Pre-scanning function for laser scanner

Abstract

The method is used for the pre-scanning function of the laser scanner. The invention relates to a laser line scanner for providing a 3D geometric scan of a surface comprising a first surface portion having a first reflectivity and a second surface portion having a second reflectivity. A laser line scanner of the present invention performs pre-scan measurements under changes in projection parameters, detector parameters, and processing parameters of a projector, a detector, and an estimation unit of the laser line scanner. And the trained machine learning model returns a suggested parameter set based on the acquired pre-scanning data.

Description

Technical Field

[0001] The present invention relates to a laser scanner, and more particularly to a laser line scanner configured to perform a pre-scan for optimizing scanning parameters. Background Technology

[0002] Laser scanners play a crucial role in parts inspection. They are capable of measuring parts with high precision in the micrometer range and are commonly used in many industries, such as aerospace, automotive, and various sectors within industrial applications, including manufacturing, machine production, and metal processing. This invention relates to a laser scanner that projects a pattern onto a surface and determines the surface geometry by analyzing the imaged pattern.

[0003] A pattern comprising one or more continuous portions will be referred to herein as a line pattern. It will be apparent to those skilled in the art that such a line pattern may comprise more than one line, and in particular, a line pattern may be a two-dimensional pattern (i.e., a grid pattern). A line pattern may also comprise points or a two-dimensional planar representation. Furthermore, the pattern may be dynamic, i.e., the geometry of the pattern may change over time, particularly for providing a sweeping scan of a surface. All these embodiments and suitable alternatives are considered line patterns in the sense of the present invention. A laser scanner configured to generate such a pattern will be referred to as a laser line scanner. In particular, WO 2014 / 091307 A2 discloses an example of such a laser line scanner that projects a pattern onto a surface. A laser line scanner differs from laser scanners that utilize rotating or oscillating beam deflection elements, such as those discussed in US 2021 / 0149030 A1.

[0004] The imaging pattern is then analyzed using triangulation methods to provide a point cloud representation of the scanned object or a suitable alternative three-dimensional digital representation. In particular, the scanner can have more than one detector element to acquire detector images from different viewpoints. Triangulation in the context of this invention also includes methods utilizing the imaging pattern and time series of the scanning instrument's movement.

[0005] The quality of 3D scan data is influenced by many factors, such as the color, gloss, geometry, ambient light, scanner resolution, and appropriate selection of the scanning segment. In particular, known surface color and gloss are significant sources of error. Existing technologies propose empirical selections of the scanner's sensor aperture and shutter speed based on the characteristics of the scanned surface. However, such selections heavily rely on the operator's skill level and accumulated experience. Knowledge transfer primarily occurs at the operator training level. That is, even when a solution to the optimal scanning parameters exists in principle, a long learning curve may be required. Even more complex are cases where the optimal scanning parameters for an unknown object must be derived through interpolation or extrapolation based on known objects.

[0006] Recently, alternatives to the entirely manual process described above have emerged. In particular, the introduction of dynamic exposure settings and automatic noise reduction allows for significant improvements in scan usability and data quality. Of particular relevance is image acquisition using high dynamic range (HDR) mode. In HDR mode, an image is created based on several (typically three) partial images. Each image is captured at a different exposure setting, specifically the exposure duration. HDR mode ensures that points from the darkest to the brightest areas are clearly visible. However, this process can sometimes make images appear noisier or coarser because it occupies a higher dynamic range at the same bit depth. Despite this drawback, the ability to capture more points in an image generally outweighs any issues with noise. However, for extremely dark or glossy objects, it may be impossible to capture all points in the point cloud, requiring manual setting.

[0007] Existing methods are also time-consuming, especially since operators can only verify scan results after data acquisition is complete. In some cases, parameters may be too far from the optimal values ​​where scans must be discarded, necessitating a new acquisition. In the worst case, iterative methods may be required, involving many individual, unsuccessful scans. This can lead not only to inefficient work but also to increased frustration and psychological stress on operators, as well as reduced job satisfaction.

[0008] Therefore, there is a strong desire to find a robust method that allows setting optimal parameters before or at the start of a scan. This would not only reduce inspection time and effort but also produce better and more reproducible results.

[0009] The purpose of this invention

[0010] In view of the above, one object of the present invention is to improve the optimization of scanning parameters of a laser line scanner, and in particular to increase the parameter space accessible in the optimization.

[0011] Another objective of this invention is to improve the accumulation and transfer of knowledge regarding the optimization of the parameters. Summary of the Invention

[0012] This invention relates to a laser line scanner for providing 3D geometric scanning of a surface. The surface includes a first surface portion having a first reflectivity and a second surface portion having a second reflectivity. The second reflectivity is less than the first reflectivity. Reflectivity, in the context of this invention, is an empirical measurement of a given surface portion under normal measurement conditions that reflects an emitted laser line toward the detector of the laser line scanner. In other words, while reflectivity can refer to a material property of the surface, such as the Fresnel reflection coefficient, and is synonymous with it for diffuse scattering surfaces, reflectivity can also be influenced by further properties, such as surface quality or surface shape. Particularly for objects with strong specular reflection, the shape of the object can also be decisive, i.e., most of the laser light may not be reflected in the direction of the detector, but rather in different directions. In the context of this invention, reflectivity specifically refers to the reflectivity over a particular wavelength range associated with the laser.

[0013] Those skilled in the art will understand that the laser line scanner of the present invention can also be used to scan surfaces with more complex geometries and material properties, such as various components with different reflectivities, colors, and scattering characteristics. Furthermore, the application of the laser scanner of the present invention is not hindered during scanning of substantially uniform surfaces.

[0014] As an example, many aspects and specific embodiments of the invention can be illustrated using an object having a glossy portion and a dark, matte portion, such as a chrome-plated bumper of a vintage car that serves as a first surface portion against a dark-painted body that serves as a second surface portion. While these objectives are challenging for prior art laser line scanners, the invention is advantageously applicable to situations where the contrast difference between the first and second surface portions is not significant by providing optimal measurement parameters.

[0015] The laser line scanner includes a projector unit, a detector unit, an estimation function, a performance evaluation function, a parameter database, a pre-scanning function, and a prediction function.

[0016] The projector unit is configured to project a laser line pattern onto a surface according to settable projection parameters. The laser line pattern includes at least one substantially continuous laser line portion. The projection parameters are used by the projector unit to set different projected laser intensities. The projection parameters may include additional parameters, particularly the geometry and / or time sequence of the laser line. The laser intensity in the context of this invention can represent an intensity distribution. In particular, the laser line pattern can be implemented as a coded pattern with spatial intensity variations.

