A laser beam contour machining detection method and system based on optical coherence tomography

Abstract

The application provides a laser beam following contour machining detection method and system based on optical coherence tomography, and relates to the technical field of laser processing.The current three-dimensional topography is obtained through optical coherence tomography, and is compared with a preset target topography to obtain a profile depth error pixel.For the error candidate area, the best incident direction is determined based on the error pixel distribution of the candidate area;the aspheric surface parameters of the aspheric surface mirror system of the flat-top light beam are taken as optimization objects to generate an initial population, the reshaped light beams under each individual vector are obtained based on the pre-aspheric surface mirror system, the aspheric surface parameters are optimized by combining the improved crossover operation genetic algorithm, the reshaped light beam under the optimal individual vector is incident into the corresponding candidate area along the best incident direction, and the laser beam following contour machining is completed, so that the profile precision of the special-shaped hole and groove is ensured.

Description

Technical Field

[0001] This invention relates to the field of laser processing technology, specifically to a laser beam conformal processing detection method and system based on optical coherence tomography. Background Technology

[0002] In industries such as aerospace, new energy vehicles, and 3C electronics, there is a widespread demand for precision structural machining of complex holes (such as irregularly shaped holes and variable cross-section holes) and grooves (such as cooling grooves and guide grooves). Laser beam conformal machining is often used for hole and groove machining, local shaping, depth compensation, and profile correction in the areas to be machined. Because the surfaces of these workpieces typically have continuously changing three-dimensional profiles and depth undulations in different areas, the machining process requires not only acquiring the current three-dimensional shape of the area to be machined, but also accurately identifying the depth and profile deviations of the current shape relative to the target shape. This allows for timely adjustment of subsequent machining parameters and improves the consistency between the machining results and the target shape.

[0003] In existing technologies, laser processing of the target area is usually performed using a pre-set incident direction and a fixed spot or a few selectable spot shapes. The processing results are then corrected by combining offline contour measurement, two-dimensional visual inspection, or simple surface height feedback. Although some surface morphology information can be obtained, it is mostly based on the overall height difference, local threshold judgment, or empirical rules. Subsequently, the optical system parameters are adjusted through single-objective optimization or conventional genetic optimization to generate the corresponding shaping beam.

[0004] However, for processing areas with complex boundaries, local deep grooves, stepped transitions, or freeform surface features, existing technologies struggle to accurately characterize the profile depth error distribution between the current 3D topography and the target topography. They also struggle to form connected local candidate regions based on error pixels and to determine the optimal incident direction based on the spatial distribution of the error regions. Furthermore, existing optimization methods often fail to simultaneously consider shape matching, boundary coverage, and energy distribution coordination between the shaping beam and the error regions, easily leading to insufficient or over-processing in certain areas, or uneven boundary trimming, thus affecting the effectiveness of laser beam conformal processing.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a laser beam conformal processing detection method and system based on optical coherence tomography to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A laser beam conformal processing detection method based on optical coherence tomography includes the following steps: S1: Obtain the current contour map of the area to be processed through optical coherence tomography, compare the current contour map with the target contour map to obtain contour depth error pixels, divide the contour depth error pixels into regions, and obtain several connected candidate regions. S2: For each candidate region, the optimal incident direction of the shaping beam for processing the candidate region is determined based on the distribution characteristics of the internal profile depth error pixels. The aspherical parameters of the aspherical mirror system used to generate the shaping beam are used as optimization variables to construct the initial population of each candidate region. S3: Based on the aspherical mirror system, determine the shaping beam corresponding to each individual in the initial population, and construct optimization index by combining the distribution of profile depth error pixels within the error candidate region; S4: Select individuals from the initial population based on the optimization index, and divide the selected individuals into sets of shape-dominant individuals, boundary-dominant individuals, and energy-dominant individuals. S5: Select one individual from each of the shape-dominant individual set, the boundary-dominant individual set, and the energy-dominant individual set to form several sets of cross-parent combinations; S6: Based on the updatable weight coefficients, perform three-parent weighted cross-crossing on each parent in each cross-parent combination to obtain cross-offspring. Adaptively update the weight coefficients in the corresponding cross-parent combination according to the improvement of each cross-offspring in the optimization index; obtain multiple cross-offspring and classify the cross-offspring into the second group. S7: After performing mutation on the crossover offspring, update the initial population. Iterate through steps S4 to S7 until the termination condition is met. When the termination condition is met, determine the optimal individual in the second population and incident the shaping beam corresponding to the optimal individual along the optimal incident direction of the corresponding candidate region to the candidate region for laser beam conformal processing.

