An underground tunneling image optimization method and system based on virtual reality fusion

Abstract

The present application relates to the technical field of downhole image processing, and particularly relates to a downhole tunneling image optimization method and system based on virtual reality fusion. The method comprises the following steps: acquiring downhole tunneling real-time images and dust, light and equipment movement information, and spatially partitioning dust-shielded, weak-light and equipment-reflective areas in the images; then, a visual recovery layer, a brightness compensation layer and a texture balance layer are respectively constructed for collaborative optimization and fusion; finally, a dynamic virtual reality scene is constructed in combination with the equipment movement trajectory, scene mapping update, perspective fusion and depth texture correction are completed, and downhole tunneling visualized optimization images are generated. The present application realizes synchronous improvement of downhole tunneling image definition, texture continuity and scene visualized reality by hierarchical collaborative optimization of downhole dust shielding, weak-light attenuation and equipment reflective areas, and in combination with dynamic virtual reality scene mapping.

Description

Technical Field

[0001] This invention relates to the field of downhole image processing technology, and in particular to a method and system for optimizing downhole tunneling images based on virtual reality fusion. Background Technology

[0002] Underground tunneling environments are generally characterized by high dust concentrations, unstable lighting conditions, and severe surface glare from equipment. This leads to problems such as blurred areas, uneven brightness, and missing textures in acquired tunneling images, affecting the clear representation of underground scenes and the identification of spatial structures. Furthermore, existing underground image processing methods mostly focus on single-image enhancement or local denoising, lacking dynamic fusion mechanisms with virtual reality scenes. During the continuous movement of tunneling equipment and real-time changes in the environment, issues such as lagging virtual scene updates and discontinuous spatial mapping can easily arise, making it difficult to meet the application requirements of visualization and dynamic interaction in underground tunneling. Summary of the Invention

[0003] Therefore, it is necessary to provide a method and system for optimizing underground tunneling images based on virtual reality fusion to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a method for optimizing underground tunneling images based on virtual reality fusion is provided, the method comprising the following steps: Step S1: Acquire real-time images of the underground tunneling face, and simultaneously collect information on dust diffusion, light attenuation, and equipment movement trajectory in the tunneling area; Step S2: Based on dust diffusion information and light attenuation information, the real-time image of underground tunneling is spatially occluded and divided into dust interference zone, weak light transition zone and equipment reflection zone. Step S3: Construct a first visual restoration layer corresponding to the dust interference area, a second brightness compensation layer corresponding to the weak light transition area, and a third texture balance layer corresponding to the equipment reflective area, and perform image hierarchy optimization and fusion to obtain a collaboratively optimized image of the downhole scene; Step S4: Construct a dynamic virtual reality scene of underground tunneling based on the equipment motion trajectory information, and synchronously map and update the virtual reality scene according to the spatial structure changes in the collaborative optimization image of the underground scene; Step S5: Perform viewpoint fusion and depth texture correction on the updated virtual reality scene to generate an optimized visualization image of underground tunneling.

[0005] The present invention also provides a virtual reality fusion-based underground tunneling image optimization system for performing the above-described virtual reality fusion-based underground tunneling image optimization method. The virtual reality fusion-based underground tunneling image optimization system includes: The image acquisition module is used to acquire real-time images of underground tunneling at the working face, and simultaneously acquire information on dust diffusion, light attenuation, and equipment movement trajectory in the tunneling area. The image partitioning module is used to spatially partition real-time images of underground tunneling based on dust diffusion information and light attenuation information, dividing them into dust interference zone, weak light transition zone and equipment reflection zone. The hierarchical optimization module is used to construct a first visual restoration layer corresponding to the dust interference area, a second brightness compensation layer corresponding to the weak light transition area, and a third texture balance layer corresponding to the equipment reflective area, and to perform image hierarchical optimization and fusion to obtain a collaboratively optimized image of the downhole scene. The scene update module is used to construct a dynamic virtual reality scene of underground tunneling based on the equipment motion trajectory information, and to synchronously map and update the virtual reality scene according to the spatial structure changes in the collaborative optimization image of the underground scene. The image correction module is used to perform viewpoint fusion and depth texture correction on the updated virtual reality scene to generate optimized images for underground tunneling visualization.

[0006] The beneficial effects of this invention lie in its ability to collaboratively partition and process various complex interferences in underground tunneling scenarios, such as dust diffusion, low-light attenuation, and equipment reflection, thereby constructing a first visual restoration layer, a second brightness compensation layer, and a third texture balance layer. This enables targeted repair of edge continuity in dust-covered areas, brightness gradation in low-light areas, and texture details in overexposed areas, effectively improving the clarity, layering, and texture integrity of underground images. Simultaneously, by combining equipment motion trajectory information to construct a dynamic virtual reality scene, and by synchronously mapping and updating the virtual reality scene and performing depth texture correction based on the collaboratively optimized image of the underground scene, the underground tunneling scene can maintain spatial structural consistency and visual continuity during dynamic changes, thereby improving the visualization realism of the underground tunneling environment, the stability of scene interaction, and the image perception capability under complex working conditions. Attached Figure Description

[0007] Figure 1 This is a flowchart illustrating the steps of a method for optimizing underground tunneling images based on virtual reality fusion. Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S2. Figure 3 This is a schematic diagram of spatial occlusion partitioning for an underground tunneling image optimization method based on virtual reality fusion proposed in this application; Figure 4 This is a schematic diagram of image hierarchy optimization and fusion for an underground tunneling image optimization method based on virtual reality fusion proposed in this application; Figure 5 for Figure 1A detailed flowchart illustrating the implementation steps of step S3. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0008] The technical method of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0009] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0010] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0011] To achieve the above objectives, please refer to Figures 1 to 5 A method for optimizing underground tunneling images based on virtual reality fusion, the method comprising the following steps: Step S1: Acquire real-time images of the underground tunneling face, and simultaneously collect information on dust diffusion, light attenuation, and equipment movement trajectory in the tunneling area; Step S2: Based on dust diffusion information and light attenuation information, the real-time image of underground tunneling is spatially occluded and divided into dust interference zone, weak light transition zone and equipment reflection zone. Step S3: Construct a first visual restoration layer corresponding to the dust interference area, a second brightness compensation layer corresponding to the weak light transition area, and a third texture balance layer corresponding to the equipment reflective area, and perform image hierarchy optimization and fusion to obtain a collaboratively optimized image of the downhole scene; Step S4: Construct a dynamic virtual reality scene of underground tunneling based on the equipment motion trajectory information, and synchronously map and update the virtual reality scene according to the spatial structure changes in the collaborative optimization image of the underground scene; Step S5: Perform viewpoint fusion and depth texture correction on the updated virtual reality scene to generate an optimized visualization image of underground tunneling.