[0017] The projector unit may also include additional light sources. These additional light sources may include marker or dot pattern projectors or diffuse light sources. The latter include LED light sources, particularly implemented as ring lights, to provide diffuse background illumination of objects. When the projector unit includes such elements, the projection parameters may also include data for setting these additional light sources. The projection parameters may specifically include data regarding the wavelength and intensity of the additional light sources.

[0018] Those skilled in the art will understand that the term "projector unit" refers to a functional unit. A projector unit may include various structural elements, particularly when it includes multiple light sources. This is similarly applicable to other components of a laser line scanner.

[0019] The detector unit is configured to provide a projected laser line pattern to a detector image about the surface according to settable detector parameters. The detector parameters are used to set different photosensitive sensitivities for the detector unit. The detector image can be a color or grayscale image about the surface. The information content of the detector image can be limited to a recorded image of the laser lines. In particular, the detector can be configured to acquire the reflected laser line pattern at pixel resolution. The detector can have a specific narrowband filter to suppress ambient light.

[0020] Laser line scanners typically include multiple detector elements spaced apart to perform triangulation of the laser line. The detector image, in the sense of this invention, can represent the totality of those images provided by those detector elements. Detector parameters can include multiple sets associated with each of these detector elements.

[0021] The estimation function is configured to provide triangulation-based 3D geometric data about the surface based on detector images and a set of processing parameters. The processing parameters include different processing settings for providing the 3D geometric data. The 3D geometric data can be specifically implemented as or configured to provide a point cloud representation of the surface.

[0022] The performance evaluation function is configured to extract key performance indicators (KPIs) and provide scan quality values ​​based on these KPIs. KPIs reflect a set of assessments regarding the quality of the surface itself and the 3D geometry data.

[0023] In the context of this invention, KPI, in one aspect, represents surface-related information including an assessment of at least a first reflectivity and a second reflectivity. KPI may include additional similar data, particularly data regarding color and shape, respectively, associated with the first and second portions. Such information can be provided from independent measurements or manually. Reflectivity can also be provided by directly analyzing the light intensity of reflected laser segments.

[0024] KPIs also include assessments reflecting the quality of detector images and / or 3D geometric data. Such data can be derived, in particular, by analyzing 3D geometric data and / or the corresponding detector images. KPIs include assessments reflecting at least the signal-to-noise ratio of the detector image, the sharpness of the detector image, and the point density of the 3D geometric data, particularly regarding outliers and voids in the 3D geometric data. Scan quality values ​​can be considered as a composite assessment based on this portion of the KPIs. KPIs can provide such assessments about areas associated with specific surface portions rather than the entire detector image or 3D geometric data.

[0025] The parameter database includes a different set of projection parameters, a different set of detector parameters, and a different set of processing parameters. It will be clear to those skilled in the art that although the projection parameters, detector parameters, and processing parameters are not completely decoupled, they represent fundamentally different aspects of the laser line scanner. Therefore, they can be adjusted largely independently of each other. Thus, finding the optimal settings is a multidimensional fitting problem.

[0026] Although the parameters are independent, the actual implementation of a laser scanner may be limited by related variations in the parameters. On the one hand, such limitations may be necessary to protect the operator and / or the equipment. For example, the maximum laser intensity may not be selected together with the maximum sensitivity of the detector unit. On the other hand, such related variations may also limit the parameter space by excluding settings that are clearly ineffective.

[0027] The pre-scan function is configured to perform a set of different pre-scan measurements. Each of the pre-scan measurements includes: (i) providing a selected set of pre-scan parameters, including projection parameters, detector parameters, and processing parameters, based on the parameter database; (ii) acquiring a pre-scan detector image associated with the pre-scan parameters by projecting the laser line pattern according to the selected projection parameters and imaging the surface according to the selected detector parameters; and (iii) deriving corresponding pre-scan 3D geometric data according to the selected processing parameters through the estimated functionality. The pre-scan function may also include deriving pre-scan KPIs through a performance evaluation function. Preferably, the steps are performed rapidly and continuously. In particular, the pre-scan consists of continuous data acquisition, during which parameters may be automatically cycled. Such a pre-scan cycle can be performed at regular intervals during the main scan, and / or the main scan sequence can be performed with a degree of variation to provide such a pre-scan cycle. However, the invention is not limited to such an implementation. Furthermore, variations in the execution of the steps are also possible, particularly since image acquisition and the derivation of 3D geometric data can be performed independently of each other.

[0028] Typically, this pre-scan must be performed within a reasonable time frame. In other words, a pre-scan can only test a representative set of parameters. While pre-scan parameters can be adapted to the type of object being studied, it is unlikely that any single pre-scan measurement will be performed with the optimal set of parameters.

[0029] The prediction function is configured to (i) evaluate performance threshold criteria for desired scan quality values, particularly for the desired signal-to-noise ratio of the detector image, the desired sharpness of the detector image, and the desired point density of the 3D geometry data; (ii) access a trained model configured to generate a set of proposed parameters and corresponding estimated scan quality values, wherein the trained model is trained based on a training database including scan data and ground-based parameters; (iii) perform inference by acquiring input data and passing the input data to the trained model, wherein the input data includes at least one of (a) pre-scan parameters, and (b) pre-scan 3D geometry data, pre-scan detector images, and pre-scan key performance indicators; and (iv) provide a set of proposed parameters as output data for inference, such that the estimated scan quality values ​​associated with the set of proposed parameters meet the performance threshold criteria.

[0030] Performance thresholds can be pre-programmed and represent aggregated optimal values ​​that satisfy certain sufficiency constraints for various performance metrics. Threshold criteria can be provided by the user to optimize specific performance metrics, such as signal-to-noise ratio and / or point cloud density and / or the minimum number of outliers.

[0031] The trained model can be provided by a convolutional neural network. In particular, a supervised machine learning model can be applied, which is trained to map pre-scanned features from the training set to their corresponding ground-based optimal parameters defined by experts.

[0032] Alternatively or additionally, adaptive learning can be used to facilitate ongoing optimization of the model. In other words, a typical operator might repeatedly scan a finite array of objects. By incorporating features that enable on-the-fly learning, the model can gradually familiarize itself with frequently scanned objects. After each scan of similar objects, the model evaluates the quality of the output, adjusts its parameters, and uses the updated model for future scans. For each loop, the model becomes better at determining the optimal settings for that particular object. The evaluation of scan quality, which is crucial to the feedback loop, can be based on a range of metrics such as point density, sharpness, quality of extracted texture, noise level, etc. Once these metrics are determined and set as KPIs, they can also be used as "rewards" in a reinforcement learning framework that guides the model to optimize the parameters of the laser scanner, thereby improving future scan results.

[0033] Therefore, machine learning models (especially when training in operation) can access a measurement pool or knowledge base during the training and retraining phases. The measurement pool can represent measurements performed by most of the organization's experienced operators. Alternatively, the measurement pool can represent an external knowledge base pre-provided by an expert organization (particularly the manufacturer or distributor of laser line scanners). In another alternative, the measurement pool can represent a closed or open knowledge pool accessible to subscribers or anyone to upload their own measurement data.

