Working method of line laser 3D profilometer integrating multiple stray light suppression algorithms

Abstract

This invention discloses a working method for a line laser 3D profilometer integrating multiple stray light suppression algorithms. The method includes: constructing a stray light sensitivity weight map based on a multi-exposure image sequence using a deep convolutional neural network combined with a point spread function; using the weight map to guide Gaussian Laplacian pyramid adaptive filtering, and adjusting the filtering intensity with a residual stability index to generate a cleaned intensity image; performing one-dimensional profile analysis based on the cleaned image and the original height map, calculating the morphological distortion index, and extracting the energy concentration through three-level wavelet packet decomposition to trigger morphological inpainting; after inpainting, the image undergoes centerline extraction and motion blur correlation calculation, and anisotropic smoothing of the height map is performed when the image exceeds a threshold; registering the processed multi-frame height maps, constructing a geometric fidelity as feedback to drive a genetic algorithm to optimize initial parameters, and restarting the entire process until the fidelity meets the accuracy requirements. This invention improves the accuracy of 3D imaging through multiple stray light suppression algorithms.

Description

Technical Field

[0001] This invention relates to the field of 3D measurement, and more particularly to a working method for a line laser 3D profilometer that integrates multiple stray light suppression algorithms. Background Technology

[0002] One of the main challenges faced by line laser 3D profilometers in practical applications is stray light interference. This interference originates from ambient light, multipath reflections, and scattered light from non-target areas. Stray light not only increases image noise and reduces the signal-to-noise ratio, affecting height measurement accuracy, but also induces abnormal structures such as false edges and stripe artifacts, misleading centerline extraction and 3D reconstruction. Especially in dynamic scanning scenarios, stray light coupled with motion easily produces time-varying artifacts and dynamic blurring, severely compromising the geometric realism and stability of the reconstruction results. While various techniques have attempted to address these issues, each has its limitations. Given that a single method cannot comprehensively and effectively solve the stray light problem, integrating multiple advanced algorithms becomes necessary. Based on this, this invention proposes a working method for a line laser 3D profilometer that integrates multiple stray light suppression algorithms. Summary of the Invention

[0003] This invention provides a working method for a line laser 3D profilometer that integrates multiple stray light suppression algorithms, including:

[0004] S10. Based on the multi-exposure original image sequence acquired by the line laser 3D profilometer, the signal-to-noise ratio gain factor and local intensity asymmetry of each pixel in the optimal exposure range are predicted by a deep convolutional neural network, and a stray light sensitivity weight map is constructed by combining the point spread function.

[0005] S20. Adaptive threshold filtering is performed on each layer of the image after decomposition of the Gaussian Laplacian pyramid using a weighted graph, and the stability index of the residual features after filtering of each layer is calculated. The stability index is used as the energy attenuation coefficient to dynamically adjust the filtering intensity of the next layer, generating a clean image while optimizing the edge retention rate.

[0006] S30. Using the purification intensity image and the original height map as input, the sharpness gradient and background flatness are calculated by one-dimensional profile analysis to obtain the morphological distortion index sequence. The sequence is then subjected to three-level wavelet packet decomposition to extract the energy concentration. The energy concentration is used as a criterion to determine whether to start the line-by-line stripe repair program based on morphological opening and closing operations.

[0007] S40. Perform centerline extraction on the striped image after the repair procedure, calculate the motion blur correlation index using the three-dimensional reconstruction algorithm. If the index exceeds the preset dynamic threshold, use the inverse diffusion equation to perform anisotropic smoothing on the current frame height map to eliminate artifacts caused by dynamic stray light.

[0008] S50. The processed multi-frame height map is registered based on the iterative nearest point method. The mean cosine of the normal vector cosine is extracted and combined with the global height fluctuation coefficient to generate geometric fidelity. The geometric fidelity is used as a feedback signal to drive the genetic algorithm to optimize the initial weights and restart the entire process until the geometric fidelity converges to its theoretical peak value and meets the preset accuracy requirements.

[0009] The line laser 3D profilometer working method integrating multiple stray light suppression algorithms, as described above, is characterized by the following steps: based on the multi-exposure raw image sequence acquired by the line laser 3D profilometer, a deep convolutional neural network is used to predict the signal-to-noise ratio gain factor and local intensity asymmetry of each pixel in the optimal exposure range, and a stray light sensitivity weight map is constructed by combining the point spread function. Specifically, it consists of the following sub-steps:

[0010] After extracting the original height map from the original image sequence and denoising and correcting it, the optimal exposure range is determined based on the multi-exposure response and physical noise. The signal-to-noise ratio gain factor and local intensity asymmetry are calculated, and the two are jointly predicted and optimized through a dual-branch convolutional neural network to achieve accurate quantification of laser stripe quality and stray light interference.

[0011] By combining physical optics simulation and experimental calibration, the imaging optical path of the line laser 3D profilometer is accurately mathematically characterized to construct a point spread function, and this function is used to simulate the light energy diffusion effect caused by stray light.

[0012] A stray light sensitivity weighting map is constructed based on the diffusion function of the combination of signal-to-noise ratio gain factor and local intensity asymmetry.

[0013] The line laser 3D profilometer method integrating multiple stray light suppression algorithms, as described above, is characterized by: using a weighted map to perform adaptive threshold filtering on each layer of the image after Gaussian-Laplacian pyramid decomposition; statistically calculating the stability index of the residual features after filtering each layer; and using the stability index as an energy attenuation coefficient to dynamically adjust the filtering intensity of the next layer, thereby generating a cleaned intensity image while optimizing the edge preservation rate. Specifically, it consists of the following sub-steps:

[0014] The original intensity image is decomposed into a Gaussian Laplacian pyramid to obtain a high-frequency layer sequence containing multi-scale detail information and a low-frequency basal layer that preserves the overall structure.

[0015] An adaptive threshold filter is designed for each layer of Laplacian detail image based on the stray light sensitivity weight map to achieve effective suppression of stray light and edge protection;

[0016] The contrast and information entropy of the residual images of each layer after adaptive filtering are calculated and fused into an energy attenuation coefficient. Based on this coefficient, the filtering intensity of the next layer is dynamically adjusted to reconstruct a clean intensity map that suppresses noise and preserves edges.

[0017] The above-described method for a line laser 3D profilometer integrating multiple stray light suppression algorithms is characterized by the following steps: Using a purified intensity image and the original height map as input, a morphological distortion index sequence is obtained by calculating the sharpness gradient and background flatness using one-dimensional profile analysis. This sequence is then subjected to three-level wavelet packet decomposition to extract energy concentration. The energy concentration is used as a criterion to determine whether to initiate a line-by-line stripe restoration procedure based on morphological opening and closing operations. Specifically, this involves the following sub-steps:

[0018] Based on the purification intensity image, the main peak of the laser stripe is located by Gaussian fitting combined with the original height map. The peak sharpness gradient variance and background flatness standard deviation in the neighborhood are calculated with the main peak as the center to construct the morphological distortion index sequence.

[0019] The high-frequency detail coefficients of the third layer of the morphological distortion index sequence are extracted by three-level wavelet packet decomposition, and the energy concentration is calculated. Severe distortion is accurately identified at the sequence level, providing a reliable criterion for subsequent restoration.