[0012] In one embodiment, a coal mine fully mechanized tunneling face (such as the S123 tunneling roadway, with a cross-section of approximately 5.2m × 3.8m) is used as an example. The tunneling machine (such as the EBZ series) is located at the tunnel face, equipped with a cutting head at the front end, and explosion-proof industrial cameras are installed on the top and sides. LED explosion-proof lighting is installed on the tunnel roof. Dust sensor arrays and inertial positioning modules are installed on both sides of the tunneling machine body to simultaneously acquire environmental and equipment motion status data.

[0013] Real-time images of underground tunneling at the working face are acquired. For example, the left camera captures images of the coal face area, and the right camera captures images of the support and conveyor belt area. These real-time images are uploaded to the edge processing terminal synchronously at 100ms intervals. Simultaneously, dust sensors detect dust spreading along the right front direction of the tunnel face and output the dust concentration gradient and diffusion direction vector; light sensors detect a localized decrease in brightness in the middle of the tunnel due to cutting disturbance; and the inertial module records the trajectory of the tunneling machine advancing at 0.6m / min with slight deflection.

[0014] Spatial occlusion partitioning of the image is performed based on dust diffusion information and illumination attenuation information. For details, please refer to... Figure 3 If a grayscale difference of less than 20 is detected in the right front area of ​​the image, or if the decrease in contrast compared to the baseline is greater than a preset ratio, it is identified as a dust interference candidate area (e.g., grayscale difference less than 20). If the dust interference candidate area is consistent with the dust diffusion direction, it is classified as a dust interference area. If the illuminance detected in the roof area in the middle of the roadway is less than 60% of the baseline illuminance, it is classified as a weak light transition area. If a cluster of bright saturated pixels is detected on the surface and the reflection direction is consistent with the light source, it is classified as an equipment reflection area.

[0015] Image optimization layers were constructed for the three types of regions and then fused. For details, please refer to [reference needed]. Figure 4 For example, a first visual restoration layer based on local contrast enhancement is constructed for dust-affected areas, and texture completion is performed by constraining the dust diffusion direction; a second brightness compensation layer based on neighborhood brightness interpolation is used for weak-light transition areas to restore the outline of the top anchor bolt from a blurred state to a clear edge; a third texture balancing layer is constructed for the equipment reflective areas through high brightness suppression and edge preservation filtering to suppress overexposed areas on the metal equipment surface and preserve structural boundaries. The results of the three layers are weighted and fused according to region masks to generate a collaboratively optimized image of the underground scene.

[0016] For example, in the fusion results, the cracks in the coal face that were originally invisible due to dust were restored and displayed, the arrangement structure of the roof anchor bolts was clearly visible, and the structural boundaries of the reflective areas involving metal equipment were also fully presented after suppressing the highlights.

[0017] A dynamic virtual reality scene is constructed based on the inertial trajectory of the tunneling machine. For example, the current position of the tunneling machine is mapped to (x=12.5m, y=3.1m) in the virtual tunnel coordinate system, and the cutting head posture and tunneling direction are synchronously updated in the 3D model. When the tunneling machine deviates slightly to the left, the virtual scene synchronously adjusts the relative position of the tunnel centerline and the equipment to make the virtual model consistent with the actual tunneling posture.

[0018] Viewpoint fusion and depth texture correction are performed on virtual reality scenes. For example, parallax matching is performed on images from the left and right cameras, depth maps are calculated, and the front and rear layers of the coal face are reconstructed. Depth consistency correction is also performed on dust areas to ensure that occlusion relationships conform to the real spatial structure. Finally, an optimized visualization image of underground tunneling is generated and displayed in a 3D rotatable view on the monitoring terminal, allowing operators to intuitively view dust distribution, equipment location, and coal face structure status.

[0019] As an example of the present invention, reference is made to... Figure 2 As shown, in this example, step S2 includes: Step S21: Analyze the regional brightness changes in the real-time image of underground tunneling based on the light attenuation information, determine the areas where the brightness continuously decreases, and generate weak light area data; Step S22: Identify blurred areas and texture occlusion areas in real-time underground tunneling images based on dust diffusion direction, determine the dust diffusion coverage area, and generate dust interference area data; Step S23: Detect the local bright areas in the real-time image of underground tunneling, and determine the reflective area on the surface of the equipment through the equipment motion trajectory information to generate equipment reflective area data; Step S24: Based on the data of dust interference area, low light area and equipment reflection area, the real-time image of underground tunneling is spatially occluded and partitioned to generate dust interference area, low light transition area and equipment reflection area.

[0020] In one embodiment, an underground fully mechanized tunneling face (such as the S123 tunneling roadway) in a coal mine is used as an example. Explosion-proof industrial cameras, dust concentration sensors, illuminance sensors, and inertial trajectory modules for the tunneling machine are installed in the roadway to collect multi-source information from the tunneling face area and to perform spatial occlusion zoning processing on the real-time underground tunneling images.

[0021] The brightness changes in real-time images of underground tunneling are analyzed based on illumination attenuation information. For example, during the cutting of the coal face by the tunneling machine, the illuminance sensor detects that the local illuminance in front of the roadway gradually decreases from 180 lx to approximately 108 lx, showing a continuous downward trend. The lowest illuminance is approximately 60% of the baseline illuminance, thus meeting the preset low-light judgment threshold. The image processing terminal calculates the brightness distribution of the corresponding video frame, identifies a continuous low-brightness extension zone at the boundary between the roadway roof and the facing coal face, and marks this area as a region of continuous brightness decrease, thereby generating low-light area data. For example, a low-light transition zone is formed in the upper 1 / 3 of the corresponding image.

[0022] This method identifies blurred and texture-occluded areas in real-time underground tunneling images based on the direction of dust diffusion. For example, if a dust sensor detects dust spreading along the right front of the tunnel (approximately 30°), the corresponding area in the image will show blurred edges and loss of coal and rock texture. By comparing and analyzing consecutive frames, regions with decreasing texture gradients are extracted, and the dust coverage area is determined by combining this with the dust diffusion vector field. For instance, a dust interference area of ​​approximately 2.0m × 1.5m is formed in front of the tunneling machine on the right, thus generating dust interference area data.