[0034] Machine learning utilizes the training database (especially with annotations of optimal parameters manually defined by experts) covering a variety of surfaces to map pre-scanned features to ground-based parameters.

[0035] During potential retraining, the model begins by making initial predictions of the optimal parameters for a given object using pre-scan measurements. A feedback loop is then created. After each scan of similar objects, the model evaluates the quality of the proposed parameters, adjusts itself, and utilizes the updated model for future scans. In other words, the training database and the trained model evolve with the accumulation of data. The model learns from multiple parts / objects, accumulating knowledge of the optimal parameters and storing it in a measurement pool. This knowledge can be used for the next scan / measurement. For each loop, the model should become better at determining the optimal settings for that particular object.

[0036] Therefore, the laser line scanner of the present invention can effectively provide optimized solutions for projection parameters, detector parameters, and processing parameters by inferring suggested parameters from pre-scan measurements. This is particularly advantageous for objects that have not been previously measured and / or for inexperienced operators. Furthermore, efficient knowledge transfer is achieved by digitizing and structuring the results of previous experience.

[0037] In some implementations, the laser line scanner is configured for handheld operation. However, even such a handheld laser scanner may include a corresponding interface for mounting the laser line scanner on a specific actuator (e.g., on a robotic arm).

[0038] In some implementations, the projection parameters include data for setting the temporal or spatial variation of the laser intensity projected by the projector unit.

[0039] In the context of this invention, spatial variation in laser intensity can mean spatial variation within a single laser line, particularly including a coded pattern comprising domains with higher and lower laser intensities. Spatial variation can also represent multiple laser lines, particularly arranged in a grid, and more particularly orthogonal grids. Clearly, the pattern represents a projection pattern. Or in other words, the projector unit is configured to set the projection direction in angular space, and the pattern appearing on the surface is also influenced by the surface geometry. Those skilled in the art will understand that the grid can be adjusted based on the shape of the surface to be scanned.

[0040] A non-exhaustive list of time variations includes (i) HDR-like projections with a stepped system variation in laser intensity, (ii) continuous deflection of the projection pattern, for example, by rotating a prism element, (iii) variations in diffuse / background illumination intensity compared to laser intensity, (iv) variations in pattern geometry, such as a projection comprising a first pattern of lines oriented to a first direction and subsequently a second pattern of lines oriented to a second direction, or (v) a projection of a first grid pattern at a first grid interval and subsequently a second grid pattern at a second grid interval. Those skilled in the art will understand that combinations of the listed embodiments (such as projecting two sets of lines and rotating them in different directions) or suitable alternatives are also possible in the context of this invention.

[0041] In some embodiments, the laser line pattern includes a first laser line and a second laser line. The first laser line has a higher intensity than the second laser line.

[0042] In some embodiments, the projector unit is configured to project a laser line pattern comprising multiple laser lines. The laser line pattern can be specifically implemented as a grid pattern of laser lines. In such embodiments, the projection parameters include data regarding the number of laser lines and the distance between them.

[0043] In some specific embodiments, the projection parameters define (i) a first laser line pattern having a first laser intensity and a first distance between laser lines, and (ii) a second laser line pattern having a second laser intensity and a second distance between laser lines. The second laser intensity and / or the second distance are different from the first laser intensity and the first distance, respectively.

[0044] In some implementations, the detector parameters include data for setting the time-varying light sensitivity of the detector unit.

[0045] In some specific embodiments, detector parameters define a first partial image and a second partial image, wherein the second partial image is acquired with higher light sensitivity. In other words, the detector operates in HDR mode. Preferably, the sensitivity is selected such that the first partial image is associated with a first surface portion, and the second partial image is associated with a second surface portion. The laser line scanner of the present invention enables optimization of the HDR mode by taking into account the reflectivity (particularly color and geometry) of the surface portions. Furthermore, the dynamic range of the image is extended only to the necessary extent. This is beneficial for optimizing the signal-to-noise ratio and / or depth of detail. This step can be generalized accordingly to surfaces with more than two feature portions.

[0046] In some specific embodiments, the proposed parameters are configured to provide a temporal change in the sensitivity of the detector unit synchronized with the temporal variation of the projected light intensity. The aforementioned HDR mode can provide even more optimized results through the synchronized variation of the projected light intensity and the sensitivity of the detector unit. Such synchronization is impossible in the prior art because the parameter space is too complex to provide meaningful results. However, the system of the present invention can perform this optimization by relying on a large number of prior measurements on similar objects and machine learning.

[0047] In some embodiments, the detector parameters include data regarding: (i) acquisition time or integration time, (ii) aperture opening, (iii) optical filter data, particularly the filter bandwidth and suppression factor, (iv) gain, and (v) noise threshold. Unlike prior art systems, specific embodiments of the laser line scanner of the present invention enable the optimization of all these parameters in an efficient and meaningful manner. Those skilled in the art will understand that alternative embodiments of the laser line scanner of the present invention may optimize only a subset of these parameters.

[0048] In some implementations, the estimation function is configured to provide noise reduction and / or outlier identification and / or smoothing and / or interpolation filling of gaps in image and / or 3D geometry data. Processing parameters include data regarding the noise reduction and / or outlier identification and / or smoothing and / or interpolation filling of the gaps.

[0049] In some implementations, the processing parameters include data regarding brightness and contrast adjustments as well as photosensitivity measurement corrections.

[0050] In some specific embodiments, the detector image includes a first partial image obtained using a first projection and detector parameters, and a second partial image obtained using a second projection and detector parameters. The processing parameters include data regarding the spatial resolution blending of the first and second partial images. In other words, the processing parameters include data regarding performing image reconstruction for HDR mode.

[0051] In some implementations, the laser line scanner is configured to (i) perform additional pre-scan measurements using a proposed set of parameters, and (ii) perform a comparison of scan quality values ​​associated with the pre-scan measurements and the additional pre-scan measurements. Specifically, the laser line scanner is configured to accept the proposed set of parameters based on the comparison. In other words, the laser line scanner is configured to verify whether the proposed set of parameters represents an optimized solution compared to the pre-scan parameters.

[0052] In some implementations, the laser line scanner includes a camera. The camera may be an RGB camera or an alternative color camera. The camera is configured to provide a camera image including radiometric measurement information about a first surface portion and a second surface portion. In the case of a color camera, the camera image also includes color information. This information can be used to provide estimates of the first and second reflectivities, and to employ pre-scan measurement and / or prediction functions accordingly.

