A rock sample panoramic interactive display method and system

Abstract

The application discloses a rock sample panoramic interactive display method and system, relates to the technical field of information visualization, and comprises the following steps: acquiring surface image data of a rock sample, pre-processing the data, including color space unification, resolution normalization and coordinate system alignment, and obtaining multi-scale image data; constructing a macroscopic layer, a mesoscopic layer and a microscopic layer model based on the multi-scale image data; aligning the mesoscopic layer and the microscopic layer to the macroscopic layer model through feature point matching and mapping, and obtaining a hierarchical three-dimensional model; calculating a detail importance score of the rock sample according to visual saliency analysis and geological knowledge annotation, generating a detail importance atlas for marking the importance of a region, identifying a user interaction intention in real time, and selecting a corresponding rendering strategy according to the identified intention and the detail importance atlas.

Description

Technical Field

[0001] This invention relates to the field of information visualization technology, and in particular to a method and system for panoramic interactive display of rock specimens. Background Technology

[0002] The digitization and visualization of rock specimens are key technologies in geological scientific research, rock and mineral identification, and geological teaching. Existing rock specimen display technologies can be mainly divided into the following categories: Display systems based on two-dimensional high-definition images, which use multi-angle photographs of the specimen for slideshow-style or simple panoramic displays. While simple to implement, these methods lack three-dimensional information, preventing interactive observation and in-depth exploration. Model display systems based on three-dimensional surface reconstruction typically use single-scale photographs or laser scan data, generating a three-dimensional surface model of the specimen through motion reconstruction or laser point cloud processing algorithms. While providing three-dimensional interactivity, the model's level of detail is fixed, making it difficult to simultaneously present both macroscopic morphology and microscopic features. When users zoom in to observe details, the model either becomes blurry due to insufficient precision or suffers from low rendering efficiency and lag due to loading a high-precision overall model. Some advanced systems attempt to use multi-level detail model technology, pre-generating multiple models of different precision and switching between them at runtime. However, this switching is usually discrete, creating noticeable visual jumps when transitioning between different model levels, disrupting the immersive and continuous viewing experience. In terms of interactive and cognitive guidance, existing systems typically only offer basic zoom, rotate, and translate operations. Their rendering strategies are static and generic, failing to dynamically adjust based on user interaction intentions or highlight key areas using geological knowledge. This makes it difficult for users to quickly locate key points when observing complex specimens, reducing the efficiency of knowledge transfer and the effectiveness of learning and exploration. Summary of the Invention

[0003] This application provides a panoramic interactive display method and system for rock specimens, which solves the problems in the prior art such as difficulty in balancing model details and rendering efficiency, visual jumps when switching scales, lack of internal structural information, and lack of intelligent cognitive guidance. It achieves smooth visualization across the entire scale from macroscopic morphology to microscopic details and even internal structure, and improves the technical effect of smooth interaction and knowledge transfer efficiency.

[0004] This application provides a method for panoramic interactive display of rock specimens, including: S1: Acquire surface image data of rock specimens, perform preprocessing on the data including color space unification, resolution normalization and coordinate system alignment to obtain multi-scale image data; S2: Based on multi-scale image data, macroscopic, mesoscopic, and microscopic models are constructed respectively; through feature point matching and mapping, the mesoscopic and microscopic models are aligned to the macroscopic model to obtain a hierarchical three-dimensional model; S3: Based on the 3D model, according to visual saliency analysis and geological knowledge annotation, calculate the detail importance score of the rock specimen and generate a detail importance map that identifies the importance of the region; the visual saliency analysis is evaluated based on color contrast, texture complexity and geometric feature density; S4: Real-time identification of user interaction intent, and selection of corresponding rendering strategy based on the identified intent and detail importance graph; the rendering strategy is to select to render macro-level, meso-level or micro-level model details according to the user's current intent, so as to achieve a smooth transition of details.

[0005] 2. The panoramic interactive display method for rock specimens as described in claim 1, wherein the surface image data includes macroscopic, mesoscopic, and microscopic scale data of the rock specimen; The macroscopic data consists of mid-to-long-range image data collected at a distance of 50-70 cm from the specimen, used to construct the overall geometric framework of the specimen; the mesoscopic data consists of close-range image data collected at a distance of 20-30 cm from the specimen, used to supplement the local detailed features of the specimen; and the microscopic data consists of macroscopic image data collected at a distance of 5-10 cm from the specimen, used to display the microscopic morphological features and details of the specimen.

[0006] Furthermore, the construction of the macroscopic, mesoscopic, and microscopic layer models includes: Based on multi-scale image data, corresponding macroscopic basic models, mesoscopic detail models, and microscopic fine models are generated through 3D reconstruction. The macroscopic basic model is used to display the overall geometric information of the specimen, the mesoscopic detail model is used to enhance the display of the geometric and texture details of local areas of the specimen, and the microscopic fine model is used to display microscopic appearance features.