[0020] Based on the energy concentration results, an energy focusing criterion is introduced to comprehensively evaluate the degree of stripe breakage by combining the filter difference and curvature enhancement term. When the energy focusing criterion is greater than the set threshold, geometric prior is combined to initiate structure-guided interpolation repair to restore the continuity of image stripes.

[0021] The line laser 3D profilometer working method integrating multiple stray light suppression algorithms, as described above, is characterized by combining geometric prior-initiated structure-guided interpolation repair to restore the continuity of image stripes. Specifically, it consists of the following sub-steps:

[0022] Based on the purification intensity image and morphological distortion index, a fracture localization mask is generated by using the low-value region in the morphological distortion index and the detected energy concentration region.

[0023] The purification intensity image is selectively smoothed and bridged along the scanning direction using multi-scale morphological opening and closing and closing-opening filters to smooth and bridge the image fracture region, and a weight map is constructed using the difference between the two to highlight the fracture location.

[0024] Guided by the mask, based on the gradient direction and width characteristics of the stripes on both sides and the geometric constraints of the original height map, linear interpolation guided by the structural tensor is used to adaptively reconstruct the fractured area. The repaired pixels are then updated to the global image, achieving effective restoration of stripe continuity and structural integrity.

[0025] The line laser 3D profilometer working method integrating multiple stray light suppression algorithms, as described above, is characterized by: extracting the centerline of the striped image after the repair procedure; calculating the motion blur correlation index using a 3D reconstruction algorithm; and if the index exceeds a preset dynamic threshold, anisotropically smoothing the current frame height map using an inverse diffusion equation to eliminate artifacts caused by dynamic stray light. Specifically, it consists of the following sub-steps:

[0026] Subpixel-level peak detection and cubic spline interpolation were performed on the repaired stripes to construct a smooth and continuous centerline trajectory, providing a geometric basis for high-precision 3D reconstruction.

[0027] Based on the centerline stripe trajectory, the cumulative sum of curvature changes and normal offset between adjacent frames are calculated through 3D reconstruction to construct a motion blur correlation index and quantify the artifact intensity caused by dynamic stray light.

[0028] Based on the motion fuzzy correlation index, the spatiotemporal anisotropic inverse diffusion equation is used to effectively filter out height field noise caused by dynamic stray light while preserving geometric edges.

[0029] The line laser 3D profilometer working method integrating multiple stray light suppression algorithms, as described above, involves registering the processed multi-frame height maps using the iterative nearest-point method, extracting the mean cosine of the normal vector's consistent angle, and combining it with the global height fluctuation coefficient to generate geometric fidelity. This geometric fidelity is then used as a feedback signal to drive a genetic algorithm to optimize the initial weights and restart the entire process until the geometric fidelity converges to its theoretical peak and meets the preset accuracy requirements. Specifically, it consists of the following sub-steps:

[0030] The iterative nearest point method based on key point matching optimization of three-dimensional scale-invariant feature transformation is adopted to perform high-precision spatiotemporal registration on the denoised multi-frame height maps, effectively overcoming motion blur and stray light interference, and forming a registered point cloud sequence.

[0031] A 3D model is generated by surface reconstruction of the point cloud sequence, and the geometric fidelity index is calculated by the mean of normal vector consistency and the coefficient of variation of height fluctuation of the static region point cloud to evaluate the overall reconstruction quality.

[0032] Using geometric fidelity as feedback, a genetic algorithm is used to dynamically optimize key parameters and restart the entire process. Through convergence judgment and dual verification, the adaptive suppression accuracy and reconstruction robustness of stray light interference are significantly improved.

[0033] This invention also provides a line laser 3D profilometer system integrating multiple stray light suppression algorithms, including:

[0034] Sensitivity weight map module: Based on the multi-exposure raw image sequence acquired by the line laser 3D profilometer, the signal-to-noise ratio gain factor and local intensity asymmetry of each pixel in the optimal exposure range are predicted by a deep convolutional neural network, and a stray light sensitivity weight map is constructed by combining the point spread function.

[0035] Purification intensity map module: Adaptive threshold filtering is performed on each layer of the image after decomposition of the Gaussian Laplacian pyramid using a weight map, and the stability index of the residual features after filtering of each layer is calculated. The stability index is used as the energy attenuation coefficient to dynamically adjust the filtering intensity of the next layer, generating a purification intensity image while optimizing the edge retention rate.

[0036] Stripe Repair Module: Taking the purified intensity image and the original height map as input, it uses one-dimensional profile analysis to calculate the sharpness gradient and background flatness to obtain the morphological distortion index sequence. The sequence is then subjected to three-level wavelet packet decomposition to extract the energy concentration. The energy concentration is used as a criterion to determine whether to start the line-by-line stripe repair program based on morphological opening and closing operations.

[0037] Anisotropic smoothing module: Performs centerline extraction on the striped image after the repair procedure, calculates motion blur correlation index using 3D reconstruction algorithm, and if the index exceeds the preset dynamic threshold, anisotropic smoothing is performed on the current frame height map using the inverse diffusion equation to eliminate artifacts caused by dynamic stray light.

[0038] Registration optimization module: The processed multi-frame height map is registered based on the iterative nearest point method. The mean cosine of the normal vector consistency angle is extracted and combined with the global height fluctuation coefficient to generate geometric fidelity. The geometric fidelity is used as a feedback signal to drive the genetic algorithm to optimize the initial weights and restart the whole process until the geometric fidelity converges to its theoretical peak value and meets the preset accuracy requirements.

[0039] The beneficial effects achieved by this invention are as follows: This invention improves the accuracy of 3D imaging through a variety of stray light suppression algorithms, ensuring the imaging accuracy of the line laser 3D profilometer. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0041] Figure 1 This is a flowchart of a line laser 3D profilometer working method that integrates multiple stray light suppression algorithms, as provided in Embodiment 1 of this application.

[0042] Figure 2This is a schematic diagram of a line laser 3D profilometer system integrating multiple stray light suppression algorithms, provided in Embodiment 2 of this application. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] Example 1

[0045] like Figure 1 As shown, Embodiment 1 of this application provides a working method for a line laser 3D profilometer that integrates multiple stray light suppression algorithms, including:

[0046] S10. Based on the multi-exposure original image sequence acquired by the line laser 3D profilometer, the signal-to-noise ratio gain factor and local intensity asymmetry of each pixel in the optimal exposure range are predicted by a deep convolutional neural network, and a stray light sensitivity weight map is constructed by combining the point spread function.

[0047] Specifically, constructing a stray light sensitivity weight map includes the following sub-steps:

[0048] S11. After extracting the original height map from the original image sequence and denoising and correcting it, the optimal exposure range is determined based on the multi-exposure response and physical noise. The signal-to-noise ratio gain factor and local intensity asymmetry are calculated, and the two are jointly predicted and optimized through a dual-branch convolutional neural network to achieve accurate quantification of laser stripe quality and stray light interference.

[0049] Based on the multi-exposure raw images acquired by a line laser 3D profilometer, cross-sectional analysis is performed on each scan line. The centroid method is used to accurately extract the center position of the laser stripes. According to the calibration parameters, the pixel displacement of the stripe center is converted into the actual height value using geometric trigonometric relationships, generating a two-dimensional raw height map. , representing the initial three-dimensional shape of the surface of the object being measured.