[0053] In some specific implementations, the pre-scan measurement further includes (i) acquiring camera images of representative portions including the first and second surface portions, (ii) deriving a first and a second reflectance based on the camera images, and (iii) having a prediction function provide pre-scan parameters based on the first and second reflectances and a trained model, such that the pre-scan parameters are associated with estimated scan quality values ​​that meet performance threshold criteria and match the first and second reflectances. Alternatively, a coarse first optimization is performed using the pre-scan, and a second fine optimization is performed based on the adjusted pre-scan parameters. In this way, effective optimization can be performed.

[0054] In some specific implementations, the pre-scan further includes (i) deriving first color data corresponding to a first surface portion and second color data corresponding to a second surface portion based on the camera image, and (ii) further providing pre-scan parameters based on the first color data and the second color data, such that the pre-scan parameters are associated with matching the first color data and the second color data.

[0055] In some implementations, the laser line scanner is configured to perform a master scan measurement, wherein the master scan measurement provides master scan 3D geometry data based on a proposed set of parameters. Specifically, the laser line scanner is configured to derive master scan KPIs and master scan quality values ​​via a performance evaluation function.

[0056] In some implementations, the laser scanner is configured to access remote computing resources, particularly cloud computing resources, and the trained model is provided by the remote computing resources.

[0057] In some specific embodiments, the laser line scanner also includes a post-scan feedback function. This post-scan feedback function is configured to (i) receive user-perceived quality feedback regarding the 3D geometric data, and (ii) provide a set of user feedback and suggested parameters, particularly key performance indicators and scan quality values ​​for the master scan, to a remote computing resource hosting the trained model and / or training database. Advantageously, this feedback can be used to improve the model. Those skilled in the art will understand that the user feedback may also include data about the user's skill level, and feedback from more experienced users may have higher weight in such retraining. That is, the feedback data can be aggregated and periodically used to retrain / refine the model over time.

[0058] In some implementations, the laser line scanner is configured to receive scan parameters provided by the user. The laser line scanner is configured to perform an alternative master scan using these user-provided scan parameters. The post-scan feedback function is also configured to provide the user-provided parameter set, key performance indicators of the alternative master scan, and scan quality values ​​of the alternative master scan to remote computing resources hosting the trained model and / or training database. In other words, an expert user can manually provide a supposedly better solution, particularly for testing solutions provided by the trained model. The user-provided solutions (especially comparisons between the user's solution and the model) can be used in a subsequent retraining framework.

[0059] The present invention also relates to a computer program product. This computer program product includes program code stored on a machine-readable medium or specifically implemented as electromagnetic waves. The program code includes program code segments and has computer-executable instructions for performing computational steps associated with pre-scanning and / or prediction functions.

[0060] The computer program product can execute, specifically, on a computing unit that controls the laser line scanner. The computing unit can be an external computer, such as a general-purpose computer implemented as a tablet, portable computer, or smartphone. The computing unit can be a computer or controller integrated into the laser line scanner. The computing unit can be a distributed computing resource with integrated and external components.

[0061] The computer program product of the present invention can also be implemented in various ways. For example, as a standalone computer program product, as a specific application or subroutine of a more general control program for a laser line scanner, or as an update to an existing pre-scanning function of a prior art control program. The present invention is not limited to any of these implementations.

[0062] Some implementations of the computer program product (particularly when installed on the control computer of a laser line scanner) when executed cause the laser scanner to (i) provide pre-scanning parameters based on a parameter database, the pre-scanning parameters including a selected set of projection parameters, detector parameters, and processing parameters; (ii) acquire a pre-scanning detector image based on the projection parameters and detector parameters; (iii) derive corresponding pre-scanning 3D geometric data based on the selected processing parameters, particularly deriving key pre-scanning performance indicators; and (iv) access performance threshold standards regarding the desired scan quality values, particularly regarding the desired signal-to-noise ratio of the detector image, the desired sharpness of the detector image, and the desired points of the 3D geometric data. The density performance threshold criterion includes: (v) accessing a trained model configured to generate a set of proposed parameters and corresponding estimated scan quality values, wherein the trained model is trained based on a training database including scan data and ground-based parameters; (vi) performing inference by acquiring input data and passing the input data to the trained model, wherein the input data includes pre-scan parameters and at least one of pre-scan 3D geometry data and / or pre-scan detector images and / or key performance indicators; and (vii) providing the set of proposed parameters as output data for the inference, such that the estimated scan quality values ​​associated with the set of proposed parameters satisfy the performance threshold criterion.

[0063] In some implementations, the computer program product also has computer-executable instructions for causing the laser line scanner to perform a scan using a recommended set of parameters. Attached Figure Description

[0064] By way of example only, specific embodiments of the invention will be described more fully below with reference to the accompanying drawings, wherein:

[0065] Figure 1 A laser line scanner is schematically shown that projects a pattern onto an object;

[0066] Figures 2a to 2l An example of a line pattern in the sense of this invention is shown;

[0067] Figure 3 An example is shown of providing 3D geometric data based on the acquisition of a first portion of the image with low light sensitivity and a second portion of the image with high light sensitivity;

[0068] Figure 4 The content of an exemplary KPI is illustrated schematically;

[0069] Figure 5 The steps of the pre-scan measurement are illustrated schematically using a flowchart;

[0070] Figure 6The implementation of the prediction function is illustrated using a flowchart;

[0071] Figure 7a This demonstrates how to train a trained model by providing expert measurements;

[0072] Figure 7b This demonstrates how a trained model is trained through supervised learning.

[0073] Figure 7c The trained model is shown through reinforcement learning. Detailed Implementation

[0074] Figure 1 An operator using a laser line scanner 1 is schematically illustrated when scanning a surface 2 of an object. Surface 2 includes a first surface portion 21 with a first reflectivity (matte white in this example) and a second surface portion 22 with a second reflectivity (matte black in this example). The laser line scanner 1 includes a projector unit 11 that projects a laser line pattern 110 (here composed of a single laser line) onto the surface 2. The laser line scanner 1 also includes a detector unit 12, which is depicted herein as a stereo camera pair. The detector unit 12 acquires a detector image of the surface 2 using the projected laser line pattern 110. Many embodiments of detector images are known to those skilled in the art. Fundamentally different is the detector image, in which only the laser line pattern 110 is imaged, and the detector image includes environmental information about the entire surface 2. Regardless of the type of detector image, challenges are shown in imaging the surface 2 with the first surface portion 21 having the first reflectivity and the second surface portion 22 having the second reflectivity. In particular, the observability of the laser line pattern 110 differs on the bright first surface portion 21 and the dark second surface portion 22.

[0075] In the depicted embodiment, one of the cameras in the stereo camera pair 12 also acts as color 17, particularly an RGB camera, which is configured to acquire color information about surface portions 21, 22. This color information can also be processed by the laser line scanner of the present invention. Those skilled in the art will understand that alternative embodiments (particularly where the color camera 17 differs spatially and functionally from the detector unit 12) are also possible.