[0007] Furthermore, the visual saliency analysis is evaluated based on color contrast, texture complexity, and geometric feature density: , in, For visual saliency score, Score the color contrast. Scoring is given for texture complexity. For geometric feature density score, , and The corresponding weights are determined based on historical data, and The color contrast score is obtained by extracting the color difference between each pixel in the image and the surrounding 3×3 pixel matrix, taking the average color difference within the region and normalizing it to [0,1]. The texture complexity score is obtained by extracting the texture entropy value, contrast and correlation from the texture features of the image, weighting the sum and normalizing it to [0,1]. The geometric feature density score is obtained by counting the number of edge points and corner points in the region and normalizing it to [0,1] based on the number of points per square millimeter.

[0008] Furthermore, the geological knowledge annotation includes generating knowledge importance scores: , in, Score the importance of knowledge. A score is assigned to each core cognitive feature. This represents the upper limit of the number of features. The corresponding weights are determined based on historical data; The detail importance score is calculated based on the visual saliency score and the knowledge importance score: , in, Score for the importance of details. and For the corresponding weights, and .

[0009] Furthermore, the detail importance map is generated based on the detail importance score and visualized using a color coding rule: areas with a score ≥ 0.8 are marked in red, areas with a score [0.6, 0.8) are marked in orange, and areas with a score < 0.6 are marked in blue; the detail importance map is updated periodically, and during the update, the detail importance map is optimized by analyzing user interaction data; The user interaction data includes dwell time, number of clicks, zoom frequency, and drag trajectory.

[0010] Furthermore, the real-time identification of user interaction intent includes: classifying intents according to zoom ratio, with zoom ratio <30% indicating global browsing intent, [30, 100]% indicating general observation intent, and >100% indicating detailed study intent; locating the area of ​​interest by mouse hover position and touch point, and setting a rectangular area occupying 20% ​​of the total screen area with the screen center as the focus area; and distinguishing between focused study mode and fast browsing mode based on the user's continuous operation sequence within the focus area.

[0011] Furthermore, the method also includes achieving a smooth visual transition: acquiring a multi-scale image data sequence with continuous scale features, constructing a scale-space continuous visual transition model for the 3D model; adjusting the current rendering scale according to user interaction behavior, and generating a smooth transition visual image through the visual transition model.

[0012] Furthermore, the method also includes: acquiring the internal structure and composition data of the rock specimen; using feature point matching and iterative nearest-point algorithms to perform cross-modal registration between the internal data and the 3D model to establish an associated model; employing transparency gradient and depth-aware hybrid rendering for seamless perspective from the surface to the internal structure; the depth-aware hybrid rendering is obtained by calculating a blending factor to calculate the cumulative color of the surface and the internal structure. , in, For the final pixel colors displayed on the screen, This is a blending factor used to control the transparency of the color in the internal volume data. The color of the surface model of the rock specimen at the current location. The cumulative color of the internal structure calculated for the current position.

[0013] A panoramic interactive display system for rock specimens, the system comprising: Data acquisition module: Acquires surface image data of rock specimens at three scales: macroscopic, mesoscopic, and microscopic. Data processing module: Preprocesses the acquired image data by unifying the color space, normalizing the resolution, and aligning the coordinate system. Based on the preprocessed data, it uses different reconstruction algorithms to construct three-layer models: macroscopic, mesoscopic, and microscopic. At the same time, it uses feature point matching and mapping technology to achieve model alignment. Intelligent Analysis Module: Performs visual saliency analysis and geological knowledge annotation, calculates detail importance scores and generates detail importance maps, and analyzes user interaction data through clustering algorithms to update the maps; Rendering engine module: Real-time identification of user interaction intent, and selection of corresponding rendering strategies based on the identified intent and detail importance graph, to achieve selective rendering of model details at different levels and smooth visual transitions; Interactive control module: Receives and processes user input such as zooming, clicking, dragging, and eye tracking, locates areas of interest and identifies usage patterns, providing the rendering engine with a basis for intent judgment.

[0014] One or more technical solutions provided in this application have at least the following technical effects or advantages: By employing multi-scale data acquisition and hierarchical modeling, integrating visual saliency with geological knowledge into a detailed importance map, and using an adaptive rendering strategy based on interactive intent, the system achieves a seamless, accurate, and cognitively guided panoramic interactive display effect for users, from macroscopic morphology to microscopic details, while ensuring high rendering efficiency and smooth interaction. Attached Figure Description

[0015] Figure 1 This is a flowchart of a panoramic interactive display method for rock specimens according to an embodiment of the present invention; Figure 2 This is a diagram illustrating the architecture of a panoramic interactive display system for rock specimens, as described in an embodiment of the present invention. Detailed Implementation

[0016] To facilitate understanding of the present invention, a more complete description of this application will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to enable a more thorough and complete understanding of the disclosure of the present invention.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0018] Example 1: As Figure 1 As shown, a method for panoramic interactive display of rock specimens.