[0050] The original multi-exposure image sequence was denoised using a bilateral filtering method. Based on the calibration parameters, the image was geometrically corrected using the camera intrinsic matrix and distortion coefficients to eliminate lens distortion. Brightness consistency was corrected using the radiometric response function to ensure that the laser stripe structure was clear and the grayscale value was accurate.

[0051] Deep convolutional neural networks are used to extract features from preprocessed image sequences, accurately extracting the signal-to-noise ratio (SNR) gain factor and local intensity asymmetry of each pixel within the optimal exposure range. The SNR gain factor reflects the intensity enhancement of the signal relative to background noise, while the local intensity asymmetry measures the asymmetry of brightness distribution in the area surrounding the pixel and is used to identify potential stray light. The specific process is as follows: a dual-branch fully convolutional neural network is designed, with the corrected image sequence as input and two output channels corresponding to the SNR gain factor and local intensity asymmetry, respectively. The network structure includes: an encoder that downsamples layer by layer to extract multi-scale features and fuses information from different exposure levels; and a decoder that upsamples layer by layer to restore spatial resolution, fuses high-dimensional features from the encoder, and preserves edges and fine structures. The loss function is the sum of three terms: the SNR gain factor prediction error, the local intensity asymmetry prediction error, and the L1 regularization term of the asymmetry gradient.

[0052] Based on this, the network is divided into two independent output branches, one for each pixel position of the input image sequence. Extract its values ​​at K different exposure times The grayscale values ​​below are used to form a multi-exposure response curve. The exposure response curve is linearized using the calibrated camera response function to restore the true irradiance estimate. Based on this, the optimal exposure range for each pixel is determined by comprehensively considering factors such as whether the grayscale value is within the ideal dynamic range, whether the gradient is significant, and whether the changes between adjacent frames are stable. The exposure level that maximizes the signal-to-noise ratio is selected as the optimal imaging reference. The specific formula is as follows: , Index the selected optimal exposure level, which is the frame that provides the highest signal-to-noise ratio for this pixel (x,y) among all eligible exposure frames. For pixels The exposure time corresponding to the kth exposure level The grayscale response value below, Let represent photon shot noise, q represent the probability that each incident photon generates a detectable photogenerated electron under illumination, and E represent the incident irradiance. To read out noise, it is obtained from sensor calibration. To measure the maximum quality gain that a pixel can achieve throughout the entire exposure sequence, a signal-to-noise ratio (SNR) gain factor is defined, expressed on a logarithmic scale as the factor by which the SNR is improved under optimal conditions relative to the worst available imaging conditions.

[0053] Another branch takes a local neighborhood window centered on each pixel and analyzes the spatial symmetry of its grayscale distribution along the normal direction of the laser stripe. It calculates the mirror brightness difference of the corresponding position in the region about the central axis and sums them by weight to obtain the local symmetry error. Gaussian weight is used to highlight the influence of the central region. Then, the difference between the local peak intensity and the background intensity is normalized to eliminate the interference of low contrast area on the asymmetry evaluation and obtain the local intensity asymmetry. This is used to quantify the degree to which the light intensity distribution around the pixel deviates from the ideal symmetry shape, thereby reflecting the risk of stripe distortion caused by stray light.

[0054] The two branches share coding features but independently complete the regression task, thereby achieving refined, pixel-by-pixel joint prediction of the signal-to-noise ratio gain factor and local intensity asymmetry.

[0055] S12. Combining physical optics simulation and experimental calibration, the imaging optical path of the line laser 3D profilometer is accurately mathematically characterized to construct a point spread function, and this function is used to simulate the light energy diffusion effect caused by stray light.

[0056] Based on the actual imaging optical path structure of a line laser 3D profilometer, the imaging optical path is precisely constructed to determine its point spread function, elevating the influence of stray light from empirical observation to a predictable physical process. Specifically, ray tracing is used to simulate the light energy diffusion distribution on the image plane after an ideal point light source passes through the lens, sensor, and internal structure. The process comprehensively considers actual parameters such as the lens's numerical aperture, focal length, aberration characteristics, optical path obstruction, lens reflectivity, and the scattering characteristics of non-ideal surfaces inside the cavity. To improve accuracy, a tiny pinhole target is used as an approximate point light source in conjunction with experimental calibration methods. Actual imaging results are collected at different field-of-view positions under conditions without main beam interference, recording the two-dimensional distribution of light energy diffusion. The measured images are registered and fused with the simulation results. Least squares fitting is used to inversely correct unknown parameters such as scattering coefficient and surface roughness in the simulation, obtaining a high-fidelity point spread function distribution.

[0057] The point spread function is used as a convolution kernel and applied layer by layer to a multi-exposure image sequence to simulate the effects of stray light on nonlocal halos, speckles, and background lifting during the imaging process. By analyzing the diffusion range, intensity overflow, and neighborhood interference patterns of pixel gray values ​​after convolution, the potential degree of stray light influence on different regions is quantified.

[0058] S13. Based on the diffusion function of the combination point of signal-to-noise ratio gain factor and local intensity asymmetry, construct a stray light sensitivity weight map.

[0059] A stray light sensitivity weight map is defined to quantify the risk of each pixel being affected by stray light. The specific formula is as follows: ,in, This is the signal-to-noise ratio gain factor; the larger the value, the stronger the original signal and the better the noise immunity. This represents the local intensity asymmetry; a larger value indicates greater susceptibility to lateral stray light intrusion. Indicates the original image The image obtained after applying the point spread function is used to measure the degree of influence of stray light on the pixel. The influence weights of the spread function term at the control points. A preset lower threshold is used, and the maximum value is taken to ensure that the denominator is not too small, preventing abnormally high values ​​from appearing in the weight graph. This yields... To create a two-dimensional weighted map of the same size as the original image, the position of each pixel in the image... The higher the calculated value, the more susceptible it is to stray light interference.

[0060] S20. Adaptive threshold filtering is performed on each layer of the image after decomposition of the Gaussian Laplacian pyramid using a weighted graph, and the stability index of the residual features after filtering of each layer is calculated. The stability index is used as the energy attenuation coefficient to dynamically adjust the filtering intensity of the next layer, generating a clean image while optimizing the edge retention rate.

[0061] Specifically, generating a purification intensity image includes the following sub-steps:

[0062] S21. Perform Gaussian Laplacian pyramid decomposition on the original intensity image to obtain a high-frequency layer sequence containing multi-scale detail information and a low-frequency basal layer that preserves the overall structure.

[0063] Acquire raw intensity images using a line laser 3D profilometer. The image is decomposed into a multi-level Laplacian pyramid of Gaussian origin, representing it in layers according to different spatial scales. This achieves the separation of high-frequency details from low-frequency structures, facilitating accurate identification and adaptive filtering of local anomalies caused by stray light at multiple resolutions. The decomposition layer number is set to L layers. As layer 0, Gaussian low-pass filtering and downsampling are applied layer by layer to obtain a series of smoothed images with the resolution halved layer by layer. The Lth layer image is the top low-frequency basal layer, preserving the overall structure and background information of the image. Subsequently, a Laplacian pyramid is constructed based on the Gaussian pyramid: from layer 1 to layer L, the Laplacian image of each layer is obtained by calculating the difference between the current layer's Gaussian image and the previous layer's Gaussian image after upsampling to the same size. This difference extracts high-frequency detail information at the corresponding scale, including edges, textures, and local abrupt changes, thus forming a set of multi-scale detail layers. Finally, the original image is decomposed into a multi-resolution representation system composed of high-frequency detail layers and low-frequency basal layers.

