image features.

[0050] When a candidate patch exhibits both red soil tone response and bare land disturbance response, but the bare land disturbance response does not cover the main edge, the candidate patch is not directly written into the set of unmineralized patches. Instead, its state to be verified is retained, and its processing result is determined based on the edge direction and the state of the adjacent retained patches during the connectivity verification of adjacent units.

[0051] S9 performs adjacent cell connectivity verification and edge closure correction on the retained patches, and outputs the boundary delineation results of lateritic bauxite ore bodies.

[0052] In this step, a boundary point sequence is generated according to the shared boundaries of adjacent retained patches in the retained patch set. The boundary point sequence is used to express the spatial orientation of the ore body boundary candidate line and is derived from the connection relationship of adjacent edge cells in the retained patch set.

[0053] When the distance between adjacent boundary points in the boundary point sequence exceeds the preset connection distance and the intermediate region has not been written into the non-mineralized patch set, connection boundary points are added according to the edge orientation of the adjacent retained patches. This addition process is only used to correct boundary breaks caused by occlusion, patch fragmentation, or missing pixels, and is not used to cross the already determined non-mineralized patch set.

[0054] When an internal hole enclosed by a sequence of boundary points belongs to a shaded cell or a vegetation-covered cell, the internal hole is marked as a region to be verified, while the outer closed boundary is preserved. The region to be verified is used to indicate that the internal hole lacks direct cell support and is not used as a basis for judging mineralization response.

[0055] When the area to be compensated is adjacent to a preserved patch in the preserved patch set, and the red soil hue response on both sides of the adjacent boundary remains continuous, the area to be compensated is used as a boundary extension area to participate in edge closure correction; when the area to be compensated is adjacent to a set of unmineralized patches, the boundary extension of the area to be compensated is stopped.

[0056] After completing the edge closure correction, the boundary point sequence is converted into vector boundary lines, and the sets of retained patches, non-mineralized patches, areas to be compensated, and areas to be verified corresponding to the vector boundary lines are written into the boundary attribute table. The boundary attribute table records the source of mineralization response, anomaly handling status, and verification markers, ultimately forming a traceable boundary delineation result for lateritic bauxite ore bodies.

[0057] Example 2

[0058] Please see Figure 1 and Figure 4 This embodiment provides a method for intelligent delineation of the boundary of lateritic bauxite ore bodies based on remote sensing images. This embodiment and Embodiment 1 are parallel implementations under the same technical object. The difference is that the generation of the intermediate layer of mineralization response emphasizes the constraint effect of terrain candidate units on texture connectivity response and terrain transition response.

[0059] In this embodiment, S1 to S3 follow the multi-band remote sensing image data acquisition, basic image raster generation, and image detection unit division methods of Embodiment 1. The difference begins before the image detection unit enters the intermediate layer generation of mineralization response, by jointly reading the basic image raster and the local elevation raster registered with the basic image raster.

[0060] The local elevation raster is used only to provide the slope extension direction and gully cutting direction. It does not change the pixel category in the base image raster, nor does it replace the source of the red soil tone response in the multi-band remote sensing image data. The base image raster provides the patch texture direction, and the registered local elevation raster provides the terrain constraint direction. Both are used together in the terrain candidate unit judgment.

[0061] S4A extracts the slope extension direction, valley cutting direction, and patch texture direction from the base image raster and the local elevation raster registered with the base image raster.

[0062] The slope extension direction is used to describe the planar extension state of the slope where the image detection unit is located; the valley cutting direction is used to describe the cutting effect of the valley on the continuity of the patch edge; the patch texture direction is used to describe the arrangement trend of similar pixels in the basic image raster within the image detection unit.

[0063] When there is a spatial inconsistency between the local elevation raster and the base image raster, alignment is first performed based on the correspondence of control points in the base image raster to ensure that the slope extension direction, valley cutting direction, and patch texture direction fall within the same image detection unit. Unaligned terrain information is not included in the terrain candidate unit evaluation to avoid misusing terrain conditions from different spatial locations in the same image detection unit.

[0064] S5A determines candidate terrain units based on terrain orientation relationships.