[0023] Locally highlighted areas in real-time images of underground tunneling are detected, and the reflective areas on the equipment surface are determined using equipment motion trajectory information. For example, a highlighted area with a grayscale value close to saturation (close to 255) is detected in areas involving metal equipment and the tunneling machine's outer shell. Simultaneously, the inertial trajectory module shows that the tunneling machine has slightly deflected at the current moment, aligning the reflection angle with the light source direction. Based on this, the highlighted area is determined to be a reflective area formed by reflection from the equipment's metal surface, and equipment reflective area data is generated, for example, mainly distributed in the left column of the support structure and the side wall area of ​​the tunneling machine.

[0024] Based on data from dust-affected areas, low-light areas, and equipment reflectivity areas, spatial occlusion zoning is performed on real-time underground tunneling images. For example, the dust-covered area on the right front is designated as the dust-affected area, the low-light areas of the tunnel roof and facing are designated as the low-light transition area, and the areas involving bright reflections from metal equipment are designated as the equipment reflectivity area.

[0025] Preferably, step S23 includes: The brightness component of the real-time image of underground tunneling is separated, and the bright pixel area in the brightness component is continuously diffused to generate a bright diffusion area. A reflective diffusion path is constructed based on the brightness attenuation direction in the high-brightness diffusion region, and the grayscale of the overexposed area is pulled back along the reflective diffusion path to generate a reflective balance region. Based on the device motion trajectory information, the position of the reflective balance area in consecutive image frames is dynamically mapped to generate a dynamic reflective mapping area. Edge texture compensation and brightness equalization are performed on the dynamic reflective mapping area to obtain the reflective area data of the device.

[0026] In one embodiment, a fully mechanized tunneling face in a coal mine (such as the S123 tunneling roadway) is used as an example for illustration. In this face, the tunneling machine is located in the facing area, and there is metallic reflection on the machine body and related metal equipment. Underground lighting units are fixedly installed on the roof of the roadway, creating localized strong light illumination conditions. Industrial cameras continuously collect real-time images of underground tunneling.

[0027] Luminance component separation is performed on real-time images of underground tunneling. For example, the RGB image is converted to HSV or YUV space, the luminance component Y is extracted, and the luminance component image is analyzed. Areas with continuous clusters of bright pixels (e.g., grayscale values ​​close to or above 240) are detected in the tunneling machine and metal lining areas, and these areas are preliminarily identified as high-brightness pixel regions.

[0028] Continuous diffusion processing is applied to the bright pixel area. For example, with the bright pixel as the center, a range of 3 to 5 pixels is extended outward according to the 8-neighborhood connectivity rule to form a bright diffusion area. This extends the original local overexposed area to the surrounding areas that may be affected by light, thereby completely covering the reflective area. Among them, 3 to 5 pixels are used to generate bright diffusion candidate areas.

[0029] A reflective diffusion path is constructed based on the brightness attenuation direction in the high-brightness diffusion region. For example, the grayscale decrease direction is gradually calculated from the center of brightness outwards, and the directional change trend of brightness gradually decreasing from 240 to 180 and 150 is identified. This direction is defined as the reflective diffusion path to describe the spatial propagation direction of light from strong to weak.

[0030] Then, the grayscale of the overexposed area is pulled back along the reflective diffusion path. For example, in the overexposed area involving metal equipment, by limiting the high brightness value and introducing the neighborhood average brightness, the grayscale value of the originally saturated metal surface is pulled back to restore it to the range where the structure can be identified, thereby generating a reflective balance area and restoring the edge structure of the support from "white and invisible" to "identifiable outline".

[0031] The reflective balance area is dynamically mapped based on the equipment's motion trajectory information. For example, the inertial trajectory module records that the tunneling machine has a lateral offset of about 0.2m and a slight rotation at the current moment. Based on this, the position of the reflective area in the continuous image frames is updated synchronously, so that the reflective area remains consistent with the actual movement of the equipment in the time series, thereby generating a dynamic reflective mapping area.

[0032] Edge texture compensation and brightness equalization are performed on the dynamically reflected areas. For example, gradient compensation is performed on the edges of metal equipment to restore structural boundaries, and local histogram equalization is performed on the metal surface to make the brightness distribution smoother and more consistent, thus finally obtaining the data of the equipment's reflective areas.

[0033] Preferably, the reflective diffusion path is constructed based on the brightness attenuation direction in the high-brightness diffusion region, and the grayscale of the overexposed area is pulled back along the reflective diffusion path, specifically as follows: The brightness components of the high-brightness diffusion region are normalized, and a brightness potential energy field is constructed based on the local pixel brightness difference to obtain a brightness gradient field that decays from the center of the high brightness outward. Using the gradient descent direction in the brightness potential field as a constraint, the high-brightness diffusion region is reconstructed into a reflective diffusion path that is consistent with the brightness decrease direction. A hierarchical brightness mapping relationship is established along the reflective diffusion path. The gray values ​​of overexposed pixels are distributed and pulled back to the average brightness of the neighborhood according to the path weight, generating a path-constrained gray redistribution result. The path consistency constraint correction is applied to the grayscale redistribution result to ensure that the grayscale change within the same reflective diffusion path satisfies a monotonically decreasing relationship, thus obtaining the reflective balance region after grayscale pullback.

[0034] In one embodiment, a fully mechanized tunneling face in a coal mine (such as the S123 tunneling roadway) is used as an example for illustration. In this face, the tunneling machine is cutting the coal wall, and the underground lighting unit is fixedly installed on the roof of the roadway, creating a localized strong light environment. The tunneling machine and the metal lining plate area show obvious bright reflection in the industrial camera image.

[0035] The brightness components of the highlighted diffusion region are normalized. For example, the image brightness values ​​are uniformly mapped to the 0-1 range, providing a unified analysis benchmark for different exposure areas. Subsequently, the brightness changes between neighboring pixels are calculated based on the local pixel brightness difference, such as calculating the gray-level difference ΔL between adjacent pixels, and a brightness potential energy field is constructed accordingly. The bright center region corresponds to a higher potential energy value, which gradually decreases outward to form a continuous brightness gradient field, thus obtaining a brightness distribution structure that decreases from the bright center outward.

[0036] Based on this, the high-brightness diffusion region is reconstructed using the gradient descent direction in the brightness potential energy field as a constraint. For example, in areas involving metal devices, a path connection is constructed from the central pixel with a gray value of approximately 240 to the surrounding gray value regions of 200 and 160, along the direction of the most stable decrease in brightness. This forms a reflective diffusion path consistent with the direction of brightness decrease, which is used to describe the diffusion direction of reflected energy in space.