[0064] S22. Based on the stray light sensitivity weight map, an adaptive threshold filter is designed for each layer of Laplacian detail image to achieve effective suppression of stray light and edge protection.

[0065] First, initialize the parameters, setting an empirical threshold for each layer. This threshold is determined based on statistical methods and represents an estimate of the noise level at that layer. An adjustment coefficient is then determined. This is used to control the influence of the stray light sensitivity weight map on the threshold. Next, an adaptive threshold is calculated. For each pixel location (x, y), the threshold is adjusted based on its value in the weight map. For the ... Laplacian image of layers Calculate the adaptive threshold as .

[0066] For each layer For each pixel (x, y), use the indicator function I to determine whether its absolute value exceeds the adaptive threshold: The first one after adaptive threshold filtering Layered Laplacian image, if If the indicator function value is 1, the original value of the pixel is retained; otherwise, if the indicator function value is 0, the pixel value is 0. The adaptive threshold design allows for filtering with a higher threshold in areas with high signal-to-noise ratio and low stray light risk, effectively removing noise without destroying true edges and details. In areas with high stray light risk, the threshold is lowered to allow more potentially useful signals to pass through.

[0067] S23. Calculate the contrast and information entropy of the residual images of each layer after adaptive filtering and fuse them into an energy attenuation coefficient. Based on this coefficient, dynamically adjust the filtering intensity of the next layer and reconstruct a clean intensity map that suppresses noise and preserves edges.

[0068] After completing the adaptive threshold filtering, the filtering effect of each layer is analyzed to provide feedback for the processing of the next layer. First, the Laplacian detail image of each layer is calculated. The residual image after filtering: It is the retained portion after adaptive thresholding, the residual. This indicates the high-frequency components that have been filtered out, including noise, abnormal textures caused by stray light, and weak signals that have been mistakenly removed.

[0069] To quantify the degree of local structural disorder reflected by these residuals, the gray-level co-occurrence matrix of each residual image is calculated, and two key texture features are extracted from it: contrast. Information entropy reflects the drastic degree of local brightness changes in an image. This measures the irregularity and complexity of the grayscale distribution in an image. The geometric mean of contrast and information entropy is defined as the first... The energy attenuation coefficient of a layer reflects the degree of structural disorder in the energy removed during the filtering process of the current layer. The larger the value, the more complex non-random, non-noise-like textures are present in the residual of that layer, indicating incompletely suppressed stray light interference or loss of effective details, which requires parameter adjustment in subsequent layers to compensate for.

[0070] Energy attenuation coefficient Used to dynamically adjust the filtering strategy of the next layer, the first The layer experience threshold is set to the same as the first layer. The energy attenuation coefficient of the layer is positively correlated, that is... The preset baseline experience threshold, This is an adjustable gain coefficient used to control the feedback strength. The larger the value, the stronger the stray light interference, and the stronger the filtering power of the next layer becomes, achieving layer-by-layer adaptive optimization to ensure that real edges and stray light artifacts can be effectively distinguished at different scales.

[0071] All adaptively filtered detail layers are progressively upsampled and added to the topmost low-frequency basal layer to complete the Laplacian pyramid reconstruction process, resulting in a cleaned intensity image. This image effectively suppresses stray light noise while preserving the true edges and geometric structure of the laser stripes to the greatest extent possible.

[0072] S30. Using the purification intensity image and the original height map as input, one-dimensional profile analysis is used to calculate the sharpness gradient and background flatness to obtain the morphological distortion index sequence. The sequence is then subjected to three-level wavelet packet decomposition to extract the energy concentration. The energy concentration is used as a criterion to determine whether to start the line-by-line stripe repair program based on morphological opening and closing operations.

[0073] Specifically, step S30 includes the following sub-steps:

[0074] S31. Based on the purification intensity image, the main peak of the laser stripe is located by Gaussian fitting combined with the original height map. The peak sharpness gradient variance and background flatness standard deviation in the neighborhood are calculated with the main peak as the center to construct the morphological distortion index sequence.

[0075] One-dimensional profile analysis is performed along each row of pixels in the purified intensity image. Gaussian fitting is used to extract the location of the maximum light intensity on each scan line as the main peak location. Since multiple local maxima caused by stray light make it difficult to determine the main peak, the height value corresponding to the current pixel provided by the original height map is used to assist in the determination. The maximum value at a reasonable height is the main peak location. A local neighborhood window is selected centered on the main peak, and two indicators are calculated within this neighborhood: peak sharpness gradient variance, obtained by taking the first derivative (gradient) of the stripe profile to obtain the slope distribution of light intensity changes, and then calculating the variance of the absolute value of the gradient to measure the steepness of the stripe edges. A small variance value indicates that the stripes become flatter and wider, due to stray light diffusion; and background flatness standard deviation, calculated by statistically analyzing the standard deviation of background gray values ​​in the regions far from the peak on both sides of the main peak, reflecting the stability and noise level of the local background. An increased standard deviation indicates spillover light from nearby bright areas or noise influence.

[0076] The morphological distortion index sequence is obtained by calculating the peak sharpness gradient variance point by point along each scan line, divided by the sum of the background flatness standard deviation and a minimal constant, where the minimal constant is used to prevent the denominator from being zero. Therefore, the morphological distortion index can sensitively characterize the morphological integrity of laser stripes at each spatial location: low-value areas correspond to stripe breakage, severe broadening, or being submerged by stray light, and can serve as the trigger for subsequent repair mechanisms.

[0077] S32. The high-frequency detail coefficients of the third layer of the morphological distortion index sequence are extracted by three-level wavelet packet decomposition, and the energy concentration is calculated. Severe distortion is accurately identified at the sequence level, providing a reliable criterion for subsequent restoration.

[0078] The morphological distortion index sequence is used to identify anomalous regions such as stripe broadening, blurring, or background interference. However, relying solely on point-by-point thresholding is susceptible to local noise or minor fluctuations, making it difficult to distinguish between true structural distortions and gradual changes. Therefore, a three-level wavelet packet decomposition is performed on the morphological distortion index sequence to extract more discriminative indicators. In the first level, the sequence is decomposed into low-frequency and high-frequency components using low-pass and high-pass filters, respectively. The former represents the overall trend of the sequence, while the latter captures local abrupt changes. In the second level, the low-frequency and high-frequency components obtained in the first level are further decomposed to obtain four finer frequency bands. In the third level, each of the four sub-bands in the second level is further decomposed to obtain eight frequency bands. Detail coefficients are extracted from the high-frequency path in the deepest layer of the third level. This coefficient reflects abrupt changes, edge or local drastic fluctuations in the morphological distortion index sequence, which precisely correspond to the locations where laser stripes break, jump, or become severely broadened.

[0079] To quantify whether these high-frequency components are concentrated in certain specific regions, their energy concentration is calculated: This is the sum of the absolute values ​​of all third-level detail coefficients. Let N be the sum of squares of all third-level detail coefficients, and N be the number of detail coefficients. When the energy of the detail coefficients is concentrated in a few positions, i.e., there is a significant abrupt change, the concentration increases significantly; conversely, if the energy distribution is uniform and there is no obvious distortion, the concentration is close to 1.

