A plug-in handheld laser scanner based on fine positioning of light spots and a method thereof

Abstract

The application relates to the technical field of laser measuring instruments, and provides a plug-in type handheld laser scanner and method based on fine positioning of a light spot, the method accesses a light spot scanning plug-in through a standardized interface of a device main body, the light spot scanning plug-in supports a transverse arrangement mode and a longitudinal arrangement mode; centimeter-level coarse positioning information is acquired by using the device main body; according to the arrangement mode of the light spot scanning plug-in, the light spot scanning plug-in is controlled to perform light spot scanning according to a corresponding scanning strategy; in the light spot scanning process, depth data and color images of a target object are synchronously collected; the centimeter-level coarse positioning information, sub-millimeter-level fine measurement data, the depth data and the color images are fused to process, and point cloud data of the target object is constructed; the detail precision and texture effect of the point cloud data are optimized based on the depth data and the color images, so that the target loss problem caused by the contradiction between efficiency and precision of a traditional laser measuring instrument is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of laser measuring instruments, and particularly relates to a plug-in handheld laser scanner based on fine positioning of light spots and a method. BACKGROUND

[0002] In the technical field of laser measuring instruments, handheld laser scanners are widely used in high-precision target modeling, complex surface detection and other scenes due to their portability and flexibility. Traditional handheld laser scanning technology usually faces the contradiction that efficiency and precision cannot be balanced: if high-density scanning is used to improve precision, it will lead to a significant decrease in data acquisition speed and an increase in device power consumption; if low-density scanning is used to ensure efficiency, it is easy to cause the loss of target surface features due to the lack of detailed information, especially in complex curved surfaces or texture change areas, the detailed precision and texture effect of point cloud data are difficult to meet the needs of high-precision scenes. In addition, the fixed arrangement of the scanning module of the traditional device cannot be flexibly adjusted according to the morphological characteristics of the target object, and there is a lack of effective fusion mechanism for coarse positioning information and fine measurement data, which leads to problems such as coordinate system conversion error accumulation and poor synchronization of multi-source data in the point cloud construction process, further affecting the final modeling accuracy. SUMMARY

[0003] The present application provides a plug-in handheld laser scanner based on fine positioning of light spots and a method, aiming at the problem that the fixed arrangement of the scanning module of the traditional device cannot be flexibly adjusted according to the morphological characteristics of the target object, and there is a lack of effective fusion mechanism for coarse positioning information and fine measurement data, which leads to problems such as coordinate system conversion error accumulation and poor synchronization of multi-source data in the point cloud construction process, further affecting the final modeling accuracy.

[0004] In a first aspect, the present application provides a plug-in handheld laser scanning method based on fine positioning of light spots, applied to a plug-in handheld laser scanner based on fine positioning of light spots, comprising:

[0005] Accessing a light spot scanning plug-in through a standardized interface of a device main body, the light spot scanning plug-in supporting a transverse arrangement mode and a longitudinal arrangement mode; acquiring centimeter-level coarse positioning information by using the device main body, the centimeter-level coarse positioning information including spatial coordinates and an orientation of the device main body;

[0006] According to the arrangement mode of the light spot scanning plug-in, combining the centimeter-level coarse positioning information to adjust a scanning path or range, controlling the light spot scanning plug-in to perform light spot scanning according to a corresponding scanning strategy, to perform sub-millimeter-level fine measurement on a target object based on the centimeter-level coarse positioning information;

[0007] In the light spot scanning process, depth data and color images of the target object are synchronously collected; the centimeter-level coarse positioning information is taken as a global spatial framework, and sub-millimeter-level fine measurement data, depth data and color images are fused to construct point cloud data of the target object; based on the depth data and color images and in combination with global spatial consistency of the centimeter-level coarse positioning information, details accuracy and texture effect of the point cloud data are optimized to obtain a point cloud model of the target object suitable for high-precision target modeling and complex surface detection scenarios.

[0008] In a second aspect, the present application provides a plug-in handheld laser scanner based on light spot fine positioning, comprising a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and realize the method provided in any embodiment of the present application when executing the computer program.

[0009] At present, there is no solution in the prior art that accesses a light spot scanning plug-in with flexible arrangement mode through a standardized interface, realizes high-efficiency and high-precision scanning in combination with centimeter-level coarse positioning and sub-millimeter-level fine measurement, or proposes a method of synchronously fusing depth data, color images and multi-precision positioning information to optimize point cloud details and texture effect. Therefore, how to realize sub-millimeter-level fine measurement while ensuring scanning efficiency and improve point cloud construction quality through multi-source data fusion has become a technical problem to be solved in this field.

[0010] The plug-in handheld laser scanner based on light spot fine positioning and the method provided in the embodiments of the present application can access a light spot scanning plug-in supporting horizontal / vertical arrangement through a standardized interface, can flexibly adjust the scanning direction according to the morphology of a target object, can adapt to diversified scanning scenarios and improve the universality of the equipment; can quickly determine a scanning area by using centimeter-level coarse positioning information provided by the equipment main body, can significantly improve local detail accuracy while ensuring overall scanning efficiency in combination with sub-millimeter-level fine measurement of the light spot scanning, and can solve the contradiction between efficiency and accuracy in traditional scanning; can construct high-precision point cloud data with color information by synchronously collecting depth data and color images and fusing the positioning information, and can effectively improve the surface detail integrity and texture reality of the point cloud model by detail accuracy optimization and texture mapping algorithm, thereby avoiding loss of target features; and can be suitable for high-precision target modeling, industrial part surface detection, cultural relic digitization protection and other scenarios with strict requirements on detail accuracy and texture effect through differential scanning strategies and data processing mechanisms.

[0011] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0012] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and all other drawings obtained by those skilled in the art without creative effort based on these drawings also belong to the protection scope of the present application.

[0013] Figure 1 is a step schematic flow chart of a plug-in handheld laser scanning method based on fine positioning of light spots provided by an embodiment of the present application;

[0014] Figure 2 is a first view structural schematic diagram of a plug-in handheld laser scanner based on fine positioning of light spots provided by an embodiment of the present application;

[0015] Figure 3 is a second view structural schematic diagram of a plug-in handheld laser scanner based on fine positioning of light spots provided by an embodiment of the present application;

[0016] Figure 4 is a third view structural schematic diagram of a plug-in handheld laser scanner based on fine positioning of light spots provided by an embodiment of the present application;

[0017] Figure 5 is a fourth view structural schematic diagram of a plug-in handheld laser scanner based on fine positioning of light spots provided by an embodiment of the present application;

[0018] Figure 6 is a structural schematic block diagram of a plug-in handheld laser scanner based on fine positioning of light spots provided by an embodiment of the present application.

[0019] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort belong to the protection scope of the present application.

[0021] The flow chart shown in the drawings is only an example description, and does not necessarily include all the contents and operations / steps, and does not necessarily be executed in the described order. For example, some operations / steps can be decomposed, combined or partially combined, so that the actual execution order can be changed according to the actual situation.

[0022] It should be understood that, in order to facilitate the clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, the terms "first", "second", etc. are used to distinguish the same items or similar items with basically the same function and role. Those skilled in the art can understand that the terms "first", "second", etc. do not limit the quantity and execution order, and the terms "first", "second", etc. also do not necessarily mean different.