[0065] When the patch texture direction is continuous with the slope extension direction, and the patch edge is not segmented into isolated fragments by the gully cutting direction, the corresponding image detection unit is identified as a topographic candidate unit. If the patch texture direction is continuous with the slope extension direction, but the gully cutting direction segments the patch edge into isolated fragments, the image detection unit enters a pending verification state. The pending verification state indicates that the image detection unit may be affected by erosion zones, gully edges, or topographic shadows, and cannot be directly used as a continuous basis for boundary closure.

[0066] If the direction of the patch texture is not continuous with the direction of the slope extension, even if the image detection unit has red-toned pixels, it will not be directly used as a terrain candidate unit. Instead, it needs to be judged jointly by the red soil tone response and the connectivity verification of adjacent units.

[0067] S6A generates texture connectivity response and terrain transition response based on the edge continuity state of terrain candidate units, and writes them into the mineralization response intermediate layer.

[0068] For a terrain candidate cell, if its edge remains continuous with that of an adjacent terrain candidate cell in the slope extension direction, a texture connectivity response that supports mineralization is generated; if its edge is segmented into unconnected isolated fragments in the valley cutting direction, the texture connectivity response is not valid.

[0069] For terrain transition response, if the edge of a terrain candidate unit is located near a slope inflection line and the edge direction is consistent with the tonal transition direction of adjacent image detection units, then the terrain transition response is valid. If the edge direction conflicts with the tonal transition direction, the terrain candidate unit enters the verification state.

[0070] After the texture connectivity response and terrain transition response are output, they are combined with the red soil tone response in Example 1 and entered into the intermediate layer of mineralization response. This intermediate layer records the spectral state and the terrain continuity state together, so that when screening non-mineralized patches, it can distinguish areas with continuous red tones but affected by valley cutting, and areas with continuous terrain but insufficient tone response.

[0071] S7A, intermediate layer of mineralization response constrained by terrain candidate units.

[0072] When the red soil tone response, texture connectivity response, and terrain transition response are all valid within the same image detection unit, the corresponding image detection unit is written into the candidate patch.

[0073] When the red soil tone response is valid but the texture connectivity response is invalid, the corresponding image detection unit is written into the patch to be verified, and the patch to be verified is prohibited from forming boundary connections across the valley cutting direction during the edge closure correction stage.

[0074] When the texture connectivity response and terrain transition response are valid, but the red soil tone response cannot be confirmed due to the influence of shadowed pixels or vegetation cover pixels, the corresponding image detection unit is written into the candidate region to be reviewed. Subsequently, it is only used as a boundary extension reference when the red soil tone response of adjacent preserved patches is continuous.

[0075] Through the above processing, the terrain candidate unit does not replace the red soil tone response alone, but serves as a source of constraints for boundary continuity and boundary closure correction, so that the intermediate layer of mineralization response maintains an interpretable judgment path under the conditions of valley cutting, slope turning and patch fragmentation.

[0076] Example 3

[0077] Please see Figure 1 and Figure 5 This embodiment provides a method for intelligent delineation of the boundary of lateritic bauxite ore bodies based on remote sensing images. After forming the intermediate layer of mineralization response in Embodiment 1 or Embodiment 2, this embodiment further explains the implementation process of non-mineralized patches, areas to be compensated, and areas to be verified participating in the boundary closure correction.

[0078] S8A reads candidate patches, patches to be verified, and anomaly marker regions from the intermediate layer of the mineralization response. Candidate patches are derived from the consistent state between the red soil tone response, texture connectivity response, and terrain transition response; patches to be verified are derived from image detection units where some responses are valid but still have the effects of terrain cutting, bare land interference, or occlusion; anomaly marker regions include areas to be compensated and areas to be verified.

[0079] Before proceeding with boundary closure correction, it is first determined whether each candidate patch has a traceable source of mineralization response. Traceability means that the candidate patch can be traced back to its corresponding image detection unit and can explain the state source of the red soil tone response, texture connectivity response, terrain transition response, and bare land disturbance response in that image detection unit. Patches that cannot be traced back to their corresponding image detection unit are not included in vector boundary line generation.

[0080] S8B, based on bare ground disturbance response screening for non-mineralized patches.