[0008] Furthermore, comparing the current contour image with the target contour image includes: The current contour map is a two-dimensional topography map containing depth information; Perform spatial registration between the current contour and the target contour so that they are located in the same coordinate system; Based on the depth difference between the current contour and the target contour at the corresponding positions, determine the contour depth error value of each pixel; The profile depth error value is compared with a preset threshold, and pixels that are greater than the preset threshold are identified as profile depth error pixels.

[0009] Furthermore, the logic for determining the optimal incident angle based on the depth difference is as follows: For each candidate region, the candidate region is divided into two regions by a straight line passing through the center of the candidate region. The sum of the depth differences of all pixels in the two regions is calculated, and the absolute difference of the sum of the depth differences of the two regions is calculated and defined as the segmentation deviation. The line is rotated continuously and the calculation is repeated. After traversing one cycle, the straight line with the smallest segmentation deviation is selected. An incident plane is constructed based on this straight line. The incident plane is a plane that is perpendicular to the current contour map plane and contains the depth bisector. Multiple candidate incident rays are arranged in the plane. The total projection length of all depth difference line segments on each candidate incident ray is counted. The ray with the largest total projection length is selected and its reverse direction is determined as the optimal incident direction.

[0010] Furthermore, the logic for projecting the depth difference along the optimal incident direction to generate a deviation projection map with energy distribution characteristics is as follows: Using the optimal incident direction as the projection axis, the profile depth error is mapped to the optimal incident direction to obtain the error projection value of each error pixel. A blank grayscale image is created based on the current contour map. The error projection values ​​of each projection deviation point are normalized and converted to grayscale to generate a deviation projection map of the standard incident space contour.

[0011] Furthermore, for each aspherical parameter, a random number is generated within its corresponding preset range. All random values ​​generated by the aspherical parameters constitute an individual, and multiple individuals are generated to form the initial population. The calculated deviation projection map and energy distribution map are registered, and the light intensity of the shaped beam is normalized and converted to grayscale to generate the energy distribution map. Texture features of the gray-level co-occurrence matrix of the deviation projection map and the energy distribution map are extracted. Texture feature vectors are constructed from the texture features of the gray-level co-occurrence matrix of the deviation projection map and the energy distribution map. The texture features include homogeneity, contrast and entropy. The cosine similarity of the texture feature vectors of the deviation projection map and the shaping beam is calculated as a shape optimization index. The edge pixels of the deviation projection map and the energy distribution map are extracted based on the Sobel operator and the corresponding edge pixel set is formed. The intersection-union ratio of the edge pixel sets of the deviation projection map and the energy distribution map is used as the boundary optimization index. The Pearson correlation coefficient between the deviation projection map and the energy distribution map is calculated and used as an energy optimization index.

[0012] Furthermore, the logic for dividing the selected individuals into sets of shape-advantageous individuals, boundary-advantageous individuals, and energy-advantageous individuals is as follows: selection refers to the selection operation in a multi-objective genetic algorithm; The selected bodies are sorted in descending order of shape optimization index, boundary optimization index, and energy optimization index, respectively, to obtain the corresponding shape sorting result, boundary sorting result, and energy sorting result. For each individual, compare its ranking in the shape ranking, boundary ranking, and energy ranking results from largest to smallest. If its ranking in the shape ranking result is the lowest, define the individual as a shape-dominant individual and form a shape-dominant individual set. If its ranking in the boundary ranking result is the lowest, define the individual as a boundary-dominant individual and form a boundary-dominant individual set. If its ranking in the energy ranking result is the lowest, define the individual as an energy-dominant individual and form a boundary-dominant individual set.