[0076] The depicted laser line scanner 1 also includes a controller 13 or computing unit. The controller 13 may provide various functions, particularly estimation, performance evaluation, pre-scanning, and prediction functions. Furthermore, the controller 13 may also include non-transitory memory where a parameter database is stored. The depicted laser line scanner 1 includes a wireless connection 149 for accessing a training database. The wireless connection 149 can be used to access a remote computing resource 148, depicted as a desktop computer, to perform some of the aforementioned functions. Alternatively or additionally, a wired connection may also be used for the same purpose. Those skilled in the art will understand that the remote computing resource 148 may be distributed, particularly a cloud-based resource.

[0077] Figure 1 A handheld laser line scanner 1 is depicted. The invention is not limited to this embodiment or application. In particular, the invention is equally applicable to mounted laser line scanners, especially those mounted to robotic arms or vehicles. The handheld laser line scanner 1 may include a suitable interface or may have specific extensions for installation and operation.

[0078] Figures 2a to 2l Some exemplary embodiments of laser line patterns are shown. The examples shown are not exhaustive, and in particular, multiple aspects from the depicted embodiments can be combined with each other.

[0079] Figure 2a A laser line pattern consisting of a single continuous segment is shown, or a laser line 150 having a substantially constant intensity within a segment 150.

[0080] Figures 2b to 2f An exemplary implementation of spatial variation is schematically depicted. Figure 2b Laser line patterns with spatial intensity variations are shown, particularly implemented as a single-line coded pattern 151. Such coded patterns are advantageous because they provide, for example, further data regarding the orientation and distance of the laser line scanner to the surface. Figure 2c A linear pattern 111 is depicted with substantially similar laser lines 150 having equidistant projections. Figure 2d A laser line pattern with two laser lines 152 and 153 is depicted. The first laser line 152 has a higher intensity than the second laser line 153. Similar to... Figure 2b The single-line coded pattern 151 and the different features of laser lines 152 and 153 provide additional information, especially additional information about the center line. Figure 2e Another alternative is described, in which different intensities of the lines and different distances between the lines 162, 163 provide such information. Figure 2fA grid pattern 112 is depicted, wherein a first set of lines facing a first direction and a second set of lines facing a second direction are projected. This grid pattern can also be provided by alternating projections of lines.

[0081] Figures 2g to 2j An example of the time-varying pattern of the laser line is provided. Figure 2g In this method, a laser line is scanned within a range by moving 164 laser lines in one direction. As an example, this can be achieved using tiltable deflecting elements such as mirrors or prisms. Figure 2h An implementation is described in which the laser intensity of laser lines 154-156 is systematically varied, for example in HDR-like lighting. Figure 2i The illustration depicts an embodiment in which two different grid patterns 113 and 114 are subsequently projected at different distances between laser lines. Alternatively or additionally, the laser intensity may also be varied. Figure 2j An embodiment is depicted in which two laser lines 157 and 158 rotate in different directions 165 and 166.

[0082] Figures 2k to 2l One embodiment is described, in which additional light also illuminates the surface. Figure 2k In particular, the marking light 171 is projected onto the surface by a light source independent of the laser line scanner, while... Figure 2l In this process, diffuse background illumination 172 is provided, for example, by LEDs integrated into the laser line scanner. The relative intensity of the laser lines 150 with respect to these additional lights 171, 172 can also be one of the projection parameters.

[0083] Those skilled in the art will understand that the projection parameters in the sense of this invention provide for, for example Figure 1 and Figures 2a to 2l The variable parameters involved in the generated laser line patterns 110-114 are shown. The laser line scanner of the present invention optimizes the projection parameters and the resulting laser line patterns 110-114 to improve the quality of the obtained 3D geometric data.

[0084] Figure 3 A surface 2 is shown, comprising a first surface portion 21 having a first reflectivity 210 and a second surface portion 22 having a second reflectivity 220. The second reflectivity 220 is less than the first reflectivity 210. That is, the first surface portion 21 produces a higher intensity of reflected light, as indicated by the darker hue of the first surface portion 21. The depicted surface 2 has an edge 29 in the second surface portion 22.

[0085] The laser line pattern 110 projected onto surface 21 (in this example, it consists of only a single line) is also affected by different reflectivities 210 and 220. The first portion of the laser line 121 reflected from the first surface portion 21 is more intense than the second portion of the laser line 122 reflected from the second surface portion 22.

[0086] To achieve optimal light intensity, detector parameters are set such that the detector sequentially acquires two partial images 201 and 202. The second partial image 202 is acquired with a higher light sensitivity than the first partial image 201; for example, during the acquisition of the second partial image 202, the acquisition time, aperture, or gain is set to be higher. The higher light sensitivity compared to the first partial image 201 increases the visibility of the second portion of the laser line 122 in the second partial image 202. As a disadvantage, the noise level of the second surface portion 22 increases, and the first portion of the laser line 121 becomes oversaturated, thus reducing its positional accuracy. Furthermore, the contrast between the laser line and the unilluminated portion of the first surface portion 21 decreases.

[0087] In the next step, estimation function 3 generates detector image 201 from partial images 201 and 202. In the depicted example, this is performed such that the first partial image 201 is associated with the first surface portion 21, and the second partial image 202 is associated with the second surface portion 22. Finally, 3D geometry data 300 is reconstructed based on detector images 200, particularly using a set of detector images 200, wherein the laser line pattern 110 occupies different locations on surface 2. Estimation function 3 may include additional steps such as noise suppression, smoothing, and interpolation of missing portions of the 3D geometry data 300.

[0088] Figure 3 Many aspects of HDR acquisition depicted in the invention are known in principle in the prior art. However, such prior art systems use preset acquisition and / or projection parameters to perform acquisition. The laser line scanner of the present invention performs the depicted acquisition and data estimation in an adaptive manner. That is, it performs a pre-scan measurement sequence under variations in projection parameters, detector parameters, and processing parameters, and infers optimized parameters based on applying a trained model to the pre-scan measurements.

[0089] Figure 4The diagram schematically illustrates the generation of KPI 500 and scan quality value 599 by performance evaluation function 5. In the depicted embodiment, KPI 500 is generated based on detector image 200. Additionally or alternatively, the KPI may include an evaluation of 3D geometric data. The depicted detector image includes data about a surface comprising a first surface portion 21 having a first reflectivity 210 and a second surface portion 22 having a second reflectivity 220. Edges 29, as geometric features, are also present in the second surface portion 22. The detector image also includes an imaged laser line pattern 110, which includes a first portion 121 reflected by the first surface portion 21 and a second portion 122 reflected by the second surface portion 22.

