Three-dimensional scanning device and method based on large base line and double view angles

Abstract

The invention discloses a large-baseline double-view-angle three-dimensional scanning device and method and a storage medium. The device comprises a first scanning unit and a second scanning unit which are arranged on the two sides of a target space, and each scanning unit comprises a zoom imaging unit with an adjustable focal length and a laser illumination unit with an adjustable divergence half angle. The calibration module is used for establishing a geometrical relationship between the two scanning units and obtaining parameters such as an external parameter matrix R and a translation vector t. The control unit is configured to realize synchronous triggering of laser emission and camera exposure; image frames under different optical parameters are obtained by adjusting the focal length f and / or the divergence half angle beta of at least one scanning unit; and inhibiting optical mutual interference between the double view angles by adopting a time-sharing and / or coding mode. And the processing unit extracts the edge of the annular or approximately annular illumination area from the image sequence, calculates three-dimensional point clouds of the two view angles based on the triangulation principle, and performs registration and fusion on the point clouds to obtain three-dimensional point cloud data of the current cross section. The method is suitable for high-precision three-dimensional shape measurement of long and narrow spaces such as tunnels, underground pipe galleries and large pipelines, and has the advantages of being large in measurement range, high in shielding resistance, high in point cloud precision, good in long-term stability and the like. The method mainly solves the problems of geometric layout of a multi-scanning unit system level, collaborative time sequence and point cloud fusion consistency, and compared with an optical parameter optimization scheme of a single scanning unit, the method has more advantages in the aspects of coverage integrity and long-term stability.

Description

1. Technical Field

[0001] This invention belongs to the fields of 3D measurement, machine vision, and active optical scanning technology, specifically relating to a system-level collaborative control 3D scanning device, method, and storage medium based on a large baseline dual-view deployment. The solution is applicable to narrow or near-rotating spaces such as tunnels, underground utility tunnels, mine roadways, long covered bridges, and long-distance factory corridors, and is suitable for 3D reconstruction of the interior walls, geometric deformation analysis, and long-term structural health monitoring. 2. Background Technology

[0002] Three-dimensional measurement and monitoring in narrow spaces, characterized by long measurement range, numerous obstructions, poor lighting conditions, strong dust and moisture interference, and high structural repeatability, places high demands on the coverage capability, geometric stability, acquisition efficiency, and long-term consistency of the scanning system.

[0003] The existing main technical routes have the following shortcomings: (1) Point-by-point measuring equipment (total station, mechanical laser scanner) Although the single-point accuracy is high, the scanning speed is low and the equipment size is large, making it difficult to meet the needs of rapid modeling or continuous monitoring of long-distance scenes; (2) The depth accuracy of conventional stereo vision decreases rapidly with the measurement range and is easily affected by light fluctuations and lack of texture; in narrow spaces, due to the large number of repetitive structures on the walls, the matching is unstable and the overall point cloud consistency is poor. (3) The structured light / line laser scheme with a fixed divergence half angle is difficult to balance near-field and far-field illumination coverage. Near-field is prone to overflow saturation, and far-field has low signal-to-noise ratio and discontinuous illumination. At the same time, the illumination mutual interference between multi-view systems is serious, resulting in blurred point cloud edges or failure of solution. (4) Adaptive improvement of single scanning unit is still limited by single viewpoint. Although there are solutions to improve the stability of single unit point cloud by linking focal length and divergence half angle, in complex and narrow space, single viewpoint is difficult to avoid point cloud gaps caused by local occlusion, surface undulation or equipment occlusion.

[0004] In multi-view systems designed for large-scale scenarios, existing technologies still generally suffer from the following system-level bottlenecks: (1) Lack of large baseline geometric layout criteria for narrow spaces, such as not clearly distinguishing the quantitative relationship between the lateral width W and the baseline length B, and not defining the minimum intersection angle or dual field of view coverage, makes the system design lack geometric stability guarantee. (2) Lack of a system-level quality control mechanism with the quality of the fused point cloud as feedback quantity. Existing technologies are difficult to maintain the consistency and stability of point clouds in the process of multi-station and multi-device collaboration. (3) The lack of reliable time-sharing / encoding collaborative control and precise triggering synchronization schemes means that illumination interference and trigger jitter between multiple lasers and multiple scanning units are still the main factors causing depth noise, model ripples and information loss. (4) Lack of a systematic long-term external parameter drift compensation mechanism Under long-term deployment conditions, there is still a lack of operable online correction schemes for changes in external parameters caused by temperature drift, base deformation or micro displacement of the support.