[0019] S1: Acquire surface image data of rock specimens, perform preprocessing on the data including color space unification, resolution normalization and coordinate system alignment to obtain multi-scale image data; Specifically, a three-tiered data acquisition system of macroscopic, mesoscopic, and microscopic levels is established. At a distance of 50-70 cm from the specimen, using a camera with at least 8 megapixels and a standard 24-70mm focal length lens, 12-16 images are taken around the specimen in a 360° arc, covering the entire specimen, including its overall morphology, color distribution, and main structural orientations. The overlap between adjacent images is controlled to within 50% to ensure the accuracy of feature point matching during 3D reconstruction. Each image is associated with metadata tags, including the shooting distance (50-70cm), lens focal length (24-70mm), scale level (L0), acquisition time, and specimen number. This serves as macroscopic-scale data for constructing the overall geometric framework of the specimen. At a distance of 20-30 cm from the specimen, using a 50mm medium focal length lens, 24-32 images were taken in different areas around the specimen, focusing on key regions such as mineral aggregates, rock structural surfaces, and sedimentary bedding. The overlap was increased to 55%, with an emphasis on capturing centimeter-level details. The metadata tags included the shooting distance of 20-30 cm, the lens focal length of 50mm, the scale level L1, and the corresponding feature area annotations, serving as mesoscale data to supplement the local detail features of the specimen. At a distance of 5-10 cm from the specimen, using a 100mm macro lens with a ring light to eliminate reflections and shadows, 36-48 images were taken to capture millimeter-level details such as cleavage surfaces, crystal textures, and mineral grain boundaries, with an overlap controlled at 60%. The metadata tags added the lighting parameters, focus distance, and micro-feature type, serving as microscale data to showcase the microscopic morphological features and details of the specimen.

[0020] Color analysis was performed on the acquired images, and color profiles for each image were extracted. All images were converted to the sRGB color space and then subjected to color space unification processing. Images at different scales were adjusted to a uniform pixel size for resolution normalization: macroscopic data was 2048×2048 pixels, mesoscopic data was 4096×4096 pixels, and microscopic data was 8192×8192 pixels. Camera parameter calibration was used to align the coordinate system of images at each scale to global space, eliminating shooting perspective bias. Multi-scale image data was obtained after preprocessing.

[0021] S2: Based on multi-scale image data, macroscopic, mesoscopic, and microscopic models are constructed respectively; through feature point matching and mapping, the mesoscopic and microscopic models are aligned to the macroscopic model to obtain a hierarchical three-dimensional model; Specifically, a macroscopic basic model, a mesoscopic detail model, and a microscopic fine model are constructed layered based on multi-scale image data. The macroscopic basic model displays the overall geometric information of the specimen, the mesoscopic detail model enhances the display of geometric and textural details in local areas of the specimen, and the microscopic fine model displays microscopic appearance features. The macroscopic (L0) basic model, based on macroscopic scale data, uses the Poisson reconstruction algorithm to generate a low-detail basic model with 50,000 to 100,000 polygons, constructing the overall geometric framework of the specimen and ensuring fast rendering in a global view. The mesoscopic (L1) detail model aligns mesoscopic scale data to the corresponding areas of the macroscopic basic model using feature point matching technology, supplementing local details and increasing the polygon count to 300,000 to 500,000, thus presenting richer texture and geometric features in normal observation modes. The microscopic (L2) fine model uses a dense point cloud reconstruction algorithm to process microscopic scale data, generating a fine model with 1 to 2 million polygons, used for detailed display during close-up observation. Key feature points of models at each scale are mapped across layers to ensure that the registration error between L0, L1, and L2 models is less than 0.1 mm.

[0022] S3: Based on a 3D model, and using visual saliency analysis and geological knowledge annotation, calculate the detail importance score of the rock specimen and generate a detail importance map that identifies the importance of different regions. Specifically, analyze the visual appeal of each region of the specimen using computer vision algorithms, including color contrast, texture complexity, and geometric feature density, to generate a visual saliency score. , in, For visual saliency score, Score the color contrast. Scoring is given for texture complexity. For geometric feature density score, , and The corresponding weighting coefficients are determined using historical data, and ,like =0.3, =0.4, =0.3. Color contrast score is obtained by extracting the color difference between each pixel in the image and its surrounding 3×3 pixel matrix, taking the average color difference within the region, and normalizing it to [0,1]. Texture complexity score is obtained by extracting three indicators—texture entropy, contrast, and correlation—from the image's textural features, weighted at 0.5 for entropy, 0.3 for contrast, and 0.2 for correlation, and then normalizing the sum to [0,1]. Geometric feature density score is obtained by counting the number of edge and corner points within the region, normalizing it to [0,1] based on the number of points per square millimeter.