[0023] It should be understood that the terms used in the present application specification are only for the purpose of describing specific embodiments and do not intend to limit the present application. As used in the present application specification and the appended claims, unless otherwise clear from the context, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0024] It should also be understood that the term "and / or" used in the present application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0025] Some embodiments of the present application will be described in detail below with reference to the accompanying drawings. The following embodiments and features in the embodiments can be combined with each other without conflict.

[0026] In the field of laser measuring instruments, handheld laser scanners are widely used in high-precision target modeling, complex surface detection and other scenarios due to their portability and flexibility. Traditional handheld laser scanning technology usually faces the contradiction that efficiency and accuracy cannot be balanced: if high-density scanning is used to improve accuracy, it will result in a significant decrease in data acquisition speed and an increase in device power consumption; if low-density scanning is used to ensure efficiency, it is easy to lose the characteristics of the target surface due to the lack of detailed information, especially in complex curved surfaces or texture changing areas, the detail accuracy and texture effect of point cloud data are difficult to meet the needs of high-precision scenarios. In addition, the scanning module arrangement of traditional devices is fixed and cannot be flexibly adjusted according to the morphological characteristics of the target object, and there is a lack of effective fusion mechanism for coarse positioning information and fine measurement data, resulting in problems such as coordinate system conversion error accumulation and poor synchronization of multi-source data in the point cloud construction process, which further affects the final modeling accuracy.

[0027] At present, there is no solution in the prior art to access a flexible switching arrangement mode spot scanning plug-in through a standardized interface, to realize efficient and high-precision scanning by combining centimeter-level coarse positioning and sub-millimeter-level fine measurement, and to propose a method of synchronously fusing depth data, color images and multi-precision positioning information to optimize point cloud details and texture effects. Therefore, how to realize sub-millimeter-level fine measurement while ensuring scanning efficiency, and improve the quality of point cloud construction through multi-source data fusion, has become a technical problem to be solved in this field.

[0028] Please refer toFigure 1 , Figure 1 is a schematic flowchart of a plug-in handheld laser scanning method based on fine positioning of a light spot provided by an embodiment of the present application. The plug-in handheld laser scanning method based on fine positioning of a light spot can be implemented by a plug-in handheld laser scanner based on fine positioning of a light spot as shown in Figures 2 to 5 The present application does not limit the specific type of the plug-in handheld laser scanner based on fine positioning of a light spot provided by the embodiments.

[0029] Specifically, as shown in Figure 1 The plug-in handheld laser scanning method based on fine positioning of a light spot provided by the present application includes steps S101 to S103, which are described in detail as follows:

[0030] Step S101. Access the light spot scanning plug-in through the standardized interface of the device body, the light spot scanning plug-in supports transverse arrangement and longitudinal arrangement; obtain centimeter-level coarse positioning information using the device body, the centimeter-level coarse positioning information includes the spatial coordinates and orientation of the device body.

[0031] Specifically, flexible switching of the scanning plug-in is achieved through modular design, and the spatial pose coarse positioning information of the device body is obtained, providing a spatial reference for subsequent fine measurement.

[0032] The standardized interface design and the light spot scanning plug-in arrangement include: the interface standardization is configured with a mechanical-electrical-data triple standardized interface through the device body (such as a handheld scanner shell). The mechanical interface adopts a magnetic / clip quick-release structure, supports two plug-in installation attitudes of transverse (the scanning direction is parallel to the device holding direction) and longitudinal (the scanning direction is perpendicular to the device holding direction), and automatically identifies the plug-in arrangement mode through a Hall sensor or a mechanical limiting structure. The electrical interface includes a power interface (5V / 12V adjustable) and a data transmission interface (USB 3.2 Gen2 / Type-C or Ethernet interface), supporting real-time data interaction between the plug-in and the main body. The data interface protocol defines a unified control instruction set (such as scanning frequency, light spot density parameters) and data output format (raw laser ranging data containing time stamp). The light spot scanning plug-in integrates a laser emission module (such as a line laser, a multi-spot array), a CMOS image sensor (for laser spot imaging), and an inertial measurement unit (IMU, optional), supporting adaptation to different target forms (such as transverse arrangement suitable for large-area plane scanning, and longitudinal arrangement suitable for deep groove and vertical curved surface scanning) through arrangement mode switching.

[0033] The centimeter-level coarse positioning information acquisition includes: sensor configuration: the device body is internally provided with a multi-modal positioning module, which includes: an inertial navigation system (INS): fusing a three-axis accelerometer and a gyroscope, the inertial navigation system (INS) can output the spatial attitude angle (pitch, roll, yaw) and the angular velocity of the device body in real time, and the update frequency is greater than or equal to 200 Hz. The visual positioning module: equipped with a binocular camera or a monocular camera + structured light, the global spatial coordinates (X, Y, Z) of the device body are obtained through feature point matching (such as the ORB algorithm) or SLAM (simultaneous localization and mapping) technology, and the positioning accuracy is ± 5 cm. UWB ultra-wideband positioning (optional): in an indoor scene, the centimeter-level positioning is realized by receiving the base station signal, and the accuracy of the visual positioning in a weak texture environment is supplemented. Data fusion: the extended Kalman filter (EKF) or particle filter algorithm is used to fuse the INS, visual / UWB data, and the centimeter-level pose information with high frequency (≥100 Hz) and low delay is output, including the global coordinates (X_w, Y_w, Z_w) of the origin of the device body coordinate system and the rotation matrix R (describing the device orientation).

[0034] Step S102. According to the arrangement mode of the light spot scanning plug-in, the scanning path or range is adjusted in combination with the centimeter-level coarse positioning information, the light spot scanning plug-in is controlled to perform light spot scanning according to the corresponding scanning strategy, and submillimeter-level fine measurement is performed on the target object based on the centimeter-level coarse positioning information.

[0035] Specifically, the scanning strategy is dynamically adjusted according to the plug-in arrangement mode, and high-precision laser ranging is realized in combination with the coarse positioning information, so that the surface details are reserved while the efficiency is ensured.

[0036] The transverse arrangement strategy is suitable for planar and shallow curved surface targets, the laser scanning direction is parallel to the device moving direction, a “zigzag” scanning path is adopted, and the scanning line spacing is dynamically adjusted (for example, according to the target curvature obtained by the coarse positioning, the spacing is less than or equal to 1 mm in a region with large curvature, and the spacing is 5-10 mm in a flat region).

[0037] The laser emission module outputs in high frequency (more than 10 kHz) pulse, and the CMOS sensor collects the light spot image at high speed, so as to realize linear laser scanning, and the single-line point cloud density is greater than or equal to 100 points / cm.

[0038] The longitudinal arrangement strategy includes: being suitable for complex structures such as deep holes and vertical edges, the scanning direction is perpendicular to the device moving direction, a “spiral” or “layered step” scanning path is adopted, and the target depth direction details are covered layer by layer. A multi-spot array (such as a 3*3 spot matrix) is enabled, multiple laser points are projected at the same time, multiple depth points are synchronously obtained through the triangulation method, and the point cloud density of the shielding area such as the deep groove bottom is improved.