[0081] When the bare land interference response covers the main edge of a candidate patch, and the outer side of this covered edge or the intended extension direction of the boundary does not form a continuous response consistent with the mineralization texture inside the candidate patch, nor does it form a continuous relationship of red soil tone with adjacent preserved patches, the edge constraint unit corresponding to the bare land interference response and its covered area are written into the set of non-mineralized patches, without using the validity of the overall texture connectivity response of the candidate patch as a criterion. This processing prevents the red-toned areas formed by the image features of roads, residential areas, floodplains, or bright pixels of bare land from being directly included in the boundary of the ore body.

[0082] When the bare land disturbance response is located in a local area at the edge of a candidate patch, but the red soil tone response is continuous with the adjacent preserved patch and the texture connectivity response is valid, the candidate patch is not directly written into the set of unmineralized patches. Instead, the position corresponding to the bare land disturbance response is avoided when the boundary point sequence is generated.

[0083] When the bare ground disturbance response overlaps with the area to be compensated, the blocking effect of the bare ground disturbance response on the boundary extension is preferentially preserved. This avoids compensation processing from crossing the identified unmineralized area due to cloud cover or shadow occlusion.

[0084] S9A generates a sequence of boundary points and performs edge closure correction.

[0085] A sequence of boundary points is generated based on the shared boundaries of adjacent preserved patches in the preserved patch set. The shared boundaries are derived from the edge cell sets of adjacent image detection units, and the sequence of boundary points is used to form the base path of the vector boundary line.

[0086] When a break occurs between adjacent boundary points in the boundary point sequence, and the intermediate region of the break is not included in the set of unmineralized patches, connecting boundary points are added based on the edge orientation of adjacent preserved patches. The purpose of connecting boundary points is to maintain the continuity of the boundary path under conditions of shading or patch fragmentation, rather than to expand the range of mineralization response.

[0087] When the intermediate region of a fault has already been incorporated into the set of unmineralized patches, no connecting boundary points are added, and this location is retained as a boundary gap or boundary turning point. Through this process, the boundary closure correction is constrained by the set of unmineralized patches, avoiding the closure of areas that clearly belong to the bare land disturbance response into the orebody boundary.

[0088] S9B processes areas awaiting compensation and areas awaiting review.

[0089] When the area to be compensated is adjacent to a preserved patch in the preserved patch set, and the laterite hue response on both sides of the adjacent boundary remains continuous, the area to be compensated is included as a boundary extension zone in the edge closure correction. This boundary extension zone only serves to connect the boundary orientation of adjacent preserved patches and is not directly used as evidence of mineralization response.

[0090] When the area to be compensated is adjacent to a set of non-mineralized patches, the boundary extension of the area to be compensated is stopped. After stopping, the area to be compensated will not be included in the region inside the vector boundary line, but will be retained in the boundary attribute table as an unextended occlusion record.

[0091] When an internal hole enclosed by a sequence of boundary points belongs to a shaded pixel or a vegetated pixel, the internal hole is marked as an area to be verified, while the outer closed boundary is preserved. The area to be verified is excluded from the mineralization response judgment, but a verification mark is added to the boundary attribute table to facilitate subsequent image verification or field verification of the area.

[0092] S9C outputs the boundary delineation results of lateritic bauxite ore bodies.

[0093] The boundary point sequence after edge closure correction is converted into vector boundary lines. Vector boundary lines are used to represent the spatial boundary of the lateritic bauxite ore body delineation results.

[0094] The set of retained patches, the set of non-mineralized patches, the region to be compensated, and the region to be verified corresponding to the vector boundary line are written into the boundary attribute table. The boundary attribute table does not change the spatial position of the vector boundary line, but is used to record the source of mineralization response, anomaly handling status, and verification markers during the boundary formation process.

[0095] When there are areas to be verified in the boundary attribute table, the boundary delineation results for lateritic bauxite ore bodies retain the outer closed boundary, while outputting the areas to be verified as attribute information. When there are unextended areas to be compensated in the boundary attribute table, these areas are output as occlusion records and do not enter the interior of the ore body boundary. Through the above output method, the ore body boundary delineation results simultaneously include vector boundary lines and traceable attributes, enabling the boundary results to be verified against the base image raster, image detection unit, and mineralization response intermediate layer.