[0023] Incorporating geological knowledge, geological knowledge annotations are applied to core cognitive features. These core cognitive features include key mineral identification characteristics such as cleavage, fracture, and cracks; important areas of rock structure such as porphyritic texture and sedimentary bedding; and typical structural features such as folds and joints. Knowledge importance score is generated. , in, Score the importance of knowledge. A score is assigned to each core cognitive feature. This represents the upper limit of the number of features. The corresponding weighting coefficients are determined based on historical data.

[0024] The detail importance score is calculated based on the visual saliency score and the knowledge importance score: , in, Score for the importance of details. and For the corresponding weights, and ,like The score is primarily based on the importance of knowledge.

[0025] Based on the detail importance score, regions with a score ≥ 0.8 are marked in red, regions with a score ≤ 0.6 and < 0.8 are marked in orange, and regions with a score < 0.6 are marked in blue, which serves as the rule for generating a detail importance map.

[0026] Collect user interaction data, including dwell time, number of clicks, zoom frequency, and drag trajectory; analyze popular observation areas using the K-means clustering algorithm; and identify usage patterns based on user operation sequences, such as zoom-dwell-click indicating focused research, and random drag-rapid zooming indicating browsing.

[0027] S4: Real-time identification of user interaction intent, and selection of corresponding rendering strategy based on the identified intent and detail importance graph; the rendering strategy is to select to render macro-level, meso-level or micro-level model details according to the user's current intent, so as to achieve a smooth transition of details.

[0028] Specifically, based on real-time user actions, the system accurately identifies interaction intents and matches corresponding rendering strategies to ensure smooth and natural interactions. Zoom ratios <30% indicate a global browsing intent, 30%-100% indicate a general observation intent, and >100% indicate a detailed study intent. The system locates the area of ​​interest by mouse hover or touch point, designating a rectangular area occupying 20% ​​of the total screen area as the focus area, with the screen center as the baseline. Three or more consecutive zooms or clicks on the same area indicate a focused study mode, while random dragging and short pauses indicate a browsing mode.

[0029] The global browsing intent renders only the L0 layer model, disables detailed lighting effects, and enables fast baked lighting to ensure smooth dragging. The regular observation intent renders the L1 layer mesoscopic details in the focus area, while retaining the L0 layer in the 80% of the surrounding area outside the focus. Layer transitions are achieved through fade-in / fade-out effects (e.g., fade-in / fade-out duration of 0.3 seconds). The detail study intent renders the L2 layer microscopic details in the focus area, prioritizing loading red areas based on detail importance scores, then orange areas, and delaying loading blue areas. Real-time global illumination is also enabled to simulate texture under natural lighting, allowing users to adjust the lighting angle to observe detail changes.

[0030] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: This application achieves full-scale information coverage of rock specimens, from overall morphology to details, by constructing a three-tiered data acquisition system encompassing macroscopic, mesoscopic, and microscopic levels. It employs layered modeling and cross-layer feature point mapping techniques to establish a hierarchical 3D model that supports on-demand loading, thereby improving rendering efficiency. By integrating visual saliency analysis and geological knowledge annotation, a detail importance map is generated. Furthermore, through real-time recognition and intent judgment of user interaction behavior, an adaptive rendering strategy is applied to match multiple modes, including global browsing, routine observation, and detailed study. Ultimately, while ensuring high-precision visual reproduction, a smooth, natural, and cognitively guided panoramic interactive display experience is achieved.

[0031] Example 2: Example 1 implemented multi-scale detail adaptive rendering, but scale switching is still based on discrete hierarchical transitions, resulting in a noticeable visual judder when users navigate between different scales. This example further supplements and explains the content of Example 1.

[0032] The method also includes achieving smooth visual transitions: acquiring multi-scale image data sequences with continuous scale features, constructing a scale-space continuous visual transition model for the 3D model; adjusting the current rendering scale according to user interaction behavior, and generating a smooth transition visual image through the visual transition model.