[0039] Laser triangulation principle briefs: after the laser spot is diffused and reflected by the target surface, it is imaged by the CMOS sensor, and the three-dimensional coordinates (X_c, Y_c, Z_c) of the target point are calculated through the pixel coordinates of the spot, the baseline distance and the included angle of the laser and the camera, and the accuracy can reach ±0.1mm (laboratory environment).

[0040] Dynamic accuracy compensation uses the coarse positioning information (device pose) of step S101 to real-time calibrate the conversion relationship between the laser scanning coordinate system and the global coordinate system, and compensate for the random error caused by device jitter. For the change of the distance between the device and the target during scanning (such as 10-100cm working distance), the laser power and camera exposure parameters are automatically adjusted to ensure the imaging clarity of the spot and avoid measurement deviation caused by light changes.

[0041] Intelligent sampling optimization is based on the preliminary shape of the target obtained by coarse positioning (such as sparse point cloud constructed by early SLAM), and automatically increases the scanning frequency (sampling interval ≤0.5mm) in the curvature mutation area (such as edge, hole edge), and reduces the sampling frequency (interval 5mm) in the flat area, balancing efficiency and accuracy.

[0042] At the same time, for the spot scanning, step S102 is a major breakthrough of the prior art, because the big pain point of the existing spot scanning is that it is easy to lose during scanning, which requires repeated scanning back and forth, causing great waste and low efficiency. At the same time, for the handheld laser scanner, the method combined with this step can also scan small objects, expanding its application range.

[0043] Step S103. In the process of spot scanning, the depth data and color image of the target object are synchronously collected; the centimeter-level coarse positioning information is taken as a global spatial framework, and the sub-millimeter-level fine measurement data, depth data and color image are fused to construct the point cloud data of the target object; based on the depth data and color image, and combined with the global spatial consistency of the centimeter-level coarse positioning information, the detail accuracy and texture effect of the point cloud data are optimized to obtain a target object point cloud model suitable for high-precision target modeling and complex surface detection scenarios.

[0044] Specifically, the depth, color and positioning data are synchronously collected, the coordinate system error is eliminated through fusion processing, the point cloud details and texture authenticity are enhanced, and the feature loss problem in traditional scanning is solved.

[0045] Hardware synchronization includes: depth data: integrated ToF (time of flight) camera or structured light module to collect target depth image at a frame rate of 30 fps or more, resolution ≥ 640*480, depth accuracy ±2mm (within 1m). Color image is captured by an RGB camera (≥10 million pixels) with the same optical axis as the depth camera or through accurate calibration of the relative position of the calibration board to ensure pixel-level alignment. Time synchronization includes: all sensors (laser scanning plug-in, depth camera, RGB camera, positioning module) realize synchronous acquisition through hardware trigger (such as GPIO synchronization signal) or software timestamp (accuracy ≤1μs), avoiding asynchronous error.

[0046] The coordinate system is established by establishing the external parameter matrix of the device main coordinate system (S), the laser scanning coordinate system (L), the depth camera coordinate system (D), and the RGB camera coordinate system (C), and the conversion relationship (rotation matrix R, translation vector T) between the coordinate systems is obtained through high-precision calibration (such as Zhang Zhengyou calibration method + laser plane calibration). The sub-millimeter level fine measurement data (laser point cloud P_L), depth data (D_img), and color image (C_img) are all converted to the global coordinate system (W), and the conversion formula is: PW=RWS*(RSL*PL+TSL)+TWS (the depth and color data are processed in the same way).

[0047] Coarse-precision positioning fusion uses centimeter-level coarse positioning information (device pose W_S) as the initial global coordinate to globally register sub-millimeter-level laser point cloud, and eliminates cumulative error through iterative closest point (ICP) algorithm or its improved version (such as Point-to-plane ICP), especially correcting the drift problem in long-distance scanning.

[0048] Detail enhancement fills the holes in the depth data (such as KNN interpolation) and filters the noise (bilateral filtering), supplements the missing points caused by occlusion or low reflectivity of laser scanning, and improves the point cloud integrity of complex surfaces (such as automobile cover part molds). The depth point cloud is fused with the laser point cloud, the high-precision points measured by the laser (error ≤0.1mm) are retained, the low-density areas are filled with depth point cloud, and a hybrid density point cloud is formed, taking into account efficiency and details.

[0049] Texture mapping maps the RGB color value to each vertex of the point cloud through coordinate conversion, uses bilinear interpolation to process non-aligned pixels, and ensures the clarity of the texture. For high light and shadow areas, light compensation is performed in combination with depth data to avoid color distortion and improve the visual reality of subsequent modeling (such as texture restoration in cultural relic digitization).

[0050] The error elimination and model construction adopt statistical filtering (such as RANSAC) to eliminate outliers, and the fused point cloud is smoothed by moving least squares (MLS), and finally a triangular mesh model or high-precision point cloud package suitable for high-precision modeling is generated, which meets the precision requirements (sub-millimeter level error) of industrial detection (such as tolerance analysis) or reverse engineering.

[0051] In some embodiments, the access to the spot scanning plug-in through the standardized interface of the device body supports the transverse arrangement mode and the longitudinal arrangement mode, comprising: setting a mechanical clamping structure and an electrical communication interface on the standardized interface of the device body, and detachably mounting the spot scanning plug-in on the front end or the side end of the device body through the mechanical clamping structure; when the spot scanning plug-in is arranged transversely, the corresponding laser emitter array is arranged along the horizontal direction of the device body, and when it is arranged longitudinally, it is arranged along the vertical direction; the electrical communication interface automatically matches the data transmission protocol according to the arrangement direction to realize the hardware configuration of the transverse or longitudinal scanning mode.

[0052] The embodiment clearly defines the hardware connection mode of the spot scanning plug-in and the device body, realizes the detachable installation of the plug-in through the standardized interface, and adjusts the direction of the laser emitter array and the data protocol according to the transverse / longitudinal arrangement to complete the hardware configuration initialization.

[0053] The mechanical interface design sets an L-shaped mechanical clamping structure on the front end / side end of the device body, which contains a elastic clamping tenon and a positioning groove, and the plug-in is embedded through horizontal sliding (transverse arrangement) or vertical plugging (longitudinal arrangement). The clamping has a built-in Hall sensor or micro switch to detect the installation state of the plug-in and feedback to the master chip.

[0054] The electrical communication interface integrates a 24-pin gold finger through the interface, which contains power pins (5V / 12V adaptive), differential signal lines and GPIO control signals. When the plug-in is installed, the interface contact piece automatically docks, the device body reads the configuration file in the EEPROM of the plug-in, identifies the current arrangement direction (transverse / longitudinal), and automatically loads the corresponding drive protocol (such as high-speed synchronization protocol for transverse scanning and depth-first transmission protocol for longitudinal scanning).

[0055] The laser emitter array layout includes: when arranged transversely, the laser emitters are arranged in a linear array along the horizontal direction (X-axis) of the device body, and the emission direction is parallel to the holding direction of the device, which is suitable for wide-surface scanning; when arranged longitudinally, the emitters are arranged along the vertical direction (Y-axis), and the emission direction is perpendicular to the holding direction, which is suitable for deep hole or vertical surface scanning. The two modes are ensured to be accurate in emission angle by hardware limiting structure.