[0013] Furthermore, based on the improvement of each crossover offspring in terms of optimization metrics, the logic for adaptively updating the weight coefficients in the corresponding crossover parent combination is as follows: Based on the updatable weight coefficients, a three-parent weighted cross is performed on each parent in each cross parent combination to obtain cross offspring. The weight coefficients in the corresponding cross parent combination are adaptively updated according to the improvement of each cross offspring in the optimization index. Multiple cross offspring are obtained and the cross offspring are assigned to the second group. Each time, a crossover parent combination is selected for a three-parent weighted crossover. In the first weighted crossover, all weight coefficients are initialized to be equal, and the sum of the weight coefficients is 1. Weighted crossover is performed according to the initialized weight coefficients to obtain crossover offspring. The optimization index of the crossover parents is weighted using the initialized weight coefficients to obtain a comprehensive optimization index. For each optimization index, the comprehensive optimization index is subtracted from the optimization index of the offspring, and the ratio of the result to the maximum value of the corresponding optimization index in the initial population is calculated to obtain the improvement amount of the optimization index. The improvement amount of the optimization index is compared with 0. If it is greater than 0, 0 is set as the penalty term for the optimization index. If it is not greater than 0, a penalty factor for the optimization index is preset, and the penalty term is set as the product of the optimization index and the penalty factor. The improvement amount of the optimization index is added to the penalty terms of the other optimization indices except for the optimization index to serve as the feedback term for the optimization index.

[0014] For each three-parent weighted cross except the first three-parent weighted cross, the average of all feedback terms before that three-parent weighted cross is obtained, which is called the average feedback term of that three-parent weighted cross. The feedback term of each optimization index in the previous three-parent weighted cross is obtained, and the average of the feedback term of the previous three-parent weighted cross and the average feedback term of that three-parent weighted cross is calculated. This average is then substituted into the Sigmoid function, and the result is multiplied by the weight coefficient of the optimization index in the previous three-parent weighted cross, which is used as the feedback weight coefficient of that three-parent weighted cross. The feedback weight coefficients of all optimization indices in that three-parent weighted cross are normalized to obtain the weight coefficient of that three-parent weighted cross.

[0015] Furthermore, the logic for determining the optimal individual is as follows: among the offspring population that meets the termination condition, the offspring population is sorted non-dominated based on the optimization index to obtain multiple non-dominated layers. In the first non-dominated layer, the individual with the largest energy optimization index is selected as the optimal individual.

[0016] This invention further provides a laser beam conformal machining detection system based on optical coherence tomography, the system being used to execute the aforementioned laser beam conformal machining detection method based on optical coherence tomography, specifically including: The error localization module is used to acquire the current contour map of the area to be processed through optical coherence tomography, compare the current contour map with the target contour map to obtain the contour depth error pixels, divide the contour depth error pixels into regions, and obtain several connected candidate regions. The beam analysis module determines the optimal incident direction of the shaping beam for processing the candidate region based on the distribution characteristics of the internal profile depth error pixels for each candidate region. The aspherical parameters of the aspherical mirror system used to generate the shaping beam are used as optimization variables to construct the initial population of each candidate region. The index construction module is used to determine the shaping beam corresponding to each individual in the initial population based on the aspherical mirror system, and to construct optimization indexes by combining the distribution of profile depth error pixels within the error candidate region. The dominance segmentation module is used to select individuals from the initial population based on optimization indicators, and to divide the selected individuals into sets of shape-dominant individuals, boundary-dominant individuals, and energy-dominant individuals. The combination cross module is used to select one individual from each of the shape advantage individual set, the boundary advantage individual set, and the energy advantage individual set to form several sets of cross parent combinations; The cross-feedback module is used to perform three-parent weighted cross-feeding on each parent in each cross-parent combination based on updatable weight coefficients to obtain cross-offspring. Based on the improvement of each cross-offspring in the optimization index, the weight coefficients in the corresponding cross-parent combination are adaptively updated; multiple cross-offspring are obtained and the cross-offspring are assigned to the second group. The beam determination module is used to update the initial population after performing mutation operations on the crossover offspring, iteratively execute the index construction module and the crossover feedback module until the termination condition is met. When the termination condition is met, the optimal individual is determined in the second population, and the shaping beam corresponding to the optimal individual is incident on the candidate region along the optimal incident direction of the corresponding candidate region for laser beam conformal processing.