[0096] Summary of Implementation Examples and Declaration of Protection

[0097] The above embodiments use multi-band remote sensing image data of the target mining area as input, and continuously process the following: basic image raster, image detection unit, mineralization response intermediate layer, preserved patch set, non-mineralized patch set, area to be compensated, area to be verified, boundary point sequence, vector boundary line, and boundary attribute table. This discloses an implementable process for intelligent delineation of the boundary of lateritic bauxite ore bodies. The preset pixel deviation, preset patch area, preset connection distance, response establishment conditions, and level boundaries in the embodiments can all be selected within the scope supported by the original disclosure. The specific processes, field states, processing order, judgment conditions, parameter sources, response generation methods, anomaly handling methods, and result output methods described in the embodiments are only used to explain the implementable methods of the present invention and should not limit the present invention to the specific embodiments listed. Without departing from the technical solution described in this invention and the scope of the original disclosure, any equivalent substitution, equivalent modification, equivalent combination, order adjustment, equivalent transformation of field names, equivalent succession of the executing entity, or equivalent change of the carrier form that can be conceived by a person skilled in the art should fall within the scope of protection of this patent; however, it shall not be extended to the unclaimed subject matter, nor shall the substantive correspondence of the technical objects be changed by changing the drawing number, step number, or terminology.

Claims

1. A method for intelligent delineation of the boundary of lateritic bauxite ore bodies based on remote sensing imagery, characterized in that, Includes the following steps: Acquire multi-band remote sensing image data of the target mining area, and perform geometric registration, radiometric correction and occlusion removal on the multi-band remote sensing image data to obtain a basic image raster. Image detection units are divided according to the differences in land cover, pixel adjacency, and local tone changes in the basic image raster, and a pixel set, an adjacency unit set, and an edge pixel set are established for each image detection unit. Extract the red soil tone response, texture connectivity response, terrain transition response and bare land interference response of the image detection unit, and generate a mineralization response intermediate layer based on the consistency between the red soil tone response, the texture connectivity response and the terrain transition response; In the intermediate layer of the mineralization response, non-mineralized patches with bare ground interference responses higher than the mineralization response are screened out. Adjacent cell connectivity checks and edge closure corrections are performed on the remaining patches, and the boundary delineation results of lateritic bauxite ore bodies are output.

2. The intelligent delineation method for the boundary of lateritic bauxite ore bodies based on remote sensing imagery according to claim 1, characterized in that, Acquiring multi-band remote sensing imagery data of the target mining area includes: Image scenes covering the target mining area are read from the local remote sensing image storage space, and each image scene carries imaging time, spatial resolution and band identification; Control points are extracted from the overlapping areas of adjacent image scenes. When the control point offset exceeds the preset pixel deviation, geometric registration is re-executed and the image scene is updated. When cloud pixels, dark shadow pixels, and water pixels continuously cover the edge of the mining area in the overlapping area, the corresponding coverage edge is marked as the area to be compensated, and the unoccluded pixels are written into the base image raster.

3. The intelligent delineation method for the boundary of lateritic bauxite ore bodies based on remote sensing imagery according to claim 2, characterized in that, The image detection unit is divided according to the land cover differences, pixel adjacency relationships, and local tone variations in the basic image raster, including: Initial segmentation patches are generated based on the category boundaries between bare ground pixels, vegetation cover pixels, and shadow occlusion pixels; The initially segmented patches with an area smaller than the preset patch area and adjacent to similar patches are merged into continuous patches; The continuous patches that cross valley lines, slope transition lines, or color change lines are divided into image detection units, and the adjacency relationship and edge direction of each image detection unit are recorded.

4. The intelligent delineation method for the boundary of lateritic bauxite ore bodies based on remote sensing imagery according to claim 3, characterized in that, One of the parallel limitations for generating a mineralization response intermediate layer is: The distribution locations of red-toned pixels, yellow-toned pixels, and dark shadow pixels within the image detection unit are statistically analyzed. When red-toned pixels are continuously distributed within the image detection unit and dark-shaded pixels do not form closed occlusions along the edges, the image detection unit is identified as a spectral candidate unit. Based on the contact boundary length between spectral candidate units and adjacent spectral candidate units, a red soil hue response is generated and written into the intermediate layer of the mineralization response.