[0037] A hierarchical brightness mapping relationship is established along the reflective diffusion path. For example, pixels on the path are divided into different attenuation levels such as L1, L2, and L3 according to their distance from the highlight center, and the gray values ​​of overexposed pixels are distributed and pulled back to the average brightness of the neighborhood according to the path weight. Specifically, the originally saturated bright pixels (gray value greater than 245) are gradually adjusted to the average brightness of the neighborhood (e.g., in the range of 180-200), so that the brightness gradually transitions along the path direction rather than abruptly recovers, thereby generating a path-constrained gray-level redistribution result.

[0038] The grayscale redistribution results are corrected by path consistency constraints. For example, the monotonicity of grayscale changes of all pixels within the same reflective diffusion path is checked. If a local pixel brightness is found to have a reverse sudden increase, the area is re-smoothed to ensure that the brightness change along the entire path maintains a monotonically decreasing relationship from high to low, thereby obtaining a stable reflective balance area after grayscale pullback.

[0039] In another embodiment, the luminance potential field can also determine the relative strength of luminance changes by analyzing the luminance differences of neighboring pixels for each pixel in the bright diffusion region. Areas where the grayscale value of the central pixel is significantly higher than that of surrounding pixels are considered to have high potential energy, while pixels with gradually decreasing luminance correspond to lower potential energy. In this way, a luminance potential field that decays outward from the center of the bright area can be formed. Alternatively, the trend of the difference in luminance between each pixel and its surrounding neighboring pixels can be statistically analyzed to determine the main direction of luminance decrease. Pixel values ​​along the most stable direction of luminance decrease are assigned a directional potential energy gradient, so that the potential field not only contains information about luminance levels but also indicates the spatial direction of luminance diffusion. Alternatively, the bright region can be divided into multiple luminance levels based on grayscale values, such as a core bright layer and a transitional decay layer. Pixels in the core bright layer are assigned the highest potential energy, and the potential energy values ​​of the outer transitional layers decrease progressively. Through hierarchical mapping and neighborhood weighting, a continuous and smooth luminance potential field is formed, allowing the reflective diffusion path to naturally extend along the direction of decreasing potential energy.

[0040] It is important to note that during the normalization process, the bright center region, the brightness transition region, and the normal brightness region are mapped to different relative brightness ranges, so that the brightness distribution is transformed from "absolute numerical expression" to "layered relationship expression". The bright center corresponds to the highest relative brightness level, the brightness transition region corresponds to the intermediate buffer level, and the normal region corresponds to the basic reference level.

[0041] Preferably, the path reconstruction of the high-brightness diffusion region, constrained by the gradient descent direction in the brightness potential field, includes: Perform direction vector extraction on each bright pixel in the brightness potential energy field to obtain the corresponding main direction of brightness attenuation; Using the main direction of brightness decay as the extension direction, the continuous brightness change relationship between adjacent pixels is reorganized by directional connectivity to generate multiple brightness-decreasing pixel chains. Path fusion is performed on multiple pixel chains with decreasing brightness, merging pixel chains with continuous brightness changes and consistent directions into a main reflective diffusion path; The brightness distribution around the reflection diffusion path is synchronously mapped along the main path of reflection diffusion, so that the gray diffusion trend on both sides of the path is consistent with the main path, thus generating a reflection diffusion path.

[0042] In one embodiment, a fully mechanized tunneling face in a coal mine (such as the S123 tunneling roadway) is used as an example for illustration. Inside this roadway, a tunneling machine is performing coal face cutting operations. Explosion-proof LED lighting is installed on the roof. The metal casing of the tunneling machine produces a noticeably bright reflective area in the camera image, resulting in localized image saturation and structural distortion.

[0043] Direction vector extraction is performed on each bright pixel in the luminance potential energy field. For example, in a region involving metal equipment, the luminance gradient direction vector (Δx, Δy) of the neighborhood of each bright pixel is calculated to characterize the direction of the fastest luminance decrease of that pixel, thereby obtaining the corresponding luminance attenuation main direction field.

[0044] Using the main direction of brightness attenuation as the extension direction, the directional connectivity and recombination of the continuous brightness change relationship between adjacent pixels are performed. For example, if adjacent pixels simultaneously satisfy "continuously decreasing brightness difference" and "directional angle less than a preset threshold (e.g., 15°)," they are connected into the same link, thereby generating multiple brightness-decreasing pixel chains. These pixel chains correspond to different local brightness attenuation paths in the reflective area, such as vertical surface chains and tilted reflective surface chains involving metal equipment.

[0045] Path fusion processing is performed on multiple pixel chains with decreasing brightness. For example, spatially adjacent pixel chains with the same direction are merged. When the brightness change trends of two chains are consistent (both gradually decreasing from 240 to below 180) and the spatial distance is less than a set threshold, they are merged into a unified reflective diffusion main path, thereby avoiding local path fragmentation.

[0046] The brightness distribution of the surrounding area is synchronously mapped along the main path of reflective diffusion. For example, with the main path as the center, the brightness change trend within a 3-5 pixel range on the left and right is constrained to ensure that the grayscale attenuation direction on both sides of the path is consistent with the main path, avoiding sudden changes or reverse fluctuations in local brightness, thereby generating a stable and continuous reflective diffusion path.

[0047] As an example of the present invention, reference is made to... Figure 5 As shown, step S3 in this example includes: Step S31: Based on the degree of texture blurring and edge attenuation in the dust interference area, construct a first visual restoration layer, and use the first visual restoration layer to restore the edge continuity and dehaze the grayscale of the dust-covered area in the real-time image of underground tunneling, and generate a first restored image; Step S32: Based on the brightness reduction relationship in the weak light transition zone, construct a second brightness compensation layer, and use the second brightness compensation layer to perform progressive brightness stretching and dark area grayscale compensation on the weak light area in the real-time image of underground tunneling, and generate a second compensation image. Step S33: Based on the high-brightness diffusion range and texture imbalance area in the reflective area of ​​the equipment, construct a third texture balance layer, and use the third texture balance layer to correct the texture continuity of the overexposed area in the real-time image of underground tunneling, and generate a third balanced image. Step S34: Perform inter-layer weight mapping on the first restored image, the second compensated image, and the third balanced image, and perform image hierarchy optimization and fusion based on the texture continuity and brightness transition relationship of each region to generate a collaboratively optimized image of the underground scene.