[0080] Set an experience threshold ,like This indicates the presence of highly concentrated high-frequency abrupt changes in the scan line, suggesting severe structural abnormalities such as stripe breakage or drastic broadening. The line is then marked for further analysis. Otherwise, the stripe continuity is considered good overall, requiring no further processing, and the repair procedure is skipped.

[0081] S33. Based on the energy concentration result, an energy focusing criterion is introduced to comprehensively evaluate the degree of stripe breakage by combining the filter difference and curvature enhancement term. When the energy focusing criterion is greater than the set threshold, the geometric prior is combined to initiate structure-guided interpolation repair to restore the continuity of the image stripes.

[0082] To accurately identify distortion locations and prevent misjudgment, an energy focusing criterion based on multi-scale morphological response is proposed. This criterion analyzes the overall continuity and local irregularities of the stripes from a broader perspective and determines whether repair should be initiated. The specific formula is as follows: ,in, The integration area is the entire row of valid pixels. The value of the shape distortion index at position x The result of performing the opening operation first, followed by the closing operation. To The results of performing the closing operation first and then the opening operation show different responses to discontinuous regions, with significant differences at the fracture point. The absolute difference between the two highlights the true structural gap. To integrate the original morphological distortion exponent over the entire scan line, use it as the denominator. It becomes a relative metric to avoid misjudgment due to changes in overall brightness. for The variance of the second derivative measures the severity of local curvature fluctuations. for The L1 norm of the gradient measures the total edge intensity of the entire line signal and is used for normalization to prevent over-response in textured regions. This is the adjustment coefficient.

[0083] like If the value exceeds a preset threshold, it indicates significant discontinuity distortion in the current scan line. A line-by-line stripe repair procedure based on morphological opening and closing operations is then initiated to significantly enhance the continuity and structural integrity of the stripes. Specifically, based on the cleansing intensity image and the morphological distortion index, a fracture localization mask is generated using the low-value regions in the morphological distortion index and the energy concentration regions detected in step S32. The cleansing intensity image is selectively smoothed and bridged along the scanning direction using multi-scale morphological opening and closing and closing-opening filters. A weight map is constructed using the difference between the two to highlight the fracture location. Guided by the mask, and based on the gradient direction and width characteristics of the stripes on both sides, as well as the geometric constraints of the original height map, linear interpolation guided by the structural tensor is used to adaptively reconstruct the fracture region. Finally, the repaired pixels are updated to the global image, effectively restoring the continuity and structural integrity of the stripes.

[0084] S40. Perform centerline extraction on the striped image after the repair procedure, and calculate the motion blur correlation index using the three-dimensional reconstruction algorithm. If the index exceeds the preset dynamic threshold, use the inverse diffusion equation to perform anisotropic smoothing on the current frame height map to eliminate artifacts caused by dynamic stray light.

[0085] Specifically, anisotropic smoothing is performed on the current frame height map, which includes the following sub-steps:

[0086] S41. Subpixel-level peak detection and cubic spline interpolation are performed on the repaired stripes to construct a smooth and continuous centerline trajectory, providing a geometric basis for high-precision 3D reconstruction.

[0087] Subpixel-level centerline extraction is performed on the restored stripe image to obtain high-precision geometric positions of the laser stripes. Specifically, the intensity peak of the restored stripes is detected along each scanning direction to initially locate the pixel-level center point. Several neighboring pixels near the peak are selected for cubic spline interpolation to construct a continuous intensity distribution curve. Subpixel-level precise positioning is achieved by solving for the extreme points of this curve. All precisely located center points are connected in scanning order to form a spatially continuous and smooth two-dimensional trajectory. , where s is the arc length parameter along the trajectory, and a represents the current scan line.

[0088] S42. Based on the centerline fringe trajectory, the cumulative sum of curvature changes and normal offset between adjacent frames are calculated through three-dimensional reconstruction to construct a motion blur correlation index and quantify the artifact intensity caused by dynamic stray light.

[0089] To evaluate the impact of dynamic stray light on the reconstruction results, a 3D reconstruction algorithm was used to perform geometric consistency analysis on the fringe centerlines at the same spatial location between adjacent frames. First, based on the sub-pixel centerline trajectory... and Calculate the curvature of each point. and And find the absolute values ​​of the changes in curvature of the two. Then, by summing the values ​​along the entire trajectory, we obtain the absolute cumulative sum of the curvature changes. This value reflects the degree of non-rigid distortion of the stripe shape between consecutive frames and is sensitive to local distortion caused by dynamic stray light. Simultaneously, the normal offset distance between corresponding points on the center lines of adjacent frames is calculated, and its root mean square value is obtained. This is used to characterize the stability of the overall fringe displacement. Based on this, a motion fuzzy correlation index is constructed as follows: , To prevent constants with a denominator of zero, this index effectively amplifies the pseudo-effects caused by dynamic lighting changes by normalizing the shape distortion intensity to the overall displacement level. If the value exceeds the preset dynamic threshold, it indicates the presence of non-physical local jitter or blurring caused by dynamic stray light.

[0090] S43. Based on the motion fuzzy correlation index, the spatiotemporal anisotropic inverse diffusion equation is used to effectively filter out height field noise caused by dynamic stray light while preserving geometric edges.

[0091] Based on the motion blur correlation index, to suppress the impact of such artifacts on the 3D reconstruction results, a spatiotemporal anisotropic inverse diffusion smoothing equation is proposed. This equation adaptively filters out height field noise caused by dynamic stray light while preserving the true geometric edges. The specific process is as follows: The height map reconstructed from the current frame... The following partial differential equation is introduced for spatiotemporal co-optimization: ,in, height map The rate of evolution with virtual time t, This is a two-dimensional height map of the current frame, calculated from the stripe center line using a three-dimensional reconstruction algorithm. This is a spatially anisotropic diffusion term, responsible for smoothing noise and preserving edges within the spatial domain. For motion-based fuzzy correlation The diffusion rate control function, For adaptive diffusion rate control coefficients, It is a divergence operator used to construct nonlinear diffusion processes and achieve local adaptive smoothing. Let H be the gradient vector field of the height map. For anisotropic diffusion-type edge preservation functions, strong spatial smoothing is allowed in flat regions to effectively smooth noise, while smoothing is suppressed in edge regions to protect the boundary. This is a time consistency constraint used to ensure a smooth transition of the height field between adjacent frames. Adjust the strength of the time constraint. For time consistency, it pushes the current frame heightmap H to the previous frame. By bringing the frames closer together, we can prevent abrupt changes caused by single-frame artifacts and improve sequence stability. This is the dynamic adjustment coefficient.

[0092] S50. The processed multi-frame height map is registered based on the iterative nearest point method. The mean cosine of the normal vector cosine is extracted and combined with the global height fluctuation coefficient to generate geometric fidelity. The geometric fidelity is used as a feedback signal to drive the genetic algorithm to optimize the initial weights and restart the entire process until the geometric fidelity converges to its theoretical peak value and meets the preset accuracy requirements.