[0033] Specifically, a progressive acquisition path planning is adopted, starting from a point 100 cm from the specimen and gradually approaching the endpoint at predetermined steps (e.g., 5 cm). Each step completes a 360° surround shot, forming a complete sequence of images at continuous scales. The overlap between adjacent acquisition points gradually increases from 60% to 85%, using this overlap gradient to improve the accuracy of feature point matching at different scales. A continuous mapping relationship between scale parameters and spatial coordinates is established to ensure spatial consistency of data at different scales. The distance data of each acquisition point is bound to image metadata to establish a visual transition model between resolution and scale parameters, achieving continuous adaptation of resolution to scale parameters. The visual transition model is as follows: , in, scale parameter The optimal resolution for the texture should be used below. This represents the maximum resolution at the microscopic scale. and These are parameters determined through calibration experiments, used to fit the visual acuity curves of the human eye at different distances.

[0034] Representing the visual transition model as a continuous function in scale space A dense point cloud is generated by progressively acquiring a full-scale image sequence. For each point in the point cloud At different scales The visibility and geometric detail are defined by the feature point response values ​​in the image at the corresponding acquisition distance. (Function) Through a three-dimensional convolution kernel To achieve scale transformation, the smoothing effect on geometric features at different observation scales was simulated. ,in This indicates a convolution operation. When... When smaller (viewed from a distance), With a large variance, it is used to smooth out details and obtain a rough model; when When enlarged (up close), The variance is reduced, thus preserving geometric details. Pre-computed discretization Model under value And establish a trilinear interpolation mechanism. When user operations trigger the scale parameter... When changes occur, real-time updates are performed on two adjacent pre-calculated models. and ≤ ≤ Interpolate the vertex coordinates, normals, and texture coordinates of the current texture to generate the current texture. A smooth transition model is used to eliminate the abruptness of hierarchical jumps. This is established through calibration experiments. The correspondence between physical dimensions. =0.1 corresponds to a 10cm observation scale, enabling the identification of centimeter-level details; =1 corresponds to a 1cm observation scale, enabling the identification of millimeter-level details.

[0035] Real-time data collection of the user's current viewpoint distance, field of view depth, and user actions; calculation of optimal scale parameters. The scale parameter changes when the user zooms in slowly and views the content carefully. It will provide a smooth, linear transition, offering detailed control. This is especially noticeable when the user zooms in or out quickly, or switches between different viewpoints. This will employ a non-linear, smooth transition to avoid any sense of acceleration or sudden stop, ensuring visual comfort. Determining the optimal scale parameters... Afterwards, the polygon complexity of the model will be based on Values ​​are adjusted in real time to avoid abrupt changes in geometry. Texture details are filtered through continuous multi-resolution filtering, resulting in a smooth, gradual transition from blurry to sharp textures. Lighting effects are also adjusted according to scale parameters. Adaptive adjustment. When viewing at close range, real-time global illumination is enabled to accurately simulate the microscopic texture and light and shadow interaction of the rock surface; while when viewing at a distance, baked lighting is switched to prioritize smooth interaction of the overall picture.

[0036] Based on monitored user operation patterns and observation habits, machine learning algorithms are used to identify users' exploration intentions and learning goals, establishing a user roaming path prediction model. Collaborative filtering algorithms are used to calculate the cosine similarity of the current user's interest vector with other users in the discussion group, identifying similar learner groups. Interest vectors include, but are not limited to, high-frequency access scale levels, key feature types, and average interaction depth. LSTM networks are used to analyze the user's roaming trajectory sequence within a single session, predicting the area the user might be interested in next. The group preferences obtained from collaborative filtering are weighted and fused with the individual short-term interests predicted by the LSTM network analysis to generate personalized recommended roaming paths. Path planning must ensure scale continuity and visual comfort; for example, a recommended path starts at a macroscopic scale of 100cm and gradually approaches a microscopic scale of 1cm with a smooth scaling animation. When entering a region with a detail importance score ≥ 0.8, the roaming speed will automatically decrease to 30%-50% of the base speed, and the system will pause at the core observation point in that region for 2-3 seconds to allow the user to conduct in-depth observation. Conversely, in regions with a detail importance score < 0.6, the system will pass through at a faster speed, thereby optimizing the efficiency of knowledge transfer within a limited time.

[0037] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: This application constructs a scale-space continuous visual transition model by designing a progressive, continuous scale acquisition path. This model is used for interpolation between different pre-computation levels to achieve a smooth, gradual transition effect. The transition rhythm is adjusted by analyzing user interaction behavior, and personalized recommended roaming paths are generated based on user behavior prediction and collaborative filtering algorithms. This effectively eliminates abrupt changes in the visuals caused by discrete level switching, achieving a natural, coherent, and cross-scale browsing experience while ensuring rendering performance.

[0038] Example 3: Example 2 achieved a smooth transition and personalized roaming for multi-scale external rendering, but still had issues with insufficient integration and visualization of internal structural data. This example further supplements and explains the content of Example 2.