[0048] In one embodiment, a fully mechanized tunneling face in a coal mine (such as the S123 tunneling roadway) is used as an example. The tunneling machine is located at the tunnel face, and explosion-proof lighting is installed on the roof. Metal equipment is deployed along both sides of the roadway. Industrial cameras are installed on the top of the tunneling machine and the sidewalls of the roadway to simultaneously capture images of the coal face and support areas. Various interference factors exist on-site, including dust diffusion and obstruction, localized low light, and glare from the metal equipment.

[0049] A first visual restoration layer is constructed for dust-affected areas. For example, in the tunneling face area, a large amount of dust is generated due to the cutting of coal and rock, causing blurred edges and texture breaks in the image. A restoration model is constructed based on the degree of texture blurring in this area (e.g., local gradient magnitude is below a threshold) and edge attenuation (decreased edge response intensity). The dust diffusion direction constraint is used to perform edge continuation and dehazing processing on the occluded areas, so that the originally broken coal wall contour gradually restores the continuous structure, thereby generating the first restored image.

[0050] A second brightness compensation layer is constructed for the low-light transition zone. For example, in the area of ​​the tunnel roof near the rear of the tunneling machine, the overall brightness drops below the baseline value due to light source attenuation (e.g., below 50% of the original brightness). A progressive compensation model is constructed based on the brightness reduction relationship, and layered brightness stretching is performed on this area. At the same time, the gray values ​​of the dark areas are compensated, so that the anchor bolts and roof crack structures that were originally not clearly visible gradually become apparent, thus generating a second compensated image.

[0051] A third texture balancing layer is constructed for reflective areas of equipment. For example, in areas involving metal equipment such as metal columns and tunneling machine casings, there are instances of high brightness saturation and texture loss. Overexposure boundaries are identified based on the high brightness diffusion range, and texture continuity correction is performed on reflective areas. The grayscale values ​​of overexposure areas are restored to the average level of the neighborhood structure, while the edge texture structure is recovered, making the outline of the metal support recognizable, thus generating a third balanced image.

[0052] Interlayer weight mapping and fusion optimization are performed on the three types of processing results. For example, the first restored image is given a higher weight based on the degree of occlusion in the dust area, the contribution weight of the second compensation image is increased in the low-light area, and the proportion of the third balancing image is enhanced in the reflective area. Subsequently, adaptive fusion is performed based on the texture continuity and brightness transition relationship of each region to maintain a smooth transition relationship between brightness and structural changes in different regions, and finally generate a collaboratively optimized image of the underground scene.

[0053] It should be noted that in some real-time images of underground tunneling, the same area may simultaneously exhibit multiple visual degradation features such as dust interference, low-light attenuation, and equipment reflection. For such composite interference areas, spatial overlap detection is first performed on the dust interference area, low-light transition area, and equipment reflection area to construct a set of region overlap determination results. Then, region membership parameters are generated based on the response intensity of various interference features. When a pixel or local area is detected to belong to two or more interference areas simultaneously, the main processing category is determined according to the priority order of equipment reflection area, dust interference area, and low-light transition area, with equipment reflection area having the highest processing priority, followed by dust interference area, and then low-light transition area. Specifically, the interference type with the highest priority is used as the main visual correction benchmark, and its corresponding visual restoration layer, brightness compensation layer, or texture balance layer is used as the main processing layer; other interference types participate in the correction and compensation as auxiliary interference items. Furthermore, based on the regional membership values ​​corresponding to each interference type, inter-layer weight allocation coefficients are constructed. These coefficients are positively correlated with membership strength, and the outputs of different processing layers are weighted and fused to form a unified pixel response value under composite interference conditions. Finally, through the aforementioned priority-driven processing and membership-weighted fusion mechanism, collaborative correction of areas where dust, low light, and reflection coexist is achieved.

[0054] In another embodiment, step S3 above is further implemented in a modular manner, including a visual restoration module, a brightness compensation module, a texture balance module, and an adaptive fusion module.

[0055] The visual restoration module performs the dust interference area processing operation in step S31 and outputs a first restored image; the brightness compensation module performs the weak light transition area processing operation in step S32 and outputs a second compensated image; the texture balancing module performs the device reflection area processing operation in step S33 and outputs a third balanced image. Based on this, a region mapping module is set up to establish the spatial correspondence between different processing results. Specifically, the same spatial location in the first restored image, the second compensated image, and the third balanced image is calibrated and aligned at the pixel level. The processing contribution sources corresponding to each image within the same spatial region are identified, and the dust interference area, the weak light transition area, and the device reflection area are mapped to unified coordinates, thereby forming a cross-image region association structure. Further, an adaptive fusion module is set up to dynamically adjust the fusion weights according to the visual quality of different regions. Specifically, this module dynamically evaluates the contribution intensity of each image based on the texture clarity, brightness stability, and edge continuity of each region. It increases the fusion ratio of the first restored image in the dust interference area, the fusion ratio of the second compensated image in the weak light transition area, and the fusion ratio of the third balanced image in the device reflection area.

[0056] Meanwhile, the adaptive fusion module also makes secondary adjustments based on the smoothness of the visual transition between adjacent regions. When a sudden change in brightness or discontinuity in texture is detected at the boundary of a region, it automatically reduces the dominant weight of the corresponding region and enhances the fusion ratio of neighboring image information, so that the transition relationship between different image layers is changed from hard switching to continuous gradual change.

[0057] Ultimately, through the synergistic effect of the region mapping module and the adaptive fusion module, dynamic weighted fusion of multiple image results is achieved under a unified spatial coordinate system, thereby generating a collaboratively optimized image of the downhole scene that combines dust suppression, low-light enhancement, and overexposure repair capabilities.

[0058] Preferably, step S33 includes: A third texture balance layer is constructed based on the high-brightness diffusion range and texture imbalance area in the reflective area of ​​the device. The third texture balance layer is then used to perform brightness layering on the overexposed area to obtain the corresponding overexposed core area and brightness transition area. Based on the grayscale variation relationship between the overexposed core region and the brightness transition region, a texture extension direction field is constructed. The texture information around the overexposed core area is extracted along the texture extension direction field, and the extracted texture information is progressively mapped and filled into the overexposed core area to generate a texture backfill area. A grayscale continuity correction is performed on the texture backfill area to maintain a consistent grayscale gradient between the texture backfill area and the brightness transition area, generating a third balanced image.

[0059] In one embodiment, a fully mechanized tunneling face in a coal mine (such as the S123 tunneling roadway) is used as an example. In this face, the tunneling machine is located at the tunnel face, and metal equipment is arranged along both sides of the roadway. Explosion-proof lighting is installed on the roof, and industrial cameras are positioned on the top of the tunneling machine and the sidewalls of the roadway to collect real-time images of the underground tunneling process at the face and in the support area. Under strong light conditions, the metal columns of the equipment and the outer shell of the tunneling machine are prone to localized overexposure areas, accompanied by texture distortion.