[0005] Therefore, there is an urgent need for a three-dimensional scanning scheme that integrates at the system level: large baseline geometric layout, dual-view collaborative control, active parameter adjustment, system-level quality constraints and long-term stability, in order to achieve high coverage, high consistency and engineering-grade reliability in narrow spatial three-dimensional measurement and monitoring.

[0006] Related Application Statement: This invention provides an independent and complete large-baseline dual-view system-level 3D scanning solution, the implementation of which does not depend on any other technical applications. This invention may optionally be deployed in conjunction with the following inventions filed / proposed by the applicant on the same day: "Multi-target optical monitoring system, method, and storage medium based on focal length-divergence angle linkage and ROI synchronization processing," "A reflective low-power passive communication target, optical communication method, and storage medium," and "A 3D scanning device and method based on focal length-divergence angle linkage."

[0007] The above-mentioned collaborative deployment method can further enhance the overall performance of the system in terms of coverage integrity, energy efficiency management, and point cloud accuracy improvement, but it is only an optional optimization configuration and not a necessary condition for the implementation of this invention.

[0008] Special Note: The technical protection scope of this invention focuses on system-level innovations such as multi-scanning unit system-level geometric layout, collaborative control mechanism, point cloud fusion, and long-term extrinsic parameter maintenance; its implementation does not depend on any specific single-scanning unit optical control strategy. Collaborative applications with single-scanning unit-based focal length-divergence angle linkage schemes are merely optional extensions and do not constitute essential technical features of this invention. 3. Summary of the Invention 3.1 Purpose of the Invention

[0009] This invention aims to propose solutions to key system-level challenges in 3D scanning applications in tunnels, underground utility tunnels, long corridors, and other narrow spaces. These spaces are characterized by long measurement ranges, numerous obstructions, complex lighting conditions, and strong environmental disturbances, which exposes the following systemic deficiencies in existing technologies: (1) Insufficient coverage of a single view: A single scanning unit is limited by the field of view direction, which can easily cause unavoidable occlusion at the arch, corner and local structure, resulting in discontinuity or local missing points in the point cloud; (2) Multi-unit cooperative interference: When multiple scanning units work at the same time, optical interference may occur between the illumination cone beams, which blurs the edge of the ring illumination area and reduces the contrast, directly affecting the accuracy and stability of the three-dimensional solution. (3) Lack of system-level quality closed loop: Existing technologies are mostly optimized based on local image quality or single-view point cloud quality, lacking a system-level control mechanism with fused point cloud as the core feedback quantity, making it difficult to maintain point cloud consistency under large-scale, multi-measurement point, and multi-view conditions. (4) The problem of extrinsic parameter drift under long-term deployment is prominent: the scanning unit is prone to extrinsic parameter matrix drift under long-term or strong environmental changes, and lacks active detection and automatic maintenance mechanisms, which causes the cumulative error to be continuously amplified.