[0039] The method further includes: acquiring the internal structure and composition data of rock specimens; performing cross-modal registration of the internal data and the three-dimensional model through feature point matching and iterative nearest point algorithm to establish an associated model; designing multi-level transparency gradients to achieve seamless perspective of the internal structure from the surface to the core; and using a hybrid rendering of transparency gradient and depth perception for seamless perspective of the surface to the internal structure. Specifically, the internal structure and composition data of the rock specimens were acquired. Three-dimensional CT scans were performed on the specimens to obtain volumetric data that clearly distinguishes pores and fractures ≥5μm, reconstructing the spatial distribution and structural characteristics of the internal minerals. Simultaneously, multispectral and ultraviolet fluorescence data were acquired to identify the characteristic responses of typical minerals and establish an internal compositional distribution map. Using the center point at the top of the specimen as the origin, with the major axis as the X-axis, the minor axis as the Y-axis, and the vertical direction as the Z-axis, all modal data were mapped to the origin coordinates. Noise reduction and artifact correction were performed on the CT data; baseline correction and normalization were performed on the spectral data to ensure data quality.

[0040] In the 3D surface model, geometric feature points of edges and corners are extracted as the external contour of the specimen. In the CT volume data, key features such as density abrupt change points and crack intersections are extracted as the internal structure. A random sampling consensus algorithm is used for feature point matching to eliminate outliers and achieve preliminary alignment. Based on the preliminary alignment, an iterative nearest-point algorithm is used for registration. Transformation parameters are optimized through translation, rotation, scaling, and least squares methods to ensure that the registration error is ≤0.1mm.

[0041] After registration, a correlation model is established between surface feature regions, internal structural units, and chemical composition. For example, surface mottled structures are correlated with corresponding feldspar crystal grains and the surrounding matrix. A structured correlation map is generated based on the correlation model, supporting reverse lookup of internal structures through surface features, or locating corresponding surface regions through internal structures.

[0042] To optimize 3D visualization, a five-level transparency gradient was designed, with a corresponding transparency range configured for each layer: the surface layer covers a depth of 0 to 2 mm, with a transparency of 0.9 to 1.0, used to highlight surface texture details; the shallow layer corresponds to a depth of 2 to 5 mm, with a transparency of 0.7 to 0.9, used to display near-surface structures; the middle layer is between 5 and 10 mm, with a transparency of 0.5 to 0.7, used to present the core area structure; the deep layer is in the range of 10 to 20 mm, with a transparency of 0.3 to 0.5, used to display the secondary core structure; the core layer has a depth exceeding 20 mm, with a transparency of 0.1 to 0.3, to expose the structure of the central area. Two interactive modes are supported: smooth transition mode achieves natural transitions between layers through a 0.2-second gradual animation, while the quick switching mode has a response time controlled within 50 milliseconds to adapt to different exploration needs and operating habits.

[0043] The rendering architecture employs a hybrid approach combining surface mesh rendering and internal volume rendering: First, the surface mesh model is rendered, and its depth information is written to a depth buffer. Then, during the ray traversal for internal volume rendering, for each sample point, the shader determines its relative spatial position to the rendered surface model and calculates a blending factor. The blending factor determines the blending ratio between the current sampled color and the accumulated colors.

[0044] The sampling point is located before the surface, that is, between the eye and the surface model, meaning the line of sight has not yet reached the specimen. In this area, the influence of the surface model should be completely ignored, and priority should be given to displaying any possible internal structures. This indicates that the current volumetric data color will be accumulated in a completely opaque manner. When the sampling point is located behind the surface model, i.e., when light has penetrated the surface and entered the specimen's interior, it will be adjusted... The goal is to achieve a natural visual transition from a semi-transparent surface to full internal visibility, ensuring the internal structure is clearly and smoothly revealed. As the viewer delves deeper, the transparency should gradually decrease, making the internal structure more apparent. This gradient effect is achieved using a linear interpolation function. , in, This is a blending factor used to control the transparency of the color in the internal volume data. Set the initial base blending factor, such as 0.3, for when light just penetrates the surface to create a semi-transparent entry effect; To reach the maximum rendering depth The desired mixing factor is set to 1.0 to indicate that volumetric data are fully displayed at depth in the specimen. For the sampling point depth, The depth of the surface intersection. For maximum rendering depth, This is a normalization factor used to map the distance of the current sampling point deeper into the interior to a range of 0 to 1. When equal When, the ratio is 0; when equal When the ratio is 1, the ratio is 1.

[0045] By surface color and internal volume drawing cumulative color According to the mixing factor The calculation formula for superposition and fusion is as follows: , in, For the final pixel colors displayed on the screen, The color of the surface model of the rock specimen at the current location. For the current position, draw the cumulative color of the calculated internal structure. As the view moves deeper inside ( Approaching 1), the internal structure color dominates, thus achieving a smooth perspective from surface texture to internal structure. When the viewing depth is ≤5mm, the surface texture accounts for 70% of the weight, and the internal volume drawing accounts for 30%; when the depth is >5mm, the weights are reversed.