5. The intelligent delineation method for the boundary of lateritic bauxite ore bodies based on remote sensing imagery according to claim 3, characterized in that, Another parallel constraint for generating the intermediate layer of mineralized response is: The slope extension direction, valley cutting direction, and patch texture direction are extracted from the base image raster and the local elevation raster registered with the base image raster. When the patch texture direction is continuous with the slope extension direction, and the patch edge is not segmented into isolated fragments by the gully cutting direction, the corresponding image detection unit is identified as a terrain candidate unit. Based on the edge continuity state of the terrain candidate unit, a texture connectivity response and a terrain transition response are generated and written into the mineralization response intermediate layer.

6. The intelligent delineation method for the boundary of lateritic bauxite ore bodies based on remote sensing imagery according to claim 4, characterized in that, The generation of red earth tone responses includes: The red tone pixels within a spectral candidate unit are compared with the red tone pixels in adjacent non-candidate units. When a continuous region of red-toned pixels within a spectral candidate unit extends to an adjacent spectral candidate unit, the mineralization continuity level of the spectral candidate unit is increased. When red-toned pixels within the spectral candidate unit are separated by bright pixels of bare land, and the separation location is consistent with the image morphology of roads, residential areas, or floodplains, the red soil tone response is updated in conjunction with the separation location.

7. The intelligent delineation method for the boundary of lateritic bauxite ore bodies based on remote sensing imagery according to claim 6, characterized in that, Screening for non-mineralized patches includes: For each candidate patch in the intermediate layer of mineralization response, read the red soil tone response, texture connectivity response, terrain transition response, and bare land disturbance response; When the red soil tone response is true and both the texture connectivity response and the terrain transition response are true, the candidate patch is written into the reserved patch set; When the bare ground interference response covers the main edge of the candidate patch and the texture connectivity response is not valid, the candidate patch is written into the set of unmineralized patches.

8. The intelligent delineation method for the boundary of lateritic bauxite ore bodies based on remote sensing imagery according to claim 7, characterized in that, Edge closure correction includes: Generate a sequence of boundary points based on the shared boundaries of adjacent preserved patches in the preserved patch set; When the distance between adjacent boundary points in the boundary point sequence exceeds the preset connection distance and the middle region is not written into the non-mineralized patch set, connection boundary points are added according to the edge direction of the adjacent retained patches. When the internal hole enclosed by the boundary point sequence belongs to a shadowed pixel or a vegetation-covered pixel, the internal hole is marked as a region to be verified, and the outer closed boundary is preserved.

9. The intelligent delineation method for the boundary of lateritic bauxite ore bodies based on remote sensing imagery according to claim 8, characterized in that, Anomaly handling for areas to be compensated and areas awaiting review includes: When the area to be compensated is adjacent to a preserved patch in the set of preserved patches, and the red soil tone response on both sides of the adjacent boundary remains continuous, the area to be compensated is used as a boundary extension area to participate in edge closure correction. When the area to be verified is surrounded by an outer closed boundary and lacks effective pixels, the area to be verified is excluded from the mineralization response judgment, and a verification mark is added to the ore body boundary delineation result. When the region to be compensated is adjacent to a set of unmineralized patches, the boundary extension of the region to be compensated is stopped.

10. The intelligent delineation method for the boundary of lateritic bauxite ore bodies based on remote sensing imagery according to claim 9, characterized in that, The output results of the boundary delineation of lateritic bauxite ore bodies include: Convert the boundary point sequence that has completed edge closure correction into a vector boundary line; Write the set of retained patches, the set of non-mineralized patches, the region to be compensated, and the region to be verified corresponding to the vector boundary line into the boundary attribute table respectively; Based on the mineralization response source, anomaly handling status, and verification marker recorded in the boundary attribute table, a traceable boundary delineation result for lateritic bauxite ore bodies is generated.

Images