[0060] Based on the high-brightness diffusion range and texture imbalance areas in the reflective area of ​​the device, brightness stratification is performed on the image. For example, the reflective area is divided into an overexposed core area and a brightness transition area. The overexposed core area is a region with a grayscale value close to saturation (e.g., >245), while the brightness transition area is an intermediate buffer zone (e.g., grayscale range of 180-240) that gradually transitions from normal brightness to high brightness. This division gives the reflective area a structured brightness hierarchy.

[0061] A texture extension direction field is constructed based on the grayscale variation relationship between the overexposed core region and the brightness transition region. For example, the gradient direction of brightness change from the transition region to the core region is statistically analyzed to determine the spatial direction in which texture information is most likely to extend, and a continuous direction vector field is formed in this direction to constrain the subsequent texture reconstruction path, ensuring that texture restoration does not deviate from the original structural trend. Specifically, the texture extension direction field can be obtained in the following ways: For example, by analyzing the grayscale variation trend between the overexposed core region and the brightness transition region, the direction in which brightness increases from the transition region to the core region is determined, and this direction is used as the possible direction of texture extension. This ensures that texture restoration extends along the most natural path of grayscale change, without deviating from the original structural trend. Another example is to identify available texture features around the overexposed core region, such as edge contours, bolt structures, and metallic textures. By analyzing the main extension directions of these neighboring textures, local direction vectors are formed, and these vectors are used to guide the filling and extension of texture within the core region. Alternatively, the grayscale gradient direction and the direction of neighboring structural features can be weighted and fused, with the weights determined based on the gradient intensity of the brightness transition and the clarity of the neighboring textures. The directional field generated in this way takes into account both the trend of brightness change and the continuity of texture, so that the texture filling conforms to the brightness logic and maintains structural coherence.

[0062] The texture information around the overexposed core area is extracted along the texture extension direction field. For example, in the edge area involving metal equipment columns, identifiable bolt structures, edge contours, and metal textures are extracted, and multiple texture paths are formed according to the direction field constraints. Then, these path information are progressively mapped and filled into the interior of the overexposed core area, so that the structure that was originally missing due to overexposure is gradually filled by the real texture of the neighboring area, thereby generating a texture backfill area.

[0063] Gray-level continuity correction is performed on the texture backfill area. For example, gray-level consistency constraints are applied to the boundary between the backfill area and the brightness transition area, so that the gray-level value smoothly transitions from the transition area (e.g., around 200) to the interior of the backfill area, avoiding abrupt changes or breaks. At the same time, the backfill texture is smoothed as a whole to keep it consistent with the brightness change trend of the original structure, thereby generating the final third balanced image.

[0064] Preferably, the third texture balance layer is used to perform brightness layering on the overexposed areas to obtain the corresponding overexposed core area and brightness transition area, including: The third texture balance layer is used to perform potential energy mapping on the brightness distribution in the overexposed area, and a brightness potential energy matrix that decreases from the center of the highlight outward is constructed. The third texture balancing layer is used to perform aggregation and convergence on the local potential energy peak regions in the brightness potential energy matrix to generate the corresponding overexposed core regions. Starting from the overexposed core area, the third texture balance layer is used to perform a ring-domain decreasing expansion on the surrounding pixels along the brightness potential energy decay trajectory to generate a brightness transition area corresponding to the brightness decay trend. The continuity of potential energy change between the overexposed core region and the brightness transition region is constrained by the third texture balance layer, so that the brightness layering result satisfies the grayscale distribution relationship that gradually decreases from the center to the outside.

[0065] In one embodiment, a fully mechanized tunneling face in a coal mine (such as the S123 tunneling roadway) is used as an example for illustration. In this face, the tunneling machine is performing coal wall cutting operations, and explosion-proof lighting is installed on the roof of the roadway. Under strong light, the metal equipment and the metal shell of the tunneling machine form obvious overexposed reflective areas. An industrial camera continuously collects real-time images of the underground tunneling in this area.

[0066] The third texture balance layer is used to perform potential energy mapping on the brightness distribution in the overexposed area. For example, the image brightness values ​​(grayscale 0-255) are converted into a brightness potential energy matrix, where bright pixels are given higher potential energy values ​​(e.g., close to 255 corresponds to high potential energy), and a potential energy propagation structure is constructed based on spatial neighborhood relationships, thereby forming a brightness potential energy field that gradually decreases from the center of brightness outwards, transforming the originally discrete reflective area into a continuous potential energy distribution structure.

[0067] Aggregation and convergence processing is performed on local potential energy peak regions in the brightness potential energy matrix. For example, in the reflective area of ​​a metal column of a metal device, multiple pixels with near-saturated gray levels form local extremum clusters in space. These clusters are then clustered and merged using neighborhood potential energy similarity analysis to remove isolated high-brightness noise points, thereby generating a structured overexposed core region, such as the continuous region with the strongest reflection at the center of the support surface.

[0068] Starting from the overexposed core area, a ring-domain decreasing expansion is performed on the surrounding pixels along the brightness potential energy decay trajectory. For example, the boundary of the core area is used as the first expansion ring, and the expansion is carried outward in increments of 1 to 3 pixels in width. The brightness weight is reduced layer by layer according to the potential energy gradient, so that the gray value of the expansion area gradually transitions from the high brightness center (such as above 240) to the medium brightness (such as the range of 180 to 210), thereby generating a brightness transition area corresponding to the brightness decay trend. Among them, 1 to 3 pixels are used for the brightness potential energy ring-domain decreasing expansion.

[0069] The continuity of potential energy changes between the overexposed core region and the brightness transition region is constrained. For example, the continuity of brightness difference between adjacent ring regions is checked to ensure that the grayscale change follows a monotonically decreasing trend from the center to the outside, avoiding sudden changes or reverse jumps in local brightness, so that the entire brightness layering result forms a stable gradient structure, thereby obtaining a clearly structured overexposed core region and brightness transition region.