[0046] Key features are automatically labeled based on detail importance scores, with features of ≥0.8 being displayed first. The labeling content includes feature name, parameter value, and geological significance description.

[0047] To enable users to effectively explore the interior of specimens, the following core interactive functions are provided: Users can define planes or use custom-shaped virtual scalpels to perform real-time sectioning of the model, observing the internal structure of any cross-section. The position and angle of the section plane can be dynamically adjusted and can be shown or hidden. A depth slider control is provided, allowing users to drag the slider to dynamically adjust the physical depth range corresponding to five levels of transparency gradients, achieving seamless drilling-like observation from the surface to the core. When users click on any point on the surface or inside, the mineral composition, physical properties, and associated geological features of that point are displayed in real time based on the correlation map.

[0048] Based on the aforementioned patented method, this application also provides a panoramic interactive display system for rock specimens, such as... Figure 2 As shown, the system includes: Data acquisition module: Acquires surface image data of rock specimens at three scales: macroscopic, mesoscopic, and microscopic; acquires continuous-scale image data through a progressive path to construct a visual transition model; and obtains internal structure and composition data of rock specimens. The data processing module preprocesses the acquired image data by unifying the color space, normalizing the resolution, and aligning the coordinate system. Based on the preprocessed data, it constructs three-layer models (macro, meso, and micro) using different reconstruction algorithms, while aligning the models using feature point matching and mapping techniques. It also constructs a scale-space continuous visual transition model based on continuous-scale image data. Furthermore, it performs registration and correction on the internal structure and composition data, aligning it with the surface 3D model through cross-modal feature matching and iterative nearest-point algorithms to establish a correlation model between surface feature regions, internal structural units, and chemical components. Intelligent Analysis Module: Performs visual saliency analysis and geological knowledge annotation, calculates detail importance scores and generates detail importance maps, and analyzes user interaction data through clustering algorithms to update the maps; Rendering engine module: Real-time identification of user interaction intent, and selection of corresponding rendering strategies based on the identified intent and detail importance graph, to achieve selective rendering of model details at different levels and smooth visual transitions; Interactive control module: Receives and processes user input such as zooming, clicking, dragging, and eye tracking, locates areas of interest and identifies usage patterns, providing the rendering engine with a basis for intent judgment.

[0049] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: This application employs a combination of non-destructive techniques to achieve high-precision data acquisition of the internal structure and composition of rock specimens. Through cross-modal feature point matching and fine registration techniques, a mapping relationship between the surface 3D model and internal data is established, forming a correlation model of surface features, internal structure, and chemical composition. By using a hybrid rendering technique combining transparency gradients and depth perception, seamless perspective and dynamic observation of the internal structure from the surface to the core are achieved. This enables visualized exploration of the internal structure, identification of mineral composition, and extraction of geological parameters, enhancing the depth and breadth of rock specimen analysis.

[0050] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for panoramic interactive display of rock specimens, characterized in that, include: S1: Acquire surface image data of rock specimens, perform preprocessing on the data including color space unification, resolution normalization and coordinate system alignment to obtain multi-scale image data; S2: Based on multi-scale image data, macroscopic, mesoscopic, and microscopic models are constructed respectively; through feature point matching and mapping, the mesoscopic and microscopic models are aligned to the macroscopic model to obtain a hierarchical three-dimensional model; S3: Based on the 3D model, according to visual saliency analysis and geological knowledge annotation, calculate the detail importance score of the rock specimen and generate a detail importance map that identifies the importance of the region; S4: Real-time identification of user interaction intent, and selection of corresponding rendering strategy based on the identified intent and detail importance graph; the rendering strategy is to select to render macro-level, meso-level or micro-level model details according to the user's current intent, so as to achieve a smooth transition of details.

2. The panoramic interactive display method for rock specimens as described in claim 1, characterized in that, The surface image data includes macroscopic, mesoscopic, and microscopic scale data of the rock specimens; The macroscopic data consists of mid-to-long-range image data collected at a distance of 50-70 cm from the specimen, used to construct the overall geometric framework of the specimen; the mesoscopic data consists of close-range image data collected at a distance of 20-30 cm from the specimen, used to supplement the local detailed features of the specimen; and the microscopic data consists of macroscopic image data collected at a distance of 5-10 cm from the specimen, used to display the microscopic morphological features and details of the specimen.