[0090] The depicted KPI 500 includes evaluations of the first surface portion 21 and the second surface portion 22 as individual elements. This detail is beneficial because it allows for more accurate predictions when the trained model is applied to pre-scan measurements. KPI 500 includes data on the first reflectivity 210 and the second reflectivity 220. This data can be derived from the detector image 200 itself, for example, by analyzing the intensity of the laser line in the image with respect to projection and imaging parameters. Additional measurements can be used to determine the reflectivity 210 and 220. In addition to reflectivity, other surface-related data can be extracted into KPI 500, particularly data specific to surface color or orientation.

[0091] KPI 500 may include image quality-related data. In the depicted example, signal-to-noise ratios 501 and 502 and sharpness 511 and 512 associated with the corresponding surface portions 21 and 22 are extracted, particularly the sharpness of edge 29. Although not shown, KPI 500 may also include data regarding 3D geometric data, such as point densities 521 and 522. Based on KPI 500, performance evaluation function 5 also provides a scan quality value 599. The scan quality value 599 can be considered as a composite index of the corresponding quality-related features of KPI 500, such as a weighted average. The actual calculation of the scan quality value 599 depends on the scanning task. For example, whether high edge contours or low noise is more important.

[0092] Figure 5The steps performed by an exemplary implementation of the pre-scan function 4 are schematically illustrated in a flowchart. Flow lines / commands are depicted with thick lines, while data lines are depicted with dashed lines. For transparency, some flow lines and / or data lines may not be shown in the schematic flowchart. Furthermore, this flowchart and any other flowcharts focus on certain aspects of the invention; that is, information relating to other aspects can be abstracted. For transparency, only a single pre-scan measurement with a single set of pre-scan parameters 40 is shown. Those skilled in the art will understand that the performed loop provides various pre-scan parameters 40 and a set of KPIs 500 (in particular, detector image 200 and 3D geometric data 300) associated with the corresponding per-scan parameter 40.

[0093] As a first step, pre-scan parameters 40, including projection parameters 41, detector parameters 42, and processing parameters 43, are provided 401 based on parameter database 400. The provision 401 of pre-scan parameters 40 can be performed using a preset sequence independent of the surface to be scanned or in an adaptive manner. In particular, pre-scan parameters 40 can be adjusted during operation based on previously determined KPIs 500. Alternatively or additionally, further input data, particularly an RGB image of the surface, can be utilized during the provision 401 of pre-scan parameters 40.

[0094] This step is followed by the actual pre-scan measurement, which includes the following steps: projecting a 119 laser line pattern according to projection parameter 41 (e.g., as shown in the image). Figures 2a to 2l As shown), acquire 129 detector images 200, and derive 139 3D geometric data 300 based on detector parameters 42 and processing parameters 43 (e.g., as shown). Figure 3 (As shown). Finally, 159 KPIs out of 500 were extracted, where each KPI includes surface-related data (including partial reflectance) and quality-related data (e.g., such as...). Figure 4 (As shown). The extraction of KPI500 159 is optional and can be omitted in an alternative implementation of the pre-scan function 4, where the proposed parameters can be inferred later based on the detector image 200 and / or 3D geometric data 300.

[0095] Figure 6The various aspects of prediction function 6 are illustrated in a schematic flowchart. Prediction function 6 accesses 601 a trained model 600. The trained model 600 is trained based on ground reality provided by experts (e.g., based on model scan parameters 640 and associated model KPIs 650 and model scan quality values ​​699). The trained model uses the data to map pre-scan features to ground reality parameters, thereby providing 661 a parameter space 660. The trained model 600 can be stored in the memory of the laser line scanner. Alternatively, the trained model 600 can be stored on external (particularly network-based) computing resources.

[0096] Parameter space 660 represents projection parameters, detector parameters, and processing parameters, which can theoretically be set by the laser line scanner. In particular, parameter space 660 also represents the type of surface under study, information provided by model KPI 650. Those skilled in the art will understand that parameter space 660 extends beyond actual measurements. Specifically, parameter space 660 covers combinations of parameters that have not been tested or have been further refined beyond the actual scan data. Due to the complexity of the possible adjustments, no simple analytical or numerical model can provide a sufficient description. Essentially, due to its generalization ability, a trained model can suggest parameter values ​​and combinations beyond what it saw during training.

[0097] In the following steps, the implementation of the described prediction function 6 accesses 49, 509 pre-scan KPIs 500 and pre-scan parameters 40. These are provided by the pre-scan function, for example, as Figure 5 As shown. Alternatively or additionally, detector images or 3D geometric data can also be provided to prediction function 6. Hereinafter, pre-scan parameters 40 with a parameter space based on pre-scan KPI 500 and pre-scan measurements are provided as input data for inference step 621. In other words, prediction function 6 positions the pre-scan KPI 500 and pre-scan parameters 40 in the parameter space. In this way, information about parameters that do not change during the pre-scan (especially those fixed by external constraints) can also be obtained. As a result, 621 is provided to “match” parameter space 622. For example, matching parameter space 622 represents the domain of projection parameters, detector parameters, and processing parameters reachable by the laser line scanner, as well as associated surface characteristics, particularly reflectivity, which is fixed.

[0098] Based on this matching parameter space 622, a set of 689 suggested parameters 680 can be provided. In particular, an optimal set of parameters can be provided within the matching parameter space 622. The advantage of this approach is that it can derive complex parameter combinations that cannot be reasonably tested through simple pre-scanning. Furthermore, the prediction function 6 can extract data on untested parameters, for example, suggesting noise thresholds based on the trained model 600 and the matching parameter space 622, even if the processed parameter does not change during pre-scanning measurements. Additionally, the trained model 600 can access previously unmeasured portions of the parameter space 660, for example, because it exceeds the usual evaluation of reasonable parameter combinations and / or has never been refined to the required degree. This is particularly advantageous during measurements of atypical surfaces, where, if possible, many iterations would be required to measure appropriately according to existing techniques.

[0099] Finally, based on the trained model 600, an estimated scan quality 690 is provided for the proposed parameter set 680. The estimated scan quality 690 is then compared 693 to a given performance threshold 670. In the depicted implementation, this is a minimum criterion; that is, if the estimated scan quality 690 exceeds the threshold 670, it is accepted. In this example, the specific calculation step 691 and the specific comparison step 693 are depicted as part of a loop. Those skilled in the art will understand that alternative implementations are also possible. In particular, the matching parameter space 622 may already contain estimated scan qualities 690 associated with the corresponding (matching) model parameters 640. In this case, a reduced parameter space can be provided consisting of the portion of the estimated scan qualities 690 in the matching parameter space 622 that exceed the performance threshold. The proposed parameter set 680 is selected from this reduced parameter space.