[0010] To address the aforementioned technical problems, this invention proposes a system-level 3D scanning device and method based on a large baseline dual-view deployment, and achieves high coverage, high consistency, and high stability in 3D measurement through the following core mechanisms: (1) System-level geometric layout mechanism: This invention introduces clear geometric layout criteria, including the quantitative relationship between the baseline length B and the horizontal width W of the target space, as well as the minimum requirement for the intersection angle of the two lines of sight, so as to ensure that the two fields of sight form a stable intersection area, thereby constructing a better triangulation geometry and reducing the sensitivity of depth calculation to noise and uncertainty. (2) Anti-interference multi-unit collaborative control mechanism: Multiple scanning units are scheduled through time-sharing, alternation or coding, and the laser emission and camera exposure are strictly synchronized in time, and the trigger deviation is controlled within the synchronization tolerance ε_t. This mechanism effectively suppresses optical interference between the two views, makes the edges of the illumination area of ​​each field of view clearly distinguishable, and ensures that the data from both views have usable quality; (3) Active parameter adjustment and system-level quality closed-loop mechanism: During the scanning process, by adjusting the focal length f and / or divergence half angle β of at least one scanning unit, the system can acquire multiple sets of image frames with complementary constraints under different optical conditions. The processing unit uses the comprehensive quality index of the fused point cloud as feedback to perform system-level linkage adjustment on the key parameters of the scanning unit, so as to achieve long-term consistency optimization and robustness enhancement between the two perspectives; (4) Automatic self-maintenance mechanism for long-term operation: The present invention further provides an automatic maintenance scheme for extrinsic parameters under long-term operation, including: extrinsic parameter drift detection, periodic correction, and automatic recalibration when necessary. This ensures the stability of extrinsic parameters in unattended or long-term deployment scenarios and avoids distortion of point clouds due to cumulative offset.

[0011] In summary, this invention aims to construct a dual-view 3D scanning solution suitable for narrow spaces, with high coverage, strong anti-interference capability, high point cloud consistency, and long-term stability at the system level, and provides its control method and storage medium accordingly to support the needs of high-precision 3D measurement and structural monitoring in multiple scenarios and multiple cycles. 3.2 System Technical Solution

[0012] The technical solution of this invention is based on a large baseline collaborative architecture with at least two scanning units. Through geometric layout optimization, collaborative triggering strategies, and system-level quality feedback, it achieves highly consistent 3D measurement. This invention provides a system-level 3D scanning device based on a large baseline dual-viewpoint, such as... Figure 1 – Figure 8 As shown, it includes core components such as a scanning unit, a system control unit, an image processing module, and a data fusion module. 3.2.1 Scanning Unit (Hardware Structure)

[0013] The system includes at least a first scanning unit and a second scanning unit, respectively positioned on the left and right sides of the target space, thus forming a geometric structure that satisfies the large baseline condition. Each scanning unit includes: (1) Imaging unit: It consists of a zoom lens and an imaging sensor, and has an adjustable focal length f; (2) Laser illumination unit: used to emit a cone-shaped illumination beam, the divergence half angle β of which is adjustable; (3) Fixed relative pose structure: The relative position and orientation between the imaging optical center and the illumination optical center are fixed by a rigid bracket to ensure the geometric stability of the dual-view triangulation. 3.2.2 System Control Unit (Timing Coordination + Active Scanning + Collaborative Decision Making)

[0014] The system control unit is responsible for the coordination and scheduling among multiple scanning units. Its core functions include: (1) Synchronous / Time-Division Acquisition Control: The synchronous triggering of the imaging unit's exposure and illumination emission is achieved through hardware triggering or precision clock synchronization; time-division, alternation, or coding methods are used for coordinated control between the two viewpoints to suppress optical interference; the synchronization deviation is controlled within the preset synchronization tolerance. ε_t Inside; (2) Active parameter adjustment (system-level active scanning): The control unit can adjust the focal length f or divergence half angle for at least one scanning unit. β This allows for the acquisition of image sequences under different optical parameters. The active scanning mechanism includes: fixed f· variable... β ,fixed β • Change f, simultaneously adjust f and β Segmented or adaptive adjustment strategies are all within the protection scope; (3) System-level closed-loop adjustment: The control unit receives the integrated quality index of the fused point cloud calculated by the image processing module. S ,according to S The changing trend of focal length f and divergence half angle β One or more parameters among exposure time, laser output power, and time-division strategy of multiple lasers are adjusted in a coordinated manner to improve the consistency, density and geometric accuracy of multi-view fused point clouds; (4) System expansion and long-term maintenance: The system can periodically detect stable marker points in the deployment scenario; when the drift of the external parameter matrix exceeds the threshold, it triggers automatic recalibration and external parameter update, thereby achieving long-term self-maintenance. 3.2.3 Image Processing Module (Calibration, Feature Extraction, 3D Solving)