[0070] Preferably, the texture extension direction field is constructed based on the grayscale variation relationship between the overexposed core region and the brightness transition region, including: The grayscale normalization is performed on the overexposed core area and the brightness transition area respectively, and the standardized grayscale value of each pixel is extracted. Gray-level difference calculation is performed along the spatial neighborhood of the brightness transition region to generate a gray-level gradient distribution that decreases outward from the overexposed core region. Based on the gray-level gradient distribution, the direction of gray-level change between adjacent pixels is vectorized to construct the initial texture extension direction candidate vector; The candidate vectors for texture extension direction are fused with directional consistency constraints to generate a texture extension direction field that is consistent with the grayscale decreasing trend.

[0071] In one embodiment, a fully mechanized tunneling face in a coal mine (such as the S123 tunneling roadway) is used as an example. In this face, the tunneling machine is located at the face, and the metal equipment is arranged along both sides of the roadway. Explosion-proof lighting is installed on the roof, and industrial cameras are deployed on the top of the tunneling machine and the sidewalls of the roadway to acquire real-time images of the underground tunneling process on the coal face and support structure. Under strong light conditions, obvious overexposed areas appear on the metal surfaces of the equipment and the outer shell of the tunneling machine, accompanied by a loss of texture information.

[0072] The overexposed core region and the brightness transition region are subjected to grayscale normalization. For example, the image grayscale values ​​are uniformly mapped to the range of 0 to 1, so that regions with different exposure intensities have a uniform calculation scale. Among them, the overexposed core region corresponds to the high grayscale value cluster area after normalization, and the brightness transition region corresponds to the buffer area that gradually transitions from medium-high grayscale to normal grayscale, thus forming a standardized brightness expression.

[0073] Gray-level difference calculations are performed along the spatial neighborhood of the brightness transition region. For example, the gray-level difference Δg between each pixel and its neighboring pixels above, below, left, and right is calculated to generate a gray-level gradient distribution that gradually decreases outward from the overexposed core region, so that the brightness change shows a continuous attenuation trend from strong to weak, thereby clarifying the spatial influence range of the overexposed region.

[0074] The direction of gray-level change between adjacent pixels is vectorized based on the gray-level gradient distribution. For example, the gray-level change rate between pixels is combined with the spatial coordinate change to construct an initial texture extension direction candidate vector (dx, dy), which is used to characterize the directional features of brightness diffusion from high-value areas to low-value areas, thereby forming multiple candidate texture extension directions.

[0075] Candidate vectors for texture extension directions are fused under direction consistency constraints. For example, vectors with directional angles exceeding a preset threshold (e.g., 15°) in their local neighborhood are eliminated or reweighted, while vectors with consistent directions that all point to the gray-level decreasing direction are fused to eliminate noise interference and local abnormal changes, thereby generating a stable and continuous texture extension direction field that is consistent with the overall gray-level decreasing trend.

[0076] The present invention also provides a virtual reality fusion-based underground tunneling image optimization system for performing the above-described virtual reality fusion-based underground tunneling image optimization method. The virtual reality fusion-based underground tunneling image optimization system includes: The image acquisition module is used to acquire real-time images of underground tunneling at the working face, and simultaneously acquire information on dust diffusion, light attenuation, and equipment movement trajectory in the tunneling area. The image partitioning module is used to spatially partition real-time images of underground tunneling based on dust diffusion information and light attenuation information, dividing them into dust interference zone, weak light transition zone and equipment reflection zone. The hierarchical optimization module is used to construct a first visual restoration layer corresponding to the dust interference area, a second brightness compensation layer corresponding to the weak light transition area, and a third texture balance layer corresponding to the equipment reflective area, and to perform image hierarchical optimization and fusion to obtain a collaboratively optimized image of the downhole scene. The scene update module is used to construct a dynamic virtual reality scene of underground tunneling based on the equipment motion trajectory information, and to synchronously map and update the virtual reality scene according to the spatial structure changes in the collaborative optimization image of the underground scene. The image correction module is used to perform viewpoint fusion and depth texture correction on the updated virtual reality scene to generate optimized images for underground tunneling visualization.

[0077] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0078] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for optimizing underground tunneling images based on virtual reality fusion, characterized in that, Includes the following steps: Step S1: Acquire real-time images of the underground tunneling face, and simultaneously collect information on dust diffusion, light attenuation, and equipment movement trajectory in the tunneling area; Step S2: Based on dust diffusion information and light attenuation information, the real-time image of underground tunneling is spatially occluded and divided into dust interference zone, weak light transition zone and equipment reflection zone. Step S3: Construct a first visual restoration layer corresponding to the dust interference area, a second brightness compensation layer corresponding to the weak light transition area, and a third texture balance layer corresponding to the equipment reflective area, and perform image hierarchy optimization and fusion to obtain a collaboratively optimized image of the downhole scene; Step S4: Construct a dynamic virtual reality scene of underground tunneling based on the equipment motion trajectory information, and synchronously map and update the virtual reality scene according to the spatial structure changes in the collaborative optimization image of the underground scene; Step S5: Perform viewpoint fusion and depth texture correction on the updated virtual reality scene to generate an optimized visualization image of underground tunneling.

2. The method for optimizing underground tunneling images based on virtual reality fusion according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Analyze the regional brightness changes in the real-time image of underground tunneling based on the light attenuation information, determine the areas where the brightness continuously decreases, and generate weak light area data; Step S22: Identify blurred areas and texture occlusion areas in real-time underground tunneling images based on dust diffusion direction, determine the dust diffusion coverage area, and generate dust interference area data; Step S23: Detect the local bright areas in the real-time image of underground tunneling, and determine the reflective area on the surface of the equipment through the equipment motion trajectory information to generate equipment reflective area data; Step S24: Based on the data of dust interference area, low light area and equipment reflection area, the real-time image of underground tunneling is spatially occluded and partitioned to generate dust interference area, low light transition area and equipment reflection area.

3. The method for optimizing underground tunneling images based on virtual reality fusion according to claim 2, characterized in that, Step S23 includes: The brightness component of the real-time image of underground tunneling is separated, and the bright pixel area in the brightness component is continuously diffused to generate a bright diffusion area. A reflective diffusion path is constructed based on the brightness attenuation direction in the high-brightness diffusion region, and the grayscale of the overexposed area is pulled back along the reflective diffusion path to generate a reflective balance region. Based on the device motion trajectory information, the position of the reflective balance area in consecutive image frames is dynamically mapped to generate a dynamic reflective mapping area. Edge texture compensation and brightness equalization are performed on the dynamic reflective mapping area to obtain the reflective area data of the device.