[0093] Specifically, step S50 includes the following sub-steps:

[0094] S51. The iterative nearest point method based on the key point matching optimization of three-dimensional scale-invariant feature transformation is adopted to perform high-precision spatiotemporal registration on the denoised multi-frame height map, effectively overcoming motion blur and stray light interference, and forming a registered point cloud sequence.

[0095] To achieve high-precision 3D dynamic reconstruction, spatiotemporal registration is performed on multi-frame height maps after pre-processing denoising and restoration. Since single-frame height fields contain minor deformations, motion blur, or local missing data, direct stitching can lead to accumulated errors and structural misalignment. Therefore, an iterative nearest-point method based on feature matching optimization is used for fine registration. First, a corresponding 3D point cloud is generated for each frame height map, and 3D scale-invariant feature transformation key points are extracted from its surface. These key points are concentrated in regions with significant curvature and rich geometric features, effectively resisting local intensity perturbations caused by stray light. Then, the 3D scale-invariant feature transformation descriptor for each key point is calculated. Corresponding point sets between adjacent frames are established through descriptor matching, and the initial rigid body transformation matrix, including rotation and translation, is solved based on these corresponding point sets.

[0096] Based on this, the iterative nearest neighbor method is executed: First, for each point in the source frame with the current pose, the nearest neighbor point is searched in the target frame to establish a correspondence; then, a weighted registration error term is constructed, and a point-to-surface distance minimization strategy is adopted to improve convergence accuracy. High-confidence feature matching points provided by the 3D scale-invariant feature transformation are given higher weights to enhance the guiding role of reliable responses in overall transformation estimation; simultaneously, geometric consistency checks such as distance thresholds and normal angles are combined to dynamically eliminate mismatched points to suppress abnormal disturbances; next, the optimal rigid body transformation is solved based on the weighted error function, and the pose of the current frame is updated; finally, it is determined whether the transformation increment is less than a preset threshold or the maximum number of iterations has been reached. If so, the iteration exits; otherwise, it returns to continue optimization until high-precision registration is achieved. The entire process is advanced frame by frame in the time series, ensuring that the height maps of each frame are accurately aligned in a unified coordinate system, forming a continuous, stable, and detailed point cloud sequence.

[0097] S52. Perform surface reconstruction on the point cloud sequence to generate a three-dimensional model, and calculate the geometric fidelity index by the mean of the normal vector consistency and the coefficient of variation of height fluctuation of the static region point cloud to evaluate the overall reconstruction quality.

[0098] After registration, the registered point cloud sequence is reconstructed using Poisson surfaces, converting the discrete point cloud into a continuous, closed triangular mesh model while preserving the original geometric details. Geometric fidelity indices are calculated to evaluate the geometric reliability and stability of the overall reconstruction results. Static regions are extracted from the registered temporal point cloud to eliminate interference from dynamic object displacement. Within these static regions, two key geometric features are calculated: first, the mean cosine of the normal vector consistency angle, obtained by averaging the cosine values ​​of the angles between the normals of corresponding points in adjacent frames, reflecting the temporal stability of surface orientation; the closer the value is to 1, the smaller the normal variation and the more consistent the reconstruction; second, the global height fluctuation variation coefficient corrected by piecewise linear transformation. This index is obtained by first fitting a piecewise linear background trend to the height field of each frame to eliminate systematic tilt or distortion, and then calculating the ratio of the standard deviation to the mean of the residuals, used to quantify the degree of non-physical fluctuation in height values. Both are normalized to the [0,1] interval and weighted to obtain the geometric fidelity index. The closer the index value is to 1, the higher the reconstruction quality.

[0099] S53. Using geometric fidelity index as feedback, the genetic algorithm is used to dynamically optimize key parameters and restart the entire process. Through convergence judgment and double verification, the adaptive suppression accuracy and reconstruction robustness of stray light interference are significantly improved.

[0100] To improve the accuracy of stray light suppression in the model, a genetic algorithm is introduced to construct a closed-loop optimization mechanism. The geometric fidelity index serves as the fitness feedback signal, driving global optimization of key parameters. By iteratively adjusting the core parameters for denoising and restoration effects and restarting the entire process for verification, the geometric fidelity index gradually converges to its peak value, ensuring an optimal balance between geometric consistency and detail fidelity in the reconstruction results. This process significantly enhances the adaptive capability of stray light suppression, especially in scenarios with drastic lighting changes or strong local interference. It can automatically match the optimal filtering strategy, effectively avoiding artifacts caused by undersmoothing or loss of detail due to oversmoothing. Furthermore, a dual verification mechanism of outlier detection and physical simulation is used. Outlier detection identifies geometric anomalies such as local distortions, oversmoothed regions, or false edges in the reconstructed point cloud, eliminating parameter combinations that, while having high indices, result in distorted details. Subsequently, physical simulation verification compares the reconstructed temporal deformation trend with a preset simple harmonic motion model to check for false fluctuations that do not conform to actual motion behavior. For static scenes, the displacement of each point is required to be close to zero; for periodic vibration scenes, the reconstructed trajectory is required to be highly consistent with a sinusoidal variation law. If the deviation exceeds the set threshold, the parameter combination is deemed invalid. Both verifications ensure that the optimized parameters suppress stray light interference while maintaining geometric rationality and physical realism, thereby achieving high-precision and robust dynamic 3D reconstruction.

[0101] Example 2

[0102] like Figure 2 As shown, Embodiment 2 of this application provides a line laser 3D profilometer system integrating multiple stray light suppression algorithms, including:

[0103] Sensitivity Weight Map Module: Based on multi-exposure raw image sequences acquired by a line laser 3D profilometer, this module uses a deep convolutional neural network to predict the signal-to-noise ratio gain factor and local intensity asymmetry of each pixel within the optimal exposure range, and combines this with a point spread function to construct a stray light sensitivity weight map. Specifically, it is divided into the following sub-modules:

[0104] The dual-branch submodule extracts the original height map from the original image sequence and performs denoising correction. Based on the multi-exposure response and physical noise, it determines the optimal exposure range, calculates the signal-to-noise ratio gain factor and local intensity asymmetry, and jointly predicts and optimizes the two through a dual-branch convolutional neural network to achieve accurate quantification of laser stripe quality and stray light interference.

[0105] Point spread function submodule: Combining physical optics simulation and experimental calibration, the imaging optical path of the line laser 3D profilometer is accurately mathematically characterized to construct the point spread function, and this function is used to simulate the light energy diffusion effect caused by stray light.

[0106] Sub-module construction: Based on the diffusion function of the combination of signal-to-noise ratio gain factor and local intensity asymmetry, a stray light sensitivity weight map is constructed.

[0107] The purification intensity map module: This module uses a weighted map to perform adaptive threshold filtering on each layer of the image after decomposition of the Gaussian Laplacian pyramid, and calculates the stability index of the residual features after filtering for each layer. This stability index is used as an energy attenuation coefficient to dynamically adjust the filtering intensity of the next layer, generating a purification intensity image while optimizing edge retention. Specifically, it is divided into the following sub-modules:

[0108] Decomposition submodule: The original intensity image is decomposed into a Gaussian Laplacian pyramid to obtain a high-frequency layer sequence containing multi-scale detail information and a low-frequency basal layer that preserves the overall structure.

[0109] The filtering submodule designs an adaptive threshold filter for each layer of Laplacian detail image based on the stray light sensitivity weight map to achieve effective suppression of stray light and edge protection.