[0056] In some embodiments, the obtaining of the centimeter-level coarse positioning information by the device body comprises: collecting acceleration data and angular velocity data in real time by an inertial navigation system (INS) built in the device body, combining with an image of an environmental feature point obtained by a visual positioning module, and using a multi-sensor fusion algorithm to perform time synchronization and error compensation on the inertial navigation data and the visual positioning data to generate real-time spatial coordinates and orientation parameters of the device body.

[0057] By fusing the inertial navigation system (INS) and the visual positioning module, centimeter-level pose estimation of the device body is achieved, and the problem of insufficient accuracy of a single sensor in a dynamic scene is solved.

[0058] The sensor configuration comprises a nine-axis INS (three-axis accelerometer, three-axis gyroscope, and three-axis magnetometer) built in the device body, and the sampling frequency is 1000 Hz; the visual positioning module comprises a binocular camera (baseline distance of 8 cm and resolution of 1920x1080) carrying an IMX490 image sensor and supporting feature point detection (ORB algorithm) and optical flow tracking.

[0059] The multi-sensor fusion process comprises: time synchronization: aligning the INS and the camera clock through a hardware PPS (pulse per second) signal; performing linear interpolation on the visual positioning result to match the high-frequency output (1 ms interval) of the INS.

[0060] Error compensation uses environmental feature points (such as SIFT features) obtained by visual positioning to construct a sparse map, fuses the motion prediction of the INS and the absolute position observation of the visual positioning through an EKF filter, and corrects the drift error of the INS (such as eliminating the accumulation of gyroscope zero offset). The magnetometer data is used to calibrate the orientation of the device, to solve the problem of heading angle drift of the INS in a visual-free scene, and finally to output the global coordinates (X, Y, Z, accuracy ±5 cm) and Euler angles (pitch, roll, yaw, accuracy ±1°) of the device body.

[0061] In some embodiments, the controlling of the light spot scanning plug-in to perform light spot scanning according to the corresponding scanning strategy based on the centimeter-level coarse positioning information for sub-millimeter-level fine measurement of the target object according to the arrangement mode of the light spot scanning plug-in comprises: if the light spot scanning plug-in is arranged horizontally, controlling the laser emission array to move in an equal-interval stepping manner along the horizontal direction, and the single stepping distance is not more than a preset distance, while the device body keeps uniform translation at the current orientation; if the light spot scanning plug-in is arranged vertically, controlling the laser emission array to move in a variable-interval adaptive scanning manner along the vertical direction, and dynamically adjusting the stepping distance according to the curvature change of the target object, and the greater the curvature, the smaller the stepping distance, so as to ensure that the projection interval of the scanning light spot on the target surface is not more than the sub-millimeter-level accuracy requirement.

[0062] According to the plug-in arrangement direction, the scanning strategy is dynamically adjusted, and sub-millimeter level precision is ensured through equidistance / variable distance scanning, and efficiency and complex surface detail capture are considered.

[0063] The transverse arrangement scanning strategy includes: the laser emission array is driven by a MEMS mirror or a linear motor, and is scanned row by row in the horizontal direction at a fixed step distance of 0.5 mm (the step error is less than or equal to ±0.05 mm), and when the device main body is translated at a constant speed of 200 mm / s, the scanning line spacing is dynamically adjusted to 1-5 mm (according to the target curvature obtained by coarse positioning, the curvature is less than 0.1 mm -1 The time interval is 5 mm, and the curvature is greater than or equal to 0.5 mm -1 The time interval is 1 mm). A single scan covers a width of 30 mm, and is suitable for fast scanning of a plane or a shallow curved surface.

[0064] The longitudinal arrangement scanning strategy uses a piezoelectric ceramic micro drive device to control the vertical movement of the laser array, and the step distance is adjusted in real time according to the target surface curvature: the local curvature is calculated through coarse positioning point cloud (such as using principal component analysis PCA), and the curvature is greater than 0.3 mm -1 The step is 0.3 mm when the curvature is less than or equal to 0.1 mm -1 The step is 1 mm, which ensures that the projection interval of the light spot on the target surface is less than or equal to 0.8 mm (considering the incident angle compensation). For deep groove structures, a “zigzag” layered scanning is enabled, each layer is vertically stepped by 5 mm, and the layer overlap rate is greater than or equal to 30% to avoid missing measurement.

[0065] In some embodiments, the depth data and color image of the target object are synchronously collected during the light spot scanning process, including: collecting the depth data of the target object through the depth camera integrated with the light spot scanning plug-in at a frequency range not less than a preset frequency range, and synchronously collecting the color image through the RGB camera at a frame rate range not less than a preset frame rate range; the depth data and the color image are time-stamped and aligned, and based on the spatial coordinates and the orientation parameters of the device main body, the coordinate systems of the depth data and the color image are unified to the global coordinate system of the device main body, to form a time-stamped synchronous data set.

[0066] The depth and color data are synchronously collected at high speed to establish a time-stamped synchronous data set, which provides a time-space alignment basis for subsequent multi-source fusion.

[0067] Hardware integration and collection parameters: the light spot scanning plug-in integrates a ToF depth camera (such as Sony DepthSenseIMX556, frame rate 120 Hz, resolution 640×480, accuracy ±2 mm@1 m) and an RGB camera (12 MP, frame rate 60 fps, supporting automatic white balance). The optical axes of the two cameras are calibrated through a precision tooling, and the included angle is less than or equal to 0.2°, which ensures that the field of view overlap rate is greater than or equal to 95%.

[0068] Synchronization process: timestamp alignment: depth camera and RGB camera are synchronized by hardware trigger, each frame is attached with nanosecond timestamp (based on high-precision clock module of device main body), for asynchronous frames (such as depth camera 120Hz / RGB camera 60fps), the timestamps are matched by nearest neighbor interpolation method, error ≤5ms.

[0069] Coordinate system: establish device main body coordinate system S, obtain the external parameter matrix of depth camera D and RGB camera C relative to S through Zhang Zhengyou calibration method (RDS, TDS; RCS, TCS), convert depth data (ZD(u, v)) and RGB pixel (C(u, v)) into three-dimensional points (XS, YS, ZS) in S coordinate system, formula is (KD is the internal parameter matrix of depth camera):

[0070] .

[0071] In some embodiments, the cm-level coarse positioning information is used as a global spatial framework, and the sub-millimeter-level fine measurement data, depth data and color image are fused to construct point cloud data of the target object, including: the cm-level coarse positioning information is used as a global spatial framework, and the sub-millimeter-level fine measurement data is converted to the global coordinate system based on the spatial coordinates and orientation parameters of the device main body to form a high-precision measurement point cloud; the depth data is matched to the coordinate system of the high-precision measurement point cloud through coordinate system conversion and scale calibration; the pixel color value of the color image is mapped to the corresponding point cloud coordinate through projection transformation to generate initial point cloud data with color information, and the spatial error of the initial point cloud data is not more than 0.3 millimeters.

[0072] Through coordinate system conversion and data registration, the high-precision laser point cloud, depth point cloud and color information are fused to generate an initial point cloud with color, and the spatial error is controlled to be in the sub-millimeter level.