4. The method for optimizing underground tunneling images based on virtual reality fusion according to claim 3, characterized in that, A reflective diffusion path is constructed based on the brightness attenuation direction in the high-brightness diffusion region, and grayscale recovery is performed on the overexposed area along the reflective diffusion path as follows: The brightness components of the high-brightness diffusion region are normalized, and a brightness potential energy field is constructed based on the local pixel brightness difference to obtain a brightness gradient field that decays from the center of the high brightness outward. Using the gradient descent direction in the brightness potential field as a constraint, the high-brightness diffusion region is reconstructed into a reflective diffusion path that is consistent with the brightness decrease direction. A hierarchical brightness mapping relationship is established along the reflective diffusion path. The gray values ​​of overexposed pixels are distributed and pulled back to the average brightness of the neighborhood according to the path weight, generating a path-constrained gray redistribution result. The path consistency constraint correction is applied to the grayscale redistribution result to ensure that the grayscale change within the same reflective diffusion path satisfies a monotonically decreasing relationship, thus obtaining the reflective balance region after grayscale pullback.

5. The method for optimizing underground tunneling images based on virtual reality fusion according to claim 4, characterized in that, Using the gradient descent direction in the luminance potential field as a constraint, path reconstruction of the high-brightness diffusion region includes: Perform direction vector extraction on each bright pixel in the brightness potential energy field to obtain the corresponding main direction of brightness attenuation; Using the main direction of brightness decay as the extension direction, the continuous brightness change relationship between adjacent pixels is reorganized by directional connectivity to generate multiple brightness-decreasing pixel chains. Path fusion is performed on multiple pixel chains with decreasing brightness, merging pixel chains with continuous brightness changes and consistent directions into a main reflective diffusion path; The brightness distribution around the reflection diffusion path is synchronously mapped along the main path of reflection diffusion, so that the gray diffusion trend on both sides of the path is consistent with the main path, thus generating a reflection diffusion path.

6. The method for optimizing underground tunneling images based on virtual reality fusion according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Based on the degree of texture blurring and edge attenuation in the dust interference area, construct a first visual restoration layer, and use the first visual restoration layer to restore the edge continuity and dehaze the grayscale of the dust-covered area in the real-time image of underground tunneling, and generate a first restored image; Step S32: Based on the brightness reduction relationship in the weak light transition zone, construct a second brightness compensation layer, and use the second brightness compensation layer to perform progressive brightness stretching and dark area grayscale compensation on the weak light area in the real-time image of underground tunneling, and generate a second compensation image. Step S33: Based on the high-brightness diffusion range and texture imbalance area in the reflective area of ​​the equipment, construct a third texture balance layer, and use the third texture balance layer to correct the texture continuity of the overexposed area in the real-time image of underground tunneling, and generate a third balanced image. Step S34: Perform inter-layer weight mapping on the first restored image, the second compensated image, and the third balanced image, and perform image hierarchy optimization and fusion based on the texture continuity and brightness transition relationship of each region to generate a collaboratively optimized image of the underground scene.

7. The method for optimizing underground tunneling images based on virtual reality fusion according to claim 6, characterized in that, Step S33 includes: A third texture balance layer is constructed based on the high-brightness diffusion range and texture imbalance area in the reflective area of ​​the device. The third texture balance layer is then used to perform brightness layering on the overexposed area to obtain the corresponding overexposed core area and brightness transition area. Based on the grayscale variation relationship between the overexposed core region and the brightness transition region, a texture extension direction field is constructed. The texture information around the overexposed core area is extracted along the texture extension direction field, and the extracted texture information is progressively mapped and filled into the overexposed core area to generate a texture backfill area. A grayscale continuity correction is performed on the texture backfill area to maintain a consistent grayscale gradient between the texture backfill area and the brightness transition area, generating a third balanced image.

8. The method for optimizing underground tunneling images based on virtual reality fusion according to claim 7, characterized in that, The third texture balance layer is used to perform brightness layering on the overexposed areas, obtaining the corresponding overexposed core area and brightness transition area, including: The third texture balance layer is used to perform potential energy mapping on the brightness distribution in the overexposed area, and a brightness potential energy matrix that decreases from the center of the highlight outward is constructed. The third texture balancing layer is used to perform aggregation and convergence on the local potential energy peak regions in the brightness potential energy matrix to generate the corresponding overexposed core regions. Starting from the overexposed core area, the third texture balance layer is used to perform a ring-domain decreasing expansion on the surrounding pixels along the brightness potential energy decay trajectory to generate a brightness transition area corresponding to the brightness decay trend. The continuity of potential energy change between the overexposed core region and the brightness transition region is constrained by the third texture balance layer, so that the brightness layering result satisfies the grayscale distribution relationship that gradually decreases from the center to the outside.

9. The method for optimizing underground tunneling images based on virtual reality fusion according to claim 7, characterized in that, Based on the grayscale variation relationship between the overexposed core region and the brightness transition region, the texture extension direction field is constructed as follows: The grayscale normalization is performed on the overexposed core area and the brightness transition area respectively, and the standardized grayscale value of each pixel is extracted. Gray-level difference calculation is performed along the spatial neighborhood of the brightness transition region to generate a gray-level gradient distribution that decreases outward from the overexposed core region. Based on the gray-level gradient distribution, the direction of gray-level change between adjacent pixels is vectorized to construct the initial texture extension direction candidate vector; The candidate vectors for texture extension direction are fused with directional consistency constraints to generate a texture extension direction field that is consistent with the grayscale decreasing trend.

10. A system for optimizing underground tunneling images based on virtual reality fusion, characterized in that, For executing the virtual reality fusion-based underground tunneling image optimization method as described in claim 1, the virtual reality fusion-based underground tunneling image optimization system comprises: The image acquisition module is used to acquire real-time images of underground tunneling at the working face, and simultaneously acquire information on dust diffusion, light attenuation, and equipment movement trajectory in the tunneling area. The image partitioning module is used to spatially partition real-time images of underground tunneling based on dust diffusion information and light attenuation information, dividing them into dust interference zone, weak light transition zone and equipment reflection zone. The hierarchical optimization module is used to construct a first visual restoration layer corresponding to the dust interference area, a second brightness compensation layer corresponding to the weak light transition area, and a third texture balance layer corresponding to the equipment reflective area, and to perform image hierarchical optimization and fusion to obtain a collaboratively optimized image of the downhole scene. The scene update module is used to construct a dynamic virtual reality scene of underground tunneling based on the equipment motion trajectory information, and to synchronously map and update the virtual reality scene according to the spatial structure changes in the collaborative optimization image of the underground scene. The image correction module is used to perform viewpoint fusion and depth texture correction on the updated virtual reality scene to generate optimized images for underground tunneling visualization.

Images