[0100] Figure 7aAn implementation is illustrated in which a training database 60 for training a model 600 is generated by an expert performing a scan of a surface 2 using a laser line scanner 1 with a manually defined set of expert parameters 899 or optimal parameters. The trained model 600 is hosted by an external computing resource 148, and the laser line scanner 1 provides at least the extracted KPIs 500, the parameter set provided by the expert, and annotations (or in other words, operator feedback 800) regarding the scan quality using the wireless connection 149. That is, the expert provides pre-scan data and ground-based parameters to the training database 60. Therefore, the training database 60, including model parameters 640 associated with the corresponding model KPIs 650 and model scan quality values ​​699, is updated in this example. Those skilled in the art will understand that the set depicted herein can be used to initialize the trained model 600. Thus, the expert could be an expert from a manufacturer, distributor, or a competency center employee. The depicted method can also be used to update an existing trained model 600 through subsequent measurements. The training database 60 and the trained model 600 can be hosted locally, for example, by an experienced operator within a company. The training database 60 and the trained model 600 can also be provided in the open access model, and the expert can be someone sufficiently qualified to provide operator feedback 800. In this way, the laser line scanner 1, which does not have the predictive function of this invention, can also provide data.

[0101] Figure 7b Supervised learning using neural network 147 (particularly a convolutional neural network) is illustrated. The laser line scanner 1 includes the pre-scanning and prediction functions 6 of this invention. That is, when scanning surface 2, the laser line scanner 1 derives a set of suggested parameters 680 based on a trained model 600. After scanning is complete, the laser line scanner 1 provides the training database 60 with neural network 147, KPI 500, the set of suggested parameters 680, and scan quality values ​​599, optionally accompanied by operator feedback 800. Figure 7a The situation is reversed; there is bidirectional communication between the trained model 600 and the laser line scanner 1. As a result, the training database 60 and the trained model 600 are updated.

[0102] Figure 7cA reinforcement learning framework using neural network 147 is illustrated. Advantageously, this learning does not require an expert operator, as feedback is provided by a prediction function 6 and a performance evaluation function 5. Specifically, the prediction function is based on a set of 6 suggested parameters provided by the trained model 600. The estimated scan quality 690 is then compared with the actual scan quality 599 (i.e., the result of a measurement performed using the set of suggested parameters 680). Advantageously, data collection in this manner is not limited by the availability of an expert operator, as operator feedback is not required. Furthermore, the collected data will be more consistent because it relies on objective quality evaluation. In other words, the trained model 600 can also be trained without human expert intervention by employing a reinforcement learning framework, in which the model is designed as an agent to learn the optimal tuning of parameters based on a reward mechanism, which can be derived from, for example, the scan quality 599.

[0103] It is particularly advantageous when the pre-scan sequence is adapted to allow the neural network 147 to freely explore the parameter space, propose specific sets of suggested parameters 680, observe their impact on scan quality 599, and learn iteratively through trial and error. The reinforcement learning agent evolves over time and improves its decision-making abilities, optimizing its actions based on cumulative rewards. In this way, the agent is given the freedom to explore the parameter space in a more dynamic and productive manner. This creates a flexible system with the ability to learn from both errors and successes, making it significantly more robust in managing unforeseen situations that may arise during the scanning process. In other words, the trained model 600 extended in this way can utilize a parameter space that would be difficult (if not impossible) to reach through expert data collection and / or supervised learning.

[0104] Although the invention has been described above with reference to some specific embodiments, it should be understood that many modifications and combinations of different features of the embodiments can be made. All such modifications are within the scope of the appended claims.

Claims

1. A laser line scanner (1) for providing a 3D geometric scan of a surface (2), the surface comprising a first surface portion (21) having a first reflectivity (210) and a second surface portion (22) having a second reflectivity (220), the second reflectivity (220) being less than the first reflectivity (210), wherein, The laser line scanner (1) includes: - Projector unit (11), the projector unit being configured to project a laser line pattern (110-114) onto the surface (2) according to a settable projection parameter (41) for different projection laser intensities set by the projector unit (11), - Detector unit (12), the detector unit is configured to provide (129) projected laser line patterns (110-114) to a detector image (200) about the surface (2) according to a settable detector parameter (42) that sets different light sensitivities of the detector unit (12). - An estimation function (3) is configured to provide (139) triangulation-based 3D geometric data (300) about the surface (2) based on the detector image (200) and a set of processing parameters (43), wherein the processing parameters (43) include processing settings for providing the 3D geometric data (300). - A performance evaluation function (5) is configured to extract (159) key performance indicators (500) and provide scan quality values ​​(599) based on the key performance indicators (500), wherein the key performance indicators (500) reflect a set of evaluations regarding the following: The first reflectivity (210) and the second reflectivity (220). The signal-to-noise ratio (501, 502) of the detector image (200). The sharpness (511, 512) of the detector image (200), and The point density (521, 522) of the 3D geometric data (300). - Parameter database (400), the parameter database includes a set of different projection parameters (41), a set of different detector parameters (42) and a set of different processing parameters (43). - Pre-scan function (4), the pre-scan function being configured to perform a set of different pre-scan measurements, wherein each of the pre-scan measurements includes Based on the parameter database (400), pre-scan parameters (401) are provided, which include a selected set of projection parameters (41), detector parameters (42) and processing parameters (43). A pre-scan detector image (200) associated with the pre-scan parameters (40) is obtained (129) by projecting (119) the laser line pattern (110-114) according to the selected projection parameters (41) and imaging the surface (2) according to the selected detector parameters (42), and The estimation function (3) derives (139) the corresponding pre-scanned 3D geometric data (300) based on the selected processing parameters (43). -Prediction function (6), the prediction function is configured to Access performance threshold (670) criteria for the required scan quality values, particularly for the required signal-to-noise ratio (501, 502) of the detector image, the required sharpness (511, 512) of the detector image, and the required point density (521, 522) of the 3D geometric data. Access (601) a trained model (600), the trained model being configured to generate a set of proposed parameters (680) and corresponding estimated scan quality values ​​(690), wherein the trained model (600) is trained based on a training database (60) including scan data and ground condition parameters. Inference (621) is performed by acquiring input data and passing the input data to the trained model (600), wherein the input data includes Pre-scan parameters (40), and Pre-scan 3D geometric data (300), and / or pre-scan detector images (200), and / or pre-scan key performance indicators (500), wherein, in particular, the pre-scan measurements also include pre-scan key performance indicators (500) derived (159) by the performance evaluation function (5). Provide (689) the suggested parameter set (680) as the output data of the inference (621) such that the estimated scan quality value (690) associated with the suggested parameter set (680) meets the performance threshold (670) criterion.

2. The laser line scanner (1) according to claim 1, wherein, The projection parameters (41) include data for setting the temporal or spatial variation of the laser intensity projected by the projector unit (11). In particular, The laser line pattern (110-114) includes a first laser line (152) and a second laser line (153), and the first laser line (152) has a higher intensity than the second laser line (153).