3. The panoramic interactive display method for rock specimens as described in claim 1, characterized in that, The construction of macroscopic, mesoscopic, and microscopic models includes: Based on multi-scale image data, corresponding macroscopic basic models, mesoscopic detail models, and microscopic fine models are generated through 3D reconstruction. The macroscopic basic model is used to display the overall geometric information of the specimen, the mesoscopic detail model is used to enhance the display of the geometric and texture details of local areas of the specimen, and the microscopic fine model is used to display microscopic appearance features.

4. The panoramic interactive display method for rock specimens as described in claim 1, characterized in that, The visual saliency analysis is evaluated based on color contrast, texture complexity, and geometric feature density. ， in, For visual saliency score, Score the color contrast. Scoring is given for texture complexity. For geometric feature density score, , and The corresponding weights are determined based on historical data, and The color contrast score is obtained by extracting the color difference between each pixel in the image and the surrounding 3×3 pixel matrix, taking the average color difference within the region and normalizing it to [0,1]. The texture complexity score is obtained by extracting the texture entropy value, contrast and correlation from the texture features of the image, weighting the sum and normalizing it to [0,1]. The geometric feature density score is obtained by counting the number of edge points and corner points in the region and normalizing it to [0,1] based on the number of points per square millimeter.

5. The panoramic interactive display method for rock specimens as described in claim 1, characterized in that, The geological knowledge annotation includes generating knowledge importance scores: ， in, Score the importance of knowledge. A score is assigned to each core cognitive feature. This represents the upper limit of the number of features. The corresponding weights are determined based on historical data; The detail importance score is calculated based on the visual saliency score and the knowledge importance score: ， in, Score for the importance of details. and For the corresponding weights, and .

6. The panoramic interactive display method for rock specimens as described in claim 1, characterized in that, The detail importance map is generated based on detail importance scores and visualized using color coding rules: regions with scores ≥ 0.8 are marked in red, regions with scores [0.6, 0.8) are marked in orange, and regions with scores < 0.6 are marked in blue; The detail importance map is updated periodically, and during the update, the detail importance map is optimized by analyzing user interaction data; The user interaction data includes dwell time, number of clicks, zoom frequency, and drag trajectory.

7. The panoramic interactive display method for rock specimens as described in claim 1, characterized in that, The real-time identification of user interaction intent includes: classifying intents according to zoom ratio, with zoom ratio <30% indicating global browsing intent, [30, 100]% indicating general observation intent, and >100% indicating detailed study intent; locating the area of ​​interest by mouse hover position and touch point, and setting a rectangular area occupying 20% ​​of the total screen area with the screen center as the focus area; and distinguishing between focused study mode and fast browsing mode based on the user's continuous operation sequence within the focus area.

8. The panoramic interactive display method for rock specimens as described in claim 1, characterized in that, The method also includes achieving smooth visual transitions: acquiring multi-scale image data sequences with continuous scale features, constructing a scale-space continuous visual transition model for the 3D model; adjusting the current rendering scale according to user interaction behavior, and generating a smooth transition visual image through the visual transition model.

9. The panoramic interactive display method for rock specimens as described in claim 1, characterized in that, The method further includes: acquiring the internal structure and composition data of the rock specimen; performing cross-modal registration between the internal data and the 3D model through feature point matching and iterative nearest-point algorithm to establish a correlation model; and employing a hybrid rendering of transparency gradient and depth perception for seamless perspective from the surface to the internal structure. The depth perception hybrid rendering is obtained by calculating a blending factor to obtain the cumulative color of the surface and the internal structure. ， in, For the final pixel colors displayed on the screen, This is a blending factor used to control the transparency of the color in the internal volume data. The color of the surface model of the rock specimen at the current location. The cumulative color of the internal structure calculated for the current position.

10. A panoramic interactive display system for rock specimens, employing the panoramic interactive display method for rock specimens as described in any one of claims 1 to 9, characterized in that, The system includes: Data acquisition module: Acquires surface image data of rock specimens at three scales: macroscopic, mesoscopic, and microscopic. Data processing module: Preprocesses the acquired image data by unifying the color space, normalizing the resolution, and aligning the coordinate system. Based on the preprocessed data, it uses different reconstruction algorithms to construct three-layer models: macroscopic, mesoscopic, and microscopic. At the same time, it uses feature point matching and mapping technology to achieve model alignment. Intelligent Analysis Module: Performs visual saliency analysis and geological knowledge annotation, calculates detail importance scores and generates detail importance maps, and analyzes user interaction data through clustering algorithms to update the maps; Rendering engine module: Real-time identification of user interaction intent, and selection of corresponding rendering strategies based on the identified intent and detail importance graph, to achieve selective rendering of model details at different levels and smooth visual transitions; Interactive control module: Receives and processes user input such as zooming, clicking, dragging, and eye tracking, locates areas of interest and identifies usage patterns, providing the rendering engine with a basis for intent judgment.

Images