[0015] The image processing module performs core computations at both the image and point cloud levels, including: (1) Joint calibration: Calculate the extrinsic parameter matrix R, translation vector t and other related quantities of the two scanning units to establish a unified coordinate system; (2) Feature extraction and dual-view 3D solution: Extract features such as the boundary of the ring or approximately ring illumination area from the images of the two scanning units; combine the imaging model and external parameters, calculate the corresponding 3D points using the principle of triangulation, and obtain the point cloud from the left and right views; (3) Point cloud registration, fusion and quality assessment: Under a unified coordinate system, the left and right point clouds are spatially weighted and fused, and the comprehensive quality index is calculated. S Feedback is sent to the control unit to perform closed-loop regulation. 3.2.4 System-level geometry and performance optimization (preferred implementation method)

[0016] To improve measurement stability, this invention can optimize the system geometry, including but not limited to: using a suitable range for the ratio of baseline length B to the lateral width W of the target space (e.g., B ∈ [0.5W, 1.0W]); and adjusting the intersection angle of the two lines of sight. θ Not less than the set minimum value (e.g.) θ ≥ 10°); Optimize the field of view overlap area. Ω coverage Cov Improve the depth stability of triangulation. G = sin( θ ). 3.3 Methodological and Technical Solutions

[0017] This invention further provides a system-level 3D scanning method based on a large baseline dual-viewpoint, comprising: (1) System layout and joint calibration: Establish the layout geometry of the scanning unit, and obtain the external parameter matrix and unified coordinate system; (2) Synchronous / Time-Division Acquisition: Acquire lighting image sequences in time-division, alternating, or encoded modes; (3) Active parameter adjustment and image acquisition: Adjust the f of at least one scanning unit and / or β To acquire multiple frames of images under different optical conditions; (4) Feature extraction and 3D solution: Extract the boundary of the illumination area and calculate the point cloud from the left and right viewpoints based on triangulation; (5) Point cloud fusion and closed-loop feedback: A unified 3D point cloud is obtained by fusion, and the quality index is calculated. S ;like S If the target is not met, perform system-level closed-loop parameter tuning; (6) (Optional) Long-term monitoring and collaborative scanning: periodically reconstruct point clouds and analyze changes; support time synchronization and spatial registration of multiple systems. 3.4 Main Technical Parameters and Variable Definitions

[0018] Symbol parameter table: Variable symbol Parameter name Definitions and Explanations f focal length The equivalent focal length of the imaging lens, measured in mm; adjustable, is one of the core parameters for changing the field of view scale during active scanning. β Diverging half angle The half-angle aperture of the laser cone illumination beam, measured in degrees, is adjustable and determines the actual range of the illuminated area. Its full angle is 2°. B Baseline length The spatial distance between the imaging optical centers of the first scanning unit and the second scanning unit is a key quantity for system-level geometric layout. W Target space horizontal width The internal width of the tunnel, utility tunnel, or similar elongated structure in the direction of deployment is used to determine the preferred configuration range of baseline B. θ Angle of sight The angle between the lines of sight drawn from the target point to the imaging optical centers of the two scanning units; used to measure the geometric stability of triangulation. R, t Extrinsic Matrix and Translation Vector The spatial transformation relationship between scanning units is obtained through joint calibration and is used to unify the dual-view point clouds into the same coordinate system. ε_t Synchronization tolerance The maximum permissible time deviation between laser emission and camera exposure, or between triggers of different scanning units, is measured in milliseconds; it is set by the system control unit. S System-level comprehensive quality indicators The feedback quantity is obtained by comprehensively evaluating the residual, consistency, density, and other characteristics of the fused point cloud and used as a system-level closed-loop parameter tuning parameter. Cov Intersection coverage The proportion of effective point cloud within the field-of-view intersection region Ω in the target area is used to quantify the effectiveness of dual-view coverage. E Fusion residual The spatial distance or residual between corresponding points after point cloud registration and fusion reflects the registration accuracy and geometric consistency. ρ Point cloud density The number of effective three-dimensional points per unit area or unit volume reflects the richness of detail in a point cloud. 3.5 Beneficial Effects of the Technical Solution