[0110] Reconstruction submodule: Calculates the contrast and information entropy of the residual images of each layer after adaptive filtering and fuses them into an energy attenuation coefficient. Based on this coefficient, it dynamically adjusts the filtering intensity of the next layer and reconstructs a clean intensity map that suppresses noise and preserves edges.

[0111] Stripe Restoration Module: Taking the purified intensity image and the original height map as input, it uses one-dimensional profile analysis to calculate the sharpness gradient and background flatness to obtain a morphological distortion index sequence. This sequence is then subjected to three-level wavelet packet decomposition to extract energy concentration. The energy concentration is used as a criterion to determine whether to initiate a line-by-line stripe restoration procedure based on morphological opening and closing operations. Specifically, it is divided into the following sub-modules:

[0112] Morphology Distortion Submodule: Based on the purification intensity image, the main peak of the laser stripe is located by Gaussian fitting combined with the original height map. The peak sharpness gradient variance and background flatness standard deviation in the neighborhood are calculated with the main peak as the center to construct the morphology distortion index sequence.

[0113] Energy Concentration Submodule: Extracts the third-level high-frequency detail coefficients of the morphological distortion index sequence through three-level wavelet packet decomposition and calculates the energy concentration, accurately identifying severe distortions at the sequence level and providing reliable criteria for subsequent restoration.

[0114] Interpolation Repair Submodule: Based on the energy concentration result, an energy focusing criterion is introduced to comprehensively evaluate the degree of stripe breakage by combining the filter difference and curvature enhancement term. When the energy focusing criterion is greater than the set threshold, the geometric prior is combined to initiate structure-guided interpolation repair to restore the continuity of the image stripes.

[0115] Anisotropic smoothing module: This module extracts the centerline of the striped image after the repair process and calculates the motion blur correlation index using a 3D reconstruction algorithm. If the index exceeds a preset dynamic threshold, anisotropic smoothing is performed on the current frame height map using the inverse diffusion equation to eliminate artifacts caused by dynamic stray light. Specifically, it is divided into the following sub-modules:

[0116] Centerline submodule: Performs subpixel-level peak detection and cubic spline interpolation on the repaired stripes to construct a smooth and continuous centerline trajectory, providing a geometric basis for high-precision 3D reconstruction.

[0117] Motion correlation submodule: Based on the centerline stripe trajectory, it calculates the cumulative sum of curvature changes and normal offset between adjacent frames through 3D reconstruction, constructs a motion blur correlation index, and quantifies the intensity of artifacts caused by dynamic stray light.

[0118] Anisotropic submodule: Based on the motion fuzzy correlation index, it uses the spatiotemporal anisotropic inverse diffusion equation to effectively filter out height field noise caused by dynamic stray light while preserving geometric edges.

[0119] Registration optimization module: This module registers the processed multi-frame height maps using the iterative nearest-point method, extracts the mean cosine of the normal vector's consistent angle, and combines it with the global height fluctuation coefficient to generate geometrical fidelity. The geometrical fidelity is then used as a feedback signal to drive a genetic algorithm to optimize the initial weights and restart the entire process until the geometrical fidelity converges to its theoretical peak and meets the preset accuracy requirements. Specifically, it is divided into the following sub-modules:

[0120] The registration submodule employs an iterative nearest-point method based on keypoint matching optimization using 3D scale-invariant feature transformation to perform high-precision spatiotemporal registration of denoised multi-frame height maps, effectively overcoming motion blur and stray light interference to form a registration point cloud sequence.

[0121] The geometric fidelity submodule performs surface reconstruction on point cloud sequences to generate 3D models, and calculates the geometric fidelity index by using the mean normal vector consistency and height fluctuation coefficient of static region point clouds to evaluate the overall reconstruction quality.

[0122] Feedback optimization submodule: Using geometric fidelity index as feedback, it dynamically optimizes key parameters using a genetic algorithm and restarts the entire process. Through convergence judgment and double verification, it significantly improves the adaptive suppression accuracy and reconstruction robustness against stray light interference.

[0123] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of the present invention should be included within the scope of protection of the present invention.

Claims

1. A working method for a line laser 3D profilometer integrating multiple stray light suppression algorithms, characterized in that: include: S10. Based on the multi-exposure original image sequence acquired by the line laser 3D profilometer, the signal-to-noise ratio gain factor and local intensity asymmetry of each pixel in the optimal exposure range are predicted by a deep convolutional neural network, and a stray light sensitivity weight map is constructed by combining the point spread function. S20. Adaptive threshold filtering is performed on each layer of the image after decomposition of the Gaussian Laplacian pyramid using the weight map, and the stability index of the residual features after filtering of each layer is calculated. The stability index is used as the energy attenuation coefficient to dynamically adjust the filtering intensity of the next layer, generating a clean image while optimizing the edge retention rate. S30. Using the purification intensity image and the original height map as input, the sharpness gradient and background flatness are calculated by one-dimensional profile analysis to obtain the morphological distortion index sequence. The sequence is then subjected to three-level wavelet packet decomposition to extract the energy concentration. The energy concentration is used as a criterion to determine whether to start the line-by-line stripe repair program based on morphological opening and closing operations. S40. Perform centerline extraction on the striped image after the repair procedure, calculate the motion blur correlation index using the three-dimensional reconstruction algorithm. If the index exceeds the preset dynamic threshold, use the inverse diffusion equation to perform anisotropic smoothing on the current frame height map to eliminate artifacts caused by dynamic stray light. S50. The anisotropically smoothed multi-frame height map is registered based on the iterative nearest point method. The mean cosine of the normal vector cosine is extracted and combined with the global height fluctuation coefficient to generate geometric fidelity. The geometric fidelity is used as a feedback signal to drive the genetic algorithm to optimize key parameters and return to S10 to restart the entire process until the geometric fidelity converges to its theoretical peak value and meets the preset accuracy requirements.

2. The working method of the line laser 3D profilometer integrating multiple stray light suppression algorithms as described in claim 1, characterized in that, Based on the multi-exposure raw image sequence acquired by a line laser 3D profilometer, a deep convolutional neural network is used to predict the signal-to-noise ratio gain factor and local intensity asymmetry of each pixel in the optimal exposure range. A stray light sensitivity weight map is then constructed using a point spread function. The process is divided into the following sub-steps: After extracting the original height map from the original image sequence and denoising and correcting it, the optimal exposure range is determined based on the multi-exposure response and physical noise. The signal-to-noise ratio gain factor and local intensity asymmetry are calculated, and the two are jointly predicted and optimized through a dual-branch convolutional neural network to achieve accurate quantification of laser stripe quality and stray light interference. By combining physical optics simulation and experimental calibration, the imaging optical path of the line laser 3D profilometer is accurately mathematically characterized to construct a point spread function, and this function is used to simulate the light energy diffusion effect caused by stray light. A stray light sensitivity weighting map is constructed based on the diffusion function of the combination of signal-to-noise ratio gain factor and local intensity asymmetry.