[0073] Global conversion of laser point cloud: the sub-millimeter-level fine measurement data (laser point cloud PL in scanning coordinate system L) is converted to global coordinate PW through the device pose (RWS, TWS, conversion from global coordinate system W to device coordinate system S) obtained by coarse positioning: PW=RWS*(RSL*PL+TSL)+TWS (RSL, TSL are fixed external parameters from scanning coordinate system L to device coordinate system S, determined by factory calibration, accuracy ±0.05mm).

[0074] Depth point cloud matching and calibration: through scale calibration of depth point cloud (correcting temperature drift error of ToF camera, calibration coefficient updated every 5 minutes), through ICP algorithm and laser point cloud for coarse registration (initial iteration number is 10 times, convergence threshold is 0.1mm), and then using point-to-plane ICP for fine registration, the coordinate system deviation of the two types of point clouds is ensured to be ≤0.2mm.

[0075] Color texture mapping finds the corresponding pixel (u, v) in the RGB image for each laser point P_W by projection transformation (R_CS, T_CS), adopts bilinear interpolation to handle sub-pixel level offset, and assigns color values (R, G, B) to the point cloud; for depth points not covered by laser points, the color mean of the neighboring laser points is searched by KNN (k = 5) to fill in, ensuring that the texture mapping error is ≤1 pixel (corresponding to actual distance ≤0.3mm@ working distance 50cm).

[0076] In some embodiments, the details accuracy and texture effect of the point cloud data are optimized based on the depth data and color image, combined with the global spatial consistency of the centimeter-level coarse positioning information, to obtain a target object point cloud model suitable for high-precision target modeling and complex surface detection scenarios, including: edge feature detection is performed on the initial point cloud data to identify the contour edges and curvature mutation regions of the target object, and sub-millimeter level measurement points missing in the region are supplemented using depth data; the color values of adjacent point clouds are smoothed and interpolated by a texture mapping algorithm of a color image to eliminate texture faults caused by changes in scanning angle; noise points are removed by an outlier filtering algorithm to ensure that the optimized point cloud data completely covers the details of the target object surface and the texture transition is natural and continuous.

[0077] Through edge detection, data supplementation, texture interpolation and noise filtering, the details and texture continuity of the point cloud are optimized, solving the feature loss and texture fault problems in traditional scanning.

[0078] Edge feature detection and data supplementation use Canny edge detection algorithm to extract the contour of the depth image, and identify the curvature mutation regions (such as edges and sharp corners) by combining point cloud normal vector analysis (calculating the eigenvalues of the local covariance matrix). For the part of the laser point cloud density <200 points / cm² in these regions, the depth point cloud is used for interpolation and supplementation: a virtual point with an interval of 0.5mm is generated in the missing area by fitting a surface through the moving least squares method (MLS), ensuring the integrity of the edge features (such as the point cloud spacing of sharp edges ≤0.3mm).

[0079] Texture smoothing interpolation uses an interpolation algorithm based on normal vector weight for the color jump of adjacent point clouds caused by changes in scanning angle (such as regions with a normal angle >30°): the normal vector angle of the target point and k nearest points (k = 10) is calculated, the weight coefficient is set as cosθ (θ is the angle), and the weighted average color value is calculated to eliminate texture faults (such as the texture transition gradient of a cylinder with large curvature ≤5%).

[0080] The noise filtering process includes: firstly removing outliers by statistical filtering (calculating the mean and standard deviation of the distance of the point cloud neighborhood, removing points with a distance exceeding μ+3σ), and then using bilateral filtering (spatial distance weight + normal vector angle weight) to smooth the point cloud surface, while preserving the edge features and keeping the surface roughness ≤0.1mm.

[0081] In some embodiments, the method further comprises: using a pre-trained deep learning model to analyze the target surface features of the real-time collected depth data and color images, and identifying the complex curved surface area and the occluded blind area of the target object; when a complex curved surface with a curvature greater than a preset threshold or an occluded blind area is detected, an encryption scanning strategy of a light spot scanning plug-in is automatically triggered, and the laser emission array is controlled to perform supplementary measurement at double scanning density in the area. The deep learning model is trained based on a large amount of historical scanning data and can identify a plurality of preset typical complex curved surface features.

[0082] By using a deep learning model to identify complex curved surfaces and occluded blind areas in real time, triggering an encryption scanning strategy, and dynamically improving the point cloud density of key areas, the adaptive deficiency problem of traditional scanning is solved.

[0083] The deep learning model is based on the UNet++ architecture, the input is a depth image (640×480) and an RGB image (channel splicing into 6 channels), and the output is a pixel-level classification result (complex curved surface area / ordinary area / occluded blind area). The training data includes 100,000 groups of scanning data of industrial parts, cultural relics sculptures and other scenes, the labeling adopts artificial marking combined with automatic labeling by curvature calculation, and the IOU of the model on the test set is ≥0.92.

[0084] Real-time detection and strategy triggering input the model at 20fps through depth and RGB images, and the inference time is ≤40ms. When a complex curved surface area (curvature >0.5mm - ¹) or an occluded blind area (depth value mutation area) is detected, the main control chip sends instructions to the scanning plug-in: in the horizontal scanning mode, the scanning line spacing of the area is reduced from 5mm to 2mm, and the step distance is reduced to 0.3mm; in the vertical scanning mode, double-spot emission is enabled (the original single-spot is changed to double-spot parallel projection), and the density is increased by 1 times. The encryption scanning area is marked in real time through coarse positioning coordinates, ensuring that the spatial alignment error of the supplementary measurement data and the historical point cloud is ≤0.2mm.

[0085] In some embodiments, the method further comprises: using a trajectory matching algorithm based on dynamic time warping to analyze the movement trajectory of the device body in real time, identifying abnormal motion trajectory segments caused by handheld shaking during scanning; for the point cloud data corresponding to the abnormal trajectory segments, using a spatio-temporal filtering algorithm combined with the sub-millimeter level fine measurement data of adjacent frames to correct the trajectory, eliminating the positioning errors introduced by handheld instability, and ensuring that the spatial error of the corrected point cloud data does not exceed the sub-millimeter level precision requirement.

[0086] Abnormal trajectories caused by handheld shaking are identified by a dynamic time warping (DTW) algorithm, and the point cloud positioning error is corrected by spatio-temporal filtering to ensure sub-millimeter level precision.

[0087] Trajectory matching and anomaly detection: by collecting inertial data (acceleration, angular velocity) of the device body to construct a motion trajectory curve, and by matching the trajectory curve with a preset ideal uniform translation trajectory (speed 200 mm / s, acceleration fluctuation ≤50 mm / s²) using a DTW algorithm, a trajectory similarity score is calculated. When the score is <0.7, it is determined to be an abnormal trajectory segment (such as acceleration / deceleration or rotation mutation caused by hand shaking). Spatio-temporal filtering and error correction: for the point cloud data corresponding to the abnormal trajectory segments, a spatio-temporal joint filtering is used: time dimension: using the stable trajectory data of the previous 5 frames and the next 5 frames, the device pose of the abnormal frame is predicted by Kalman filtering to correct the high-frequency noise in the coarse positioning (such as attitude error caused by angular velocity mutation). Spatial dimension: for the point cloud in the abnormal area, the sub-millimeter level laser point cloud of the adjacent frames is used for rigid transformation registration (using quaternion method to solve rotation and translation parameters), to eliminate the coordinate offset caused by shaking (correction accuracy ≤0.1 mm). Finally, the spatial error of all point cloud data is ensured to be ≤0.3 mm in the global coordinate system, meeting the high-precision detection requirement.