[0019] This invention constructs a complete system-level technology chain from geometric deployment and data fusion to quality closed-loop management and long-term stability maintenance, significantly improving coverage integrity, anti-occlusion capability, and long-term consistency in large-scale scenes. It has the following beneficial effects: (1) The large baseline dual-view layout significantly improves geometric stability and coverage integrity: effectively overcomes the inherent local occlusion problem of single view, and makes the depth calculation accuracy more uniform throughout the entire measurement range; (2) Time-sharing / encoded collaborative control improves system-level anti-interference capability and timing consistency: fundamentally eliminates mutual interference between the two sides of the illumination, making the light spot edge clear and the feature stable; (3) Actively adjust f / β to acquire multiple optical condition data and improve the robustness of three-dimensional solution: enhance the system's adaptability under different measurement ranges, different material reflection characteristics and complex lighting conditions; (4) Based on the system-level closed loop of the fused point cloud quality S, global consistency optimization is achieved: By comprehensively evaluating indicators such as coverage Cov, fusion residual E, and point cloud density ρ, the parameters are adjusted in a coordinated manner to ensure that high consistency and stability are maintained in the data collection of multiple stations and multiple time points. (5) Provide long-term automatic external parameter maintenance and drift compensation capabilities: By periodically identifying stable structural features and detecting external parameter drift, the system can maintain stable three-dimensional measurement performance in long-term deployment scenarios. 4. Description of the attached drawings

[0020] Note: The accompanying drawings are schematic diagrams intended to aid in understanding the principles of the present invention; modules related to multi-laser deployment, scanning execution mechanism, long-term external parameter maintenance, etc., are preferably indicated by dashed boxes or 'optional' markings in the drawings to illustrate dependent claims or preferred embodiments, and are not essential technical features of the device described in claim 1.

[0021] The attached diagram is described as follows: Figure 1 This is a schematic diagram of the overall structure of the large baseline dual-view three-dimensional scanning system of the present invention; Figure 2 This is a schematic diagram comparing the working principles of the two active scanning modes of the present invention. (a) First scanning mode: fixed focal length f, changing the divergence half angle β; (b) Second scanning mode: fixed divergence half angle β, changing the focal length f; Figure 3 This is a schematic diagram of the structure and optical axis arrangement of the scanning unit of the present invention; Figure 4 This is a front view schematic diagram of the multiple lasers arranged around the perimeter in this invention; Figure 5 This is a timing diagram of the synchronization of multi-laser time-division triggering and camera exposure in this invention; Figure 6 This is a schematic diagram illustrating the principle of dual-view feature extraction and triangulation three-dimensional solution of the present invention; Figure 7 This is a schematic diagram of the geometric relationship of the large baseline dual-view layout of the present invention; Figure 8 This is a flowchart of the three-dimensional scanning method of the present invention. 5. Detailed Implementation 5.1 Example 1: Overall System Structure (corresponding to) Figure 1 )

[0022] This invention provides a three-dimensional scanning system based on a large baseline dual-view layout, comprising: a first scanning unit (1), a second scanning unit (2), a system control unit (3), an image processing module (4), and an optional data exchange module and host computer. The system may optionally further include a scanning execution mechanism for driving the first scanning unit (1) and the second scanning unit (2) to move along the axial direction of a tunnel, underground utility tunnel, or pipeline and / or to oscillate around the device mounting point for scanning.

[0023] Each scanning unit includes: (1) Zoom imaging unit (1-1 / 2-1): Focal length f is adjustable; (2) Laser illumination unit (1-2 / 2-2): The divergence half-angle β is adjustable and can be one or more lasers; (3) Fixed optical mechanism structure (1-3 / 2-3): Ensure the relative geometric relationship between the imaging optical center and the illumination optical center. 5.2 Example 2: Scanning Unit Structure and Optical Axis Layout (corresponding to) Figure 3 )

[0024] The imaging optical axis and the illumination optical axis can be designed to be substantially parallel, or have a small angle of no more than 5°. The conical beam emitted by the laser illumination unit intersects with the inner wall of the target to form a ring or near-ring illumination area, which is covered by the field of view of the imaging unit. 5.3 Example 3: Active Scanning Strategy (corresponding to...) Figure 2 )