3. The laser line scanner (1) according to any one of claims 1 or 2, wherein -The projector unit is configured to project (119) a laser line pattern (110-114) comprising multiple laser lines (150-158), particularly wherein, The laser line pattern (110-114) is specifically implemented as a grid pattern (112-114) of the laser lines (150-158), and - The projection parameters (41) include data regarding the number of laser lines (150-158) and the distances (162, 163) between the laser lines. In particular, The projection parameter (41) is defined - A first laser line pattern, the first laser line pattern having a first laser intensity and a first distance between the laser lines, and - A second laser line pattern, the second laser line pattern having a second laser intensity and a second distance between the laser lines that are different from the first laser intensity and / or the first distance.

4. The laser line scanner (1) according to any one of the preceding claims, wherein, The detector parameters (42) include data on the time-varying changes in the light sensitivity of the detector unit (12). Especially among them The detector parameters (42) define a first partial image (201) and a second partial image (202), wherein the second partial image (202) is acquired with higher light sensitivity, and - The first partial image (201) is associated with the first surface portion (21), and the second partial image (202) is associated with the second surface portion (22).

5. The laser line scanner (1) according to claims 2 and 4, wherein, The proposed parameter set (680) is configured to provide a time-varying sensitivity of the detector unit (12) that is synchronized with the time-varying intensity of the projected light.

6. The laser line scanner (1) according to any one of the preceding claims, wherein, The detector parameters (42) include data regarding the following: -Get time, and -hole opening, and - Optical filter data, and - Gain, and - Noise threshold.

7. The laser line scanner (1) according to any one of the preceding claims, wherein - The estimation function (3) is configured to provide Noise reduction, and / or Outlier identification, and / or Smoothing, and / or Interpolation filling of gaps in the image and / or 3D geometric data (300), and - The processing parameters (43) include data regarding the noise reduction, and / or outlier identification, and / or smoothing and / or gap filling.

8. The laser line scanner (1) according to any one of the preceding claims, wherein, The processing parameters (43) include data regarding the following: - Brightness and contrast adjustment, and -Sensing measurement correction In particular, The detector image (200) includes a first partial image (201) obtained using a first projection and detector parameters (42) and a second partial image (202) obtained using a second projection and detector parameters (42), and the processing parameters (43) include spatially resolved mixed data about the first partial image (201) and the second partial image (202).

9. The laser line scanner (1) according to any one of the preceding claims, wherein, The laser line scanner (1) is configured as follows: - Perform additional pre-scan measurements using the recommended parameter set (680), and - Perform a comparison of the scan quality values ​​(599) associated with the pre-scan measurement and the additional pre-scan measurement. In particular, The laser line scanner (1) is configured to accept the proposed set of parameters (680) based on the comparison.

10. The laser line scanner (1) according to any one of the preceding claims, the laser line scanner further comprising a camera, particularly an RGB camera (17), the laser line scanner being configured to provide camera images including radiometric measurements and color information about the first surface portion (21) and the second surface portion (22).

11. The laser line scanner (1) according to claim 10, wherein, The pre-scan measurement also includes: -Acquire camera images of representative portions including the first surface portion (21) and the second surface portion (22). - Derive the first reflectance (210) and the second reflectance (220) based on the camera image. The prediction function (6) provides the pre-scan parameters (40) based on the first reflectance (210) and the second reflectance (220) and the trained model (600), such that the pre-scan parameters (40) are associated with an estimated scan quality value (690) that meets the performance threshold (670) criterion and matches the first reflectance (210) and the second reflectance (220). In particular, the pre-scan also includes: -Based on the camera image, derive first color data corresponding to the first surface portion (21) and second color data corresponding to the second surface portion (22). - Further, the pre-scan parameters (40) are provided based on the first color data and the second color data, such that the pre-scan parameters (40) are associated with matching the first color data and the second color data.

12. The laser line scanner (1) according to any one of the preceding claims, wherein the laser line scanner is further configured to perform a main scan measurement, wherein, The main scan measurement provides main scan 3D geometry data (300) based on the recommended parameter set (680). In particular, the main scan key performance index (500) and the main scan scan quality value (599) are derived through the performance evaluation function (5).

13. The laser line scanner (1) according to any one of the preceding claims, wherein, The laser scanner is configured to access remote computing resources (148), particularly cloud computing resources, and the trained model (600) is provided by the remote computing resources (148).

14. The laser line scanner (1) according to claims 12 and 13, further comprising a post-scan feedback function, the post-scan feedback function being configured to: - Receive user feedback (800) regarding the quality perception of the 3D geometric data (300), and - Provide the user feedback (800) and the suggested parameter set (680), in particular the main scan key performance indicators (500) and the main scan scan quality value (599), to the remote computing resource (148) hosting the trained model (600) and / or the training database (60). Especially among them -The laser line scanner (1) is configured as Receive scan parameters provided by the user, and Perform an alternative main scan using the scan parameters provided by the user. - The post-scan feedback function is also configured to provide the user-supplied set of parameters, alternative master scan key performance indicators and alternative master scan scan quality values ​​(599) to the remote computing resource (148) hosting the trained model (600) and / or the training database (60).

15. A computer program product for use with a laser scanner according to any one of claims 1 to 14, the computer program product comprising program code stored on a machine-readable medium or embodied by an electromagnetic wave including program code segments, and the computer program product having computer-executable instructions for causing the laser line scanner (1): - Based on the parameter database (400), pre-scan parameters (40) are provided, including a selected set of projection parameters (41), detector parameters (42) and processing parameters (43). - Obtain (129) a pre-scan detector image (200) based on the projection parameters (41) and detector parameters (42). - Based on the selected processing parameters (43), export the corresponding pre-scan 3D geometric data (300), and in particular export (159) the pre-scan key performance indicators (500). - Access performance threshold (670) standards regarding the required scan quality values, particularly performance threshold standards regarding the required signal-to-noise ratio (501, 502) of the detector image, the required sharpness (511, 512) of the detector image, and the required point density (521, 522) of the 3D geometric data. - Access (601) the trained model (600), the trained model being configured to generate a set of proposed parameters (680) and corresponding estimated scan quality values ​​(690), wherein, The trained model (600) is trained based on a training database (60) that includes scan data and ground condition parameters. - Inference (621) is performed by acquiring input data and passing the input data to the trained model (600), wherein the input data includes Pre-scan parameters (40), and Pre-scanned 3D geometry data (300), and / or pre-scanned detector images (200), and / or key performance indicators (500). - Provide the proposed parameter set (680) as the output data of the inference, such that the estimated scan quality value (690) associated with the proposed parameter set (680) satisfies the performance threshold (670) criterion. In particular, the computer program product also has computer-executable instructions for causing the laser line scanner (1) to perform scanning using the recommended parameter set (680).

Images