[0088] Currently, there is no solution in the prior art that can access a flexible switching arrangement mode spot scanning plug-in through a standardized interface, combine centimeter level coarse positioning and sub-millimeter level fine measurement to achieve high efficiency and high precision scanning, or propose a method of synchronously fusing depth data, color image and multi-precision positioning information to optimize point cloud details and texture effect. Therefore, how to realize sub-millimeter level fine measurement while ensuring scanning efficiency, and improve the quality of point cloud construction through multi-source data fusion, has become a technical problem to be solved in this field.

[0089] In some embodiments, the full-automatic identification and arrangement mode determination of the spot scanning plug-in are realized based on computer vision and machine learning, the multi-sensor fusion positioning algorithm is optimized based on self-supervised learning, and the limitations of traditional mechanical limit identification and the positioning drift problem in complex environments are solved.

[0090] The plug-in type intelligent identification includes a micro RGB-D camera (resolution 480x480) built in the device main body, which collects images in real time in the plug-in installation area, and identifies the plug-in type (horizontal / vertical / special function plug-in) and installation posture through a lightweight convolutional neural network (such as MobileNetV3). The network model is pre-trained through self-supervised learning (features are extracted through contrast learning using unlabeled plug-in images), and only a small amount of labeled data is needed for fine-tuning during deployment, with an identification accuracy of ≥99.5%. After identification, the control protocol and scanning parameter library of the corresponding plug-in are automatically loaded.

[0091] Self-supervised multi-sensor fusion positioning: Construct a self-supervised learning framework to optimize the EKF filter: Unsupervised error prediction: Use the LSTM network to learn the historical sequence features of the inertial navigation data (acceleration, angular velocity), and predict the short-term motion trend. When the visual positioning data is missing (such as weak texture scenes), the predicted value is used to compensate the positioning output, reducing the drift error (compared with traditional EKF, the long-time positioning accuracy is improved by 30%). Dynamic weight distribution: Through the attention mechanism, the fusion weights of INS, vision, and UWB data are automatically adjusted, for example, increasing the INS weight when moving quickly (response frequency 1kHz), and relying on UWB / vision data correction when stationary (positioning accuracy ≤3cm). The weight parameters are obtained through offline reinforcement learning training.

[0092] In some embodiments, by introducing deep reinforcement learning (DRL) to dynamically decide the scanning strategy, the optimal scanning path and parameters are generated in real time based on the target surface features, solving the efficiency-accuracy imbalance problem of traditional rule-based scanning in complex curved surface scenes.

[0093] The reinforcement learning environment construction includes: the state space S includes: the current device pose (X, Y, Z, Euler angle), the target surface curvature distribution (calculated through the previous coarse scanning point cloud), the scanning plug-in type (horizontal / vertical), and the remaining battery power; the action space A includes: the scanning step distance (0.3-5mm), the laser emission frequency (5-20kHz), and the camera exposure parameters (ISO, shutter speed); the reward function R is designed as: (effective point cloud density x 0.6 + scanning speed x 0.3 - energy consumption x 0.1), encouraging to increase the density in high curvature areas and to increase the speed in flat areas.

[0094] Real-time strategy generation: The PPO (Proximal Policy Optimization) algorithm is used to train the DRL model, the input layer is a state vector (12 dimensions), and the output layer is an action parameter (continuous value). The model inference delay is ≤10ms during deployment, and the scanning parameters are dynamically adjusted according to the curvature heat map of the current scanning area (generated in real time through Gaussian process regression). When the curvature is detected to be >0.4mm -When the sharp edge of 1 is detected, the system automatically switches to the longitudinal insert (if the current one is transverse) and reduces the step distance to 0.3 mm, and the emission frequency is increased to 15 kHz; for the planar area, the "skip scanning" mode is enabled, the step distance is 5 mm, and the area without features (determined by the edge detection algorithm) is skipped, and the scanning efficiency is increased by 40% while maintaining the key feature point density ≥ 200 points / cm².

[0095] In some embodiments, by modeling the spatial relationship of the point cloud with a graph neural network (GNN) and combining a generative adversarial network (GAN) to complete the details of the occluded area, the distortion problem of traditional interpolation methods in complex structures (such as deep holes and overlapping surfaces) is solved.

[0096] The graph neural network fusion framework includes: point cloud graph construction: converting laser point cloud and depth point cloud into an undirected graph, with three-dimensional points as nodes (features including coordinates, normal vectors, and reflectivity intensity), and k-neighbor connections (k=20) as edges, and edge features as the distance and angle difference between nodes. Feature propagation and fusion: aggregate neighborhood features through GCN (graph convolution network) layers to update the global descriptor of the node, realize multi-source data feature alignment (such as combining the high-precision coordinates of laser points with the dense distribution characteristics of depth points), and output the fused point cloud feature vector (dimension 128).

[0097] The intelligent completion of the occluded area is achieved by constructing a conditional GAN model, with the input being the point cloud with missing areas (labeled mask) and the coarse positioning pose, and the output being the completed point cloud coordinates: the generator uses the PointNet++ architecture to predict the coordinates of missing points based on the surrounding effective point cloud and the device pose, with the constraint being the curvature continuity of the completed area (achieved through a differential geometry loss function); the discriminator distinguishes between real point clouds and generated point clouds, and the training data comes from real scans of occluded scenes (such as inside bolt holes and back of blades), with a completion accuracy of ≤0.5 mm within a 10 cm depth range, a 60% reduction in error compared to traditional interpolation methods.

[0098] In some embodiments, by introducing a Meta-Learning algorithm, the device quickly adapts to new scanning scenes (such as cultural relics sculptures and aerospace special-shaped parts), and through the "learning to learn" mechanism, the parameters of the whole process from insert configuration to point cloud generation are optimized.

[0099] The Meta-Learning task definition defines different scanning scenes as sub-tasks (such as plane detection, deep hole scanning, and free-form surface modeling), each task including: insert arrangement method, scanning strategy parameters, and data fusion weights. The goal of Meta-Learning is to quickly adjust the model parameters through a small number of samples (1-2 scans) to meet the point cloud accuracy (sub-millimeter level error) in new scenes.

[0100] Adopting MAML (Model-Agnostic Meta-Learning) framework, the base model includes: plug-in configuration network: predicting the best plug-in type (horizontal / vertical / multi-spot combination) according to the scene image (uploaded by mobile phone or captured by device-mounted camera), parameter update step 0.01; scanning strategy network: generating parameters such as stepping distance and emission frequency based on scene curvature distribution heat map, supporting continuous value output; fusion weight network: dynamically adjusting the fusion weights of laser point cloud, depth data and color image, taking scene reconstruction error as feedback signal. In actual application, the user uploads the local scanning data (about 1000 points) of the new scene, and the device completes the meta-learning update within 5 seconds to generate a customized scanning scheme, which improves the efficiency by 80% compared with traditional manual configuration, and the first scanning accuracy directly reaches the optimal level of the target scene.