[0017] Compared with the prior art, the beneficial effects of the present invention are: Optical coherence tomography (OCT) is used to acquire the current three-dimensional shape of the area to be processed and compare it with the target shape to obtain profile depth error pixels. Based on this, the profile depth error pixels are divided into regions to obtain several connected candidate regions. The optimal incident direction of the shaping beam for processing the candidate region can be determined based on the distribution characteristics of the profile depth error pixels within each candidate region. This allows for targeted identification and processing control of depth and profile errors in different local regions, which is beneficial for improving the error detection accuracy, shaping accuracy, and processing accuracy of complex three-dimensional surfaces. By constructing an initial population using aspherical parameters as optimization variables, determining the shaping beam corresponding to each body based on the finite element model, and constructing optimization indices by combining the pixel distribution of profile depth error within the candidate region, each body is further divided into a set of shape-advantageous individuals, a set of boundary-advantageous individuals, and a set of energy-advantageous individuals. The optimal individuals can be obtained through three-parent weighted crossover, adaptive weight update, and mutation iteration. Thus, a more coordinated optimization can be achieved between the shape adaptation, boundary matching, and energy distribution of the shaping beam, making the obtained shaping beam more closely fit the depth and profile error distribution of the candidate region, thereby improving the effect of laser beam conformal processing. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a schematic diagram of the three-parent generation weighted crossover process of the present invention; Figure 3 This is a schematic diagram illustrating the adaptive update of the weight coefficients in this invention; Figure 4 This is a schematic diagram of the overall system structure of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0020] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0021] Example: Please see Figure 1 The present invention provides a technical solution: A laser beam conformal processing detection method based on optical coherence tomography includes the following steps: S1: Obtain the current contour map of the area to be processed through optical coherence tomography, compare the current contour map with the target contour map to obtain contour depth error pixels, divide the contour depth error pixels into regions, and obtain several connected candidate regions. Furthermore, comparing the current contour image with the target contour image includes: The current contour map is a two-dimensional topography map containing depth information; Perform spatial registration between the current contour and the target contour so that they are located in the same coordinate system; Based on the depth difference between the current contour and the target contour at the corresponding positions, determine the contour depth error value of each pixel; The profile depth error value is compared with a preset threshold, and pixels that are greater than the preset threshold are identified as profile depth error pixels.

[0022] The current contour map is a two-dimensional topography map containing depth information, and each pixel in the two-dimensional topography map corresponds to the depth value of a sampling position in the area to be processed; the target contour map is a two-dimensional topography map containing target depth information generated based on the design model, processing path model or preset target topography data of the area to be processed.

[0023] During the comparison, the current contour image and the target contour image are first spatially registered so that the pixels in the current contour image and the pixels in the target contour image are located in the same coordinate system. Specifically, the planar coordinate system of the target contour image is used as the reference coordinate system, and the current contour image is translated, rotated and scaled so that the reference feature points, boundary feature points or preset marker points in the current contour image coincide with the corresponding positions in the target contour image. After completing the spatial registration, the current contour image and the target contour image are resampled so that they have the same pixel resolution and the same pixel index range.

[0024] For any registered pixel, obtain the current depth value of the current contour map at that pixel and the target depth value of the target contour map at the corresponding pixel. Use the difference between the current depth value and the target depth value as the depth difference of the pixel and the absolute value of the depth difference as the profile depth error value of the pixel. Thus, a profile depth error map with the same size as the current contour map can be obtained.

[0025] Furthermore, the profile depth error value of each pixel is compared with a preset threshold. When the profile depth error value of a certain pixel is greater than the preset threshold, the pixel is determined as a profile depth error pixel. When the profile depth error value of a certain pixel is less than or equal to the preset threshold, the current profile at that pixel is considered to meet the processing accuracy requirements, and it is not regarded as a profile depth error pixel.

[0026] The preset threshold is used to distinguish between effective profile error and detection noise or allowable processing deviation. The preset threshold is not a fixed empirical value, but is determined based on the repeatability error of the optical coherence tomography (OCT) equipment, the design allowable error of the workpiece to be processed, and the minimum effective shaping depth of the processing technology. Specifically, before formal inspection, a known and stable reference area is selected in the area to be processed. The OCT equipment is used to repeatedly scan the reference area to obtain the depth measurement value of the same pixel position in multiple scans, and the standard deviation of the depth measurement value of each pixel position is calculated. The average of the standard deviations of all reference pixel positions is used to obtain the repeatability noise value of the equipment.

[0027] Simultaneously, the allowable depth deviation value of the target contour is obtained according to the design requirements of the workpiece to be processed, and the minimum effective shaping depth that can be stably generated by a single laser shaping is determined according to the energy control accuracy and material removal characteristics of the laser beam conformal processing system. Then, the maximum value among the preset multiple of the repeated measurement noise value of the equipment, the allowable depth deviation value of the target contour, and the minimum effective shaping depth is determined as the preset threshold.

[0028] In specific settings, if the device repeatedly measures the noise value, it is recorded as follows: The allowable depth deviation value for the target contour is denoted as The minimum effective shaping depth is denoted as Then the preset threshold