[0025] To improve the robustness of 3D solution, this invention can employ active optical parameter adjustment to acquire multi-frame data, including the following two typical modes: (1) First scanning mode (fixed f, adjusted β): focal length f is fixed at the preset value f 0 Laser divergence half-angle β Adjust in segments within a predetermined range; each β One frame of image is obtained accordingly; (2) Second scanning mode (fixed β, adjusted f): divergence half angle β Fixed as β 0 By adjusting the zoom level (f), the imaging resolution is gradually changed; image frames corresponding to different zoom levels are acquired. The two modes mentioned above can be used individually or in combination. 5.4 Example 4: Image Feature Extraction and 3D Solving (corresponding to...) Figure 6 )

[0026] The image feature extraction and 3D solution process includes, but is not limited to, the following steps: (1) Image preprocessing: including brightness equalization, noise suppression, etc.; (2) Ring edge extraction: The boundary pixels of the ring or near-ring illumination region are extracted by edge detection algorithms (such as Canny, LoG, structured gradient method, etc.); (3) Dual-view three-dimensional coordinate calculation: Based on the external parameter matrix R and translation vector t obtained by joint calibration, the corresponding edge points in the left and right views are projected into two spatial rays, and the three-dimensional point coordinates P(X,Y,Z) are obtained by finding the best intersection point between the ray pairs. 5.5 Example 5: Point Cloud Fusion and Geometric Registration (corresponding to...) Figure 8 )

[0027] This embodiment includes: (1) Cross-section level point cloud fusion: For the point clouds obtained by the two scanning units respectively, weighted fusion based on spatial distance, eigenvalue fusion based on point cloud covariance matrix or other applicable point cloud fusion algorithms are adopted; (2) Axial scanning and multi-section point cloud stitching: When the system moves along the axis, it scans repeatedly at multiple cross-section positions; adjacent cross-sections can be spatially aligned using point cloud registration algorithms (such as ICP algorithm or other algorithms) to construct a continuous three-dimensional point cloud model. 5.6 Example 6: Time-sharing / encoded collaborative control and time synchronization (corresponding to...) Figure 5 )

[0028] This invention supports the following strategies for suppressing multi-unit optical interference and maintaining timing consistency: (1) Time-division triggering: Each scanning unit emits lasers in turn according to the time slice; (2) Encoding trigger: Different encoding modes are used to distinguish the lighting source, and the corresponding image is decoded by the processing unit; (3) Exposure synchronization: The exposure time of all scanning units is strictly aligned with the illumination pulse, and the deviation does not exceed the synchronization tolerance. ε_t . 5.7 Example 7: 3D Scanning Method Flow (corresponding to...) Figure 8 )

[0029] The three-dimensional scanning method of the present invention may include the following steps: (1) S1: Calibration and geometric relationship establishment: Obtain the intrinsic and extrinsic parameters of the two scanning units and establish a unified coordinate system; (2) S2: Timing Coordination Setting: Execute a synchronization or time-division / encoding trigger strategy to make the exposure and illumination timing of each scanning unit consistent; (3) S3: Actively adjust optical parameters and acquire images: Adjust the f of at least one scanning unit and / or β To acquire image frames under different optical conditions; (4) S4: Ring edge extraction: Extracts the ring or near-ring edge pixels of the illuminated area; (5) S5: 3D solution and cross-sectional point cloud generation: Based on the two scanning unit images and extrinsic parameters, the cross-sectional point cloud is obtained through triangulation. (6) S6: Point cloud fusion and system-level closed-loop parameter tuning: fuse dual-view point clouds and calculate system-level quality indicators S .like S If the preset requirements are not met, adjust f. β Set the parameters and repeat steps S3–S6.

[0030] The scope of protection of this invention is limited to the technical features defined in the claims.