[0101] In some embodiments, by combining self-supervised learning with super-resolution reconstruction algorithm, the point cloud details are enhanced in real time during scanning, solving the problem of point cloud noise and insufficient resolution caused by motion blur or low reflectivity surface in handheld scanning.

[0102] Self-supervised noise robustness training uses unlabelled data collected by the scanning device to train the denoising model through contrastive learning: adding Gaussian noise (σ=0.5mm) and salt and pepper noise (density 5%) to the input point cloud as negative samples, and the original point cloud as positive samples, the model (PointDenoiser) learns to restore the true point coordinates, and the root mean square error (RMSE) of the denoised point cloud is ≤0.08mm.

[0103] Super-resolution point cloud reconstruction uses a Transformer-based super-resolution network (PointSR) for sparse scanning areas (such as low-density point clouds of flat surfaces): the input is a low-resolution point cloud (spacing 5mm), and the output is a high-density point cloud (spacing 0.5mm), the network captures long-distance geometric dependencies through self-attention mechanism, and maintains global structural consistency combined with coarse positioning pose information. After ICP precise registration, the coincidence degree of the reconstructed point cloud with the original high-precision point cloud is ≥99.2%, effectively improving the surface smoothness of subsequent modeling (such as NURBS surface fitting error ≤0.1mm).

[0104] The embodiment of the application provides a plug-in handheld laser scanner based on spot fine positioning and a method thereof. The method accesses a spot scanning plug-in supporting horizontal / vertical arrangement through a standardized interface, can flexibly adjust a scanning direction according to a target object form, adapts to diversified scanning scenes, and improves equipment versatility. The scanning area is quickly determined by using the centimeter-level coarse positioning information provided by the device main body, and the sub-millimeter-level fine measurement of the spot scanning is combined, so that the local detail precision is significantly improved while the overall scanning efficiency is ensured, and the contradiction between efficiency and precision in the traditional scanning is solved. The high-precision point cloud data with color information is constructed by synchronously collecting depth data and color images and fusing with the positioning information. The surface detail integrity and texture reality of the point cloud model are effectively improved by the detail precision optimization and texture mapping algorithm, and target feature loss is avoided. The differential scanning strategy and data processing mechanism are suitable for high-precision target modeling, industrial part surface detection, cultural relic digitization protection and other scenes with high requirements for detail precision and texture effect.

[0105] The embodiment of the application further provides a plug-in handheld laser scanning device based on spot fine positioning. The plug-in handheld laser scanning device based on spot fine positioning is used to execute the steps of the plug-in handheld laser scanning method based on spot fine positioning shown in the above-mentioned embodiments. The plug-in handheld laser scanning device based on spot fine positioning can be a single server or a server cluster, or the plug-in handheld laser scanning device based on spot fine positioning can be a terminal, which can be a handheld terminal, a notebook computer, a wearable device or a robot.

[0106] The plug-in handheld laser scanning device based on spot fine positioning comprises:

[0107] A plug-in access unit is configured to access a spot scanning plug-in through a standardized interface of a device main body, and the spot scanning plug-in supports a horizontal arrangement mode and a vertical arrangement mode. The device main body is configured to obtain centimeter-level coarse positioning information, and the centimeter-level coarse positioning information comprises spatial coordinates and an orientation of the device main body.

[0108] A scanning control unit is configured to adjust a scanning path or range according to an arrangement mode of the spot scanning plug-in in combination with the centimeter-level coarse positioning information, control the spot scanning plug-in to perform spot scanning according to a corresponding scanning strategy, and perform sub-millimeter-level fine measurement on a target object based on the centimeter-level coarse positioning information.

[0109] The optimization solution unit is configured to: in a spot scanning process, synchronously collect depth data and color images of a target object; take the centimeter-level coarse positioning information as a global spatial framework, fuse sub-millimeter-level fine measurement data, depth data and color images, and construct point cloud data of the target object; based on the depth data and color images, and in combination with global spatial consistency of the centimeter-level coarse positioning information, optimize details precision and texture effects of the point cloud data, and obtain a target object point cloud model suitable for high-precision target modeling and complex surface detection scenarios.

[0110] It should be noted that, for the convenience and brevity of description, the specific working processes of the above-described plug-in handheld laser scanning device based on spot fine positioning and modules can refer to the corresponding processes in the above-described plug-in handheld laser scanning method embodiments based on spot fine positioning, which will not be described herein.

[0111] The above-described plug-in handheld laser scanning method based on spot fine positioning can be implemented in the form of a computer program, which can run on the device provided in the present application.

[0112] As shown in Figure 6 The present application also provides a plug-in handheld laser scanner based on spot fine positioning. The plug-in handheld laser scanner based on spot fine positioning comprises a processor, a storage and a network interface connected through a device bus, wherein the storage can include a storage medium and an internal memory.

[0113] The storage medium can store an operating device and a computer program. The computer program comprises program instructions, which, when executed, can cause the processor to execute any one of the plug-in handheld laser scanning methods based on spot fine positioning.

[0114] The processor is configured to provide computing and control capabilities to support the operation of the entire plug-in handheld laser scanner based on spot fine positioning.

[0115] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium, which, when executed by the processor, can cause the processor to execute any one of the plug-in handheld laser scanning methods based on spot fine positioning.

[0116] The network interface is configured to perform network communication, such as sending the assigned task, etc. Those skilled in the art can understand that the structure shown in the above embodiments is only part of the structure related to the scheme of the present application, and does not constitute a limitation on the terminal to which the scheme of the present application is applied. Specifically, the plug-in handheld laser scanner based on fine positioning of a light spot can include more or fewer components than those mentioned in the embodiments, or combine certain components, or have a different component arrangement.

[0117] It should be understood that the processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0118] In one embodiment, the processor is configured to run a computer program stored in the memory to perform the following steps:

[0119] Access the light spot scanning plug-in through a standardized interface of the device main body, the light spot scanning plug-in supports a transverse arrangement mode and a longitudinal arrangement mode; obtain centimeter-level coarse positioning information by using the device main body, the centimeter-level coarse positioning information including spatial coordinates and an orientation of the device main body;

[0120] Adjust a scanning path or range according to the arrangement mode of the light spot scanning plug-in and the centimeter-level coarse positioning information, control the light spot scanning plug-in to perform light spot scanning according to a corresponding scanning strategy, and perform sub-millimeter-level fine measurement on a target object based on the centimeter-level coarse positioning information;

[0121] In the light spot scanning process, depth data and color images of the target object are synchronously collected; the centimeter-level coarse positioning information is taken as a global spatial framework, sub-millimeter-level fine measurement data, depth data and color images are fused, and point cloud data of the target object is constructed; based on the depth data and the color images and in combination with global spatial consistency of the centimeter-level coarse positioning information, the details accuracy and texture effect of the point cloud data are optimized, and a target object point cloud model suitable for high-precision target modeling and complex surface detection scenarios is obtained.

[0122] The embodiment of the present application also provides a computer readable storage medium, which stores a computer program, and the computer program comprises program instructions, and the processor executes the program instructions to realize the steps of the plug-in handheld laser scanning method based on spot fine positioning provided by each embodiment of the present application.

[0123] The computer readable storage medium can be an internal storage unit of the plug-in handheld laser scanner based on spot fine positioning, for example, a hard disk or a memory of the plug-in handheld laser scanner based on spot fine positioning. The computer readable storage medium can also be an external storage device of the plug-in handheld laser scanner based on spot fine positioning, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card and the like.