3. The working method of the line laser 3D profilometer integrating multiple stray light suppression algorithms as described in claim 1, characterized in that, The weighted image is used to perform adaptive threshold filtering on each layer of the image after decomposition of the Gaussian Laplacian pyramid. The stability index of the residual features after filtering of each layer is calculated. The stability index is used as the energy attenuation coefficient to dynamically adjust the filtering intensity of the next layer, generating a clean image while optimizing the edge preservation rate. The process is divided into the following sub-steps: The original intensity image is decomposed into a Gaussian Laplacian pyramid to obtain a high-frequency layer sequence containing multi-scale detail information and a low-frequency basal layer that preserves the overall structure. An adaptive threshold filter is designed for each layer of Laplacian detail image based on the stray light sensitivity weight map to achieve effective suppression of stray light and edge protection; The contrast and information entropy of the residual images of each layer after adaptive filtering are calculated and fused into an energy attenuation coefficient. Based on this coefficient, the filtering intensity of the next layer is dynamically adjusted to reconstruct a clean intensity map that suppresses noise and preserves edges.

4. The working method of the line laser 3D profilometer integrating multiple stray light suppression algorithms as described in claim 1, characterized in that, Using the purification intensity image and the original height map as input, a morphological distortion index sequence is obtained by calculating the sharpness gradient and background flatness using one-dimensional profile analysis. This sequence is then subjected to three-level wavelet packet decomposition to extract the energy concentration. The energy concentration is used as a criterion to determine whether to initiate a line-by-line stripe restoration procedure based on morphological opening and closing operations. The specific steps are as follows: Based on the purification intensity image, the main peak of the laser stripe is located by Gaussian fitting combined with the original height map. The peak sharpness gradient variance and background flatness standard deviation in the neighborhood are calculated with the main peak as the center to construct the morphological distortion index sequence. The high-frequency detail coefficients of the third layer of the morphological distortion index sequence are extracted by three-level wavelet packet decomposition, and the energy concentration is calculated. Severe distortion is accurately identified at the sequence level, providing a reliable criterion for subsequent restoration. Based on the energy concentration results, an energy focusing criterion is introduced to comprehensively evaluate the degree of stripe breakage by combining the filter difference and curvature enhancement term. When the energy focusing criterion is greater than the set threshold, geometric prior is combined to initiate structure-guided interpolation repair to restore the continuity of image stripes.

5. The working method of the line laser 3D profilometer integrating multiple stray light suppression algorithms as described in claim 4, characterized in that, By combining geometric prior-driven structure-guided interpolation incision to restore the continuity of image stripes, the process is divided into the following sub-steps: Based on the purification intensity image and morphological distortion index, a fracture localization mask is generated by using the low-value region in the morphological distortion index and the detected energy concentration region. The image fracture region is selectively smoothed and bridged by multi-scale morphological opening and closing and closing-opening filtering along the scanning direction of the purification intensity image. A weight map is constructed using the difference between the opening-closing filtering result and the closing-opening filtering result to highlight the fracture location. Guided by the mask, based on the gradient direction and width characteristics of the stripes on both sides and the geometric constraints of the original height map, linear interpolation guided by the structural tensor is used to adaptively reconstruct the fractured area. The repaired pixels are then updated to the global image, achieving effective restoration of stripe continuity and structural integrity.

6. The working method of the line laser 3D profilometer integrating multiple stray light suppression algorithms as described in claim 1, characterized in that, The centerline of the striped image after the repair procedure is extracted, and the motion blur correlation index is calculated using a 3D reconstruction algorithm. If the index exceeds a preset dynamic threshold, the height map of the current frame is anisotropically smoothed using the inverse diffusion equation to eliminate artifacts caused by dynamic stray light. The specific steps are as follows: Subpixel-level peak detection and cubic spline interpolation were performed on the repaired stripes to construct a smooth and continuous centerline trajectory, providing a geometric basis for high-precision 3D reconstruction. Based on the centerline trajectory, the cumulative sum of curvature changes and normal offset between adjacent frames are calculated through 3D reconstruction to construct a motion blur correlation index and quantify the artifact intensity caused by dynamic stray light. If the motion blur correlation index exceeds the preset dynamic threshold, the spatiotemporal anisotropic inverse diffusion equation is used to anisotropically smooth the current frame height map.

7. The working method of the line laser 3D profilometer integrating multiple stray light suppression algorithms as described in claim 1, characterized in that, The anisotropically smoothed multi-frame height maps are registered using the iterative nearest point method. The mean cosine of the normal vector consistency angle is extracted and combined with the global height fluctuation coefficient to generate geometric fidelity. The geometric fidelity is then used as a feedback signal to drive the genetic algorithm to optimize key parameters and return to S10 to restart the entire process until the geometric fidelity converges to its theoretical peak and meets the preset accuracy requirements. The specific steps are as follows: The iterative nearest point method based on key point matching optimization of three-dimensional scale-invariant feature transformation is adopted to perform high-precision spatiotemporal registration on the denoised multi-frame height maps, effectively overcoming motion blur and stray light interference, and forming a registered point cloud sequence. A 3D model is generated by surface reconstruction of the point cloud sequence, and the geometric fidelity index is calculated by the mean of normal vector consistency and the coefficient of variation of height fluctuation of the static region point cloud to evaluate the overall reconstruction quality. Using geometric fidelity as feedback, a genetic algorithm is used to dynamically optimize key parameters and restart the entire process until the geometric fidelity converges to its theoretical peak and meets the preset accuracy requirements, significantly improving the adaptive suppression accuracy and reconstruction robustness against stray light interference.

8. A line laser 3D profilometer system integrating multiple stray light suppression algorithms for performing the method as described in any one of claims 1-7, characterized in that, include: Sensitivity weight map module: Based on the multi-exposure raw image sequence acquired by the line laser 3D profilometer, the signal-to-noise ratio gain factor and local intensity asymmetry of each pixel in the optimal exposure range are predicted by a deep convolutional neural network, and a stray light sensitivity weight map is constructed by combining the point spread function. Purification intensity map module: The weight map is used to perform adaptive threshold filtering on each layer of the image after decomposition of the Gaussian Laplacian pyramid, and the stability index of the residual feature after filtering of each layer is calculated. The stability index is used as the energy attenuation coefficient to dynamically adjust the filtering intensity of the next layer, generate a purification intensity image and optimize the edge retention rate at the same time. Stripe Repair Module: Taking the purified intensity image and the original height map as input, it uses one-dimensional profile analysis to calculate the sharpness gradient and background flatness to obtain the morphological distortion index sequence. The sequence is then subjected to three-level wavelet packet decomposition to extract the energy concentration. The energy concentration is used as a criterion to determine whether to start the line-by-line stripe repair program based on morphological opening and closing operations. Anisotropic smoothing module: Performs centerline extraction on the striped image after the repair procedure, calculates motion blur correlation index using 3D reconstruction algorithm, and if the index exceeds the preset dynamic threshold, anisotropic smoothing is performed on the current frame height map using the inverse diffusion equation to eliminate artifacts caused by dynamic stray light. Registration optimization module: Registers the anisotropically smoothed multi-frame height maps based on the iterative nearest point method, extracts the mean cosine of the normal vector consistency angle and combines it with the global height fluctuation coefficient to generate geometric fidelity, and uses the geometric fidelity as a feedback signal to drive the genetic algorithm to optimize key parameters and return to S10 to restart the entire process until the geometric fidelity converges to its theoretical peak value and meets the preset accuracy requirements.

Images