Claims

1. A large-baseline dual-view three-dimensional scanning device for measuring the three-dimensional morphology of the inner walls of tunnels, underground utility tunnels, large pipelines, and other narrow or near-rotating spaces, characterized in that, include: a. A first scanning unit and a second scanning unit are arranged on both sides of the target space to form a baseline, each scanning unit comprising: (i) A zoom imaging unit with an adjustable focal length f; (ii) A laser illumination unit with adjustable divergence half-angle β; b. A calibration module, configured to establish the geometric relationship between the two scanning units, and to obtain the extrinsic parameter matrix R, the translation vector t, and / or the baseline length B; c. The control unit is configured as follows: (i) Achieve synchronous triggering of laser emission and camera exposure within each scanning unit; (ii) During the scanning process, multiple sets of image frame sequences with complementary geometric constraints are acquired by coordinating and adjusting the optical operating parameters of the dual scanning units, including at least one of focal length f and divergence half angle β. (iii) A time-division and / or coding cooperative control strategy is adopted between the two scanning units to suppress mutual optical interference; d. The processing unit is configured as follows: (i) Obtain image sequences from two scanning units and extract the edges of the annular or near-annular illumination regions formed by the intersection of the conical illumination beam and the inner wall in the images; (ii) Based on the imaging model, calibration parameters and external parameters (R, t) of each scanning unit, the three-dimensional point cloud of the inner wall is calculated by triangulation. (iii) Perform registration and fusion on the point clouds of the two scanning units to generate the three-dimensional point cloud data of the current section.

2. The three-dimensional scanning device according to claim 1, characterized in that, The control unit is further configured to perform: a. First scanning mode: Adjust the divergence half-angle β while keeping the focal length f at a preset constant value; b. Second scanning mode: Adjust the focal length f while keeping the divergence half-angle β at a preset constant value; The processing unit is further configured to perform fusion on the cross-sectional point cloud data generated by the first scanning mode and the second scanning mode to improve coordinate accuracy and robustness; wherein the quality index S is calculated based on at least one of coverage, fusion residual and point cloud density, and is used for system-level closed-loop parameter tuning.

3. The three-dimensional scanning device according to claim 1 or 2, characterized in that, The baseline length B and the target space lateral width W satisfy the following relationship: B is 0.5W to 1.0W.

4. The three-dimensional scanning apparatus according to any one of claims 1 to 3, characterized in that, The device further includes a scanning actuator configured to move along the axial direction of a tunnel, underground utility tunnel, or pipeline and / or oscillate around the device mounting point for scanning; the processing unit is configured to perform registration and stitching on cross-sectional point clouds obtained from multiple measuring point locations to generate a continuous three-dimensional point cloud model.

5. The three-dimensional scanning apparatus according to any one of claims 1 to 4, characterized in that, The laser illumination unit includes multiple lasers arranged circumferentially around the optical axis of the zoom imaging unit; the control unit drives each laser using a time-division triggering and / or encoding differentiation method, and synchronizes with the camera exposure sequence; the processing unit performs correlation calculation and fusion between the illumination area in the image and the corresponding laser based on the trigger sequence or encoding information, so as to improve coverage integrity and anti-occlusion capability.

6. The three-dimensional scanning apparatus according to any one of claims 1 to 5, characterized in that, The conical illumination beam forms a structured light spot with identifiable edges on the inner wall, which is a ring-shaped or near-ring-shaped illumination area.

7. A large baseline dual-view three-dimensional scanning method, applied to the three-dimensional scanning device according to any one of claims 1 to 6, characterized in that, include: S1: Obtain the calibration parameters of the two scanning units and establish the geometric relationship; S2: Implement time-division and / or encoding-coordinated triggering of the two scanning units; S3: Adjust the focal length f and / or divergence half angle β of at least one scanning unit to acquire a series of image frames under different optical parameters; S4: Extract edge pixels of the ring or near-ring illumination region; S5: Perform triangulation to solve the 3D point cloud of the two scanning units respectively; S6: Perform registration and fusion on the two sets of point clouds to generate the current cross-sectional point cloud data.

8. The method according to claim 7, characterized in that, The first scanning mode and the second scanning mode as described in claim 2 are executed sequentially at the same cross-sectional position, and the two sets of cross-sectional point cloud data are fused.

9. A computer-readable storage medium having stored thereon program instructions that, when executed by a processor, cause the processor to perform the steps of the method of claim 7 or 8 and output three-dimensional point cloud data of the inner wall.

Images