[0124] The above merely provides a specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present application, and these modifications or replacements should be encompassed in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A plug-in handheld laser scanning method based on fine positioning of light spots, characterized in that, The application is applied to a plug-in handheld laser scanner based on spot fine positioning, and the method comprises the following steps: Accessing a spot scanning plug-in through a standardized interface of a device main body, wherein the spot scanning plug-in supports a transverse arrangement mode and a longitudinal arrangement mode; acquiring centimeter-level coarse positioning information by using the device main body, wherein the centimeter-level coarse positioning information comprises spatial coordinates and an orientation of the device main body; Adjusting a scanning path or range according to the arrangement mode of the spot scanning plug-in in combination with the centimeter-level coarse positioning information, and controlling the spot scanning plug-in to perform spot scanning according to a corresponding scanning strategy, so as to perform sub-millimeter-level fine measurement on a target object based on the centimeter-level coarse positioning information; During the spot scanning process, synchronously collecting depth data and color images of the target object; taking the centimeter-level coarse positioning information as a global spatial framework, fusing the sub-millimeter-level fine measurement data, the depth data and the color images, and constructing point cloud data of the target object; and based on the depth data and the color images and in combination with the global spatial consistency of the centimeter-level coarse positioning information, optimizing the detail accuracy and texture effect of the point cloud data, so as to obtain a point cloud model of the target object suitable for high-precision target modeling and complex surface detection scenarios.

2. The method of claim 1, wherein, The step of accessing the spot scanning plug-in through the standardized interface of the device main body, wherein the spot scanning plug-in supports the transverse arrangement mode and the longitudinal arrangement mode, comprises the following steps: Providing a mechanical clamping structure and an electrical communication interface on the standardized interface of the device main body, and detachably mounting the spot scanning plug-in on the front end or the side end of the device main body through the mechanical clamping structure; When the spot scanning plug-in is arranged transversely, corresponding laser emission arrays are arranged along the horizontal direction of the device main body, and when the spot scanning plug-in is arranged longitudinally, the laser emission arrays are arranged along the vertical direction; the electrical communication interface automatically matches a data transmission protocol according to the arrangement direction, so as to realize hardware configuration in the transverse or longitudinal scanning mode.

3. The method of claim 1, wherein, The step of acquiring the centimeter-level coarse positioning information by using the device main body comprises the following steps: Real-time collecting acceleration data and angular velocity data by using an inertial navigation system built in the device main body, combining with environment feature point images obtained by a visual positioning module, and using a multi-sensor fusion algorithm to perform time synchronization and error compensation on the inertial navigation data and the visual positioning data, so as to generate real-time spatial coordinates and orientation parameters of the device main body.

4. The method of claim 1, wherein, The step of adjusting the scanning path or range according to the arrangement mode of the spot scanning plug-in in combination with the centimeter-level coarse positioning information, and controlling the spot scanning plug-in to perform spot scanning according to a corresponding scanning strategy, so as to perform sub-millimeter-level fine measurement on a target object based on the centimeter-level coarse positioning information, comprises the following steps: If the spot scanning plug-in is arranged transversely, the laser emission arrays are controlled to move along the horizontal direction in an equal-interval stepping manner, and the single stepping distance is not more than a preset distance, and meanwhile the device main body is kept to move at a constant speed in the current orientation; If the spot scanning plug-in is arranged longitudinally, the laser emission arrays are controlled to move along the vertical direction in a variable-interval adaptive scanning manner, and the stepping distance is dynamically adjusted according to the curvature variation of the target object, and the greater the curvature is, the smaller the stepping distance is, so as to ensure that the projection interval of the scanning spot on the target surface is not more than the sub-millimeter-level accuracy requirement.

5. The method of claim 1, wherein, The step of synchronously collecting depth data and color images of the target object during the spot scanning process comprises the following steps: The depth camera integrated with the spot scanning plug-in collects the depth data of the target object at a frequency not lower than a preset frequency range, and synchronously collects the color image through the RGB camera at a frame rate not lower than a preset frame rate range; The depth data and the color image are time-stamped and aligned, and the coordinate systems of the depth data and the color image are unified to the global coordinate system of the device body based on the spatial coordinates and the orientation parameters of the device body, to form a time-labeled synchronous data set.

6. The method of claim 1, wherein, The cm-level coarse positioning information is used as a global spatial framework to fuse the sub-millimeter-level fine measurement data, the depth data and the color image, and to construct the point cloud data of the target object, including: The cm-level coarse positioning information is used as a global spatial framework, and the sub-millimeter-level fine measurement data is converted to the global coordinate system based on the spatial coordinates and the orientation parameters of the device body, to form a high-precision measurement point cloud. The depth data is matched to the coordinate system of the high-precision measurement point cloud through coordinate system conversion and scale calibration. The pixel color values of the color image are mapped to the corresponding point cloud coordinates through projection transformation, to generate initial point cloud data with color information, and the spatial error of the initial point cloud data is not more than 0.3 millimeters.

7. The method of claim 1, wherein, The details and texture of the point cloud data are optimized based on the global spatial consistency of the depth data and the color image combined with the cm-level coarse positioning information, to obtain a target object point cloud model suitable for high-precision target modeling and complex surface detection scenarios, including: Edge feature detection is performed on the initial point cloud data to identify the contour edges and curvature mutation regions of the target object, and depth data is used to supplement the missing sub-millimeter-level measurement points in the regions; Through a texture mapping algorithm of the color image, the color values of adjacent point clouds are smoothly interpolated to eliminate texture faults caused by changes in scanning angles; Through an outlier filtering algorithm, noise points are removed to ensure that the optimized point cloud data completely covers the details of the target object surface and the texture transition is natural and continuous.

8. The method of claim 1, wherein, The method further includes: A pre-trained deep learning model is used to analyze the target surface features of the real-time collected depth data and color image, to identify complex curved surface regions and occluded blind areas of the target object; When a complex curved surface with a curvature greater than a preset threshold or an occluded blind area is detected, an encryption scanning strategy of the spot scanning plug-in is automatically triggered to control the laser emission array to perform supplementary measurement at double the scanning density in the region, and the deep learning model is trained based on a large amount of historical scanning data and can identify a plurality of preset typical complex curved surface features.

9. The method of claim 1, wherein, The method further includes: A trajectory matching algorithm based on dynamic time warping is used to analyze the moving trajectory of the device body in real time, to identify abnormal trajectory segments caused by handheld shaking during scanning; For the point cloud data corresponding to the abnormal trajectory segments, a spatio-temporal filtering algorithm is used to correct the trajectory in combination with the sub-millimeter-level fine measurement data of adjacent frames, to eliminate the positioning errors introduced by unstable handheld, and to ensure that the spatial error of the corrected point cloud data is not more than the sub-millimeter-level precision requirement.

10. A plug-in handheld laser scanner based on fine positioning of a light spot, characterized in that, A computer program product comprising a computer readable medium, the computer readable medium having stored thereon the computer program of claim 10. A computer program comprising program code adapted to perform the method of any one of claims 1 to 9 when the program is executed on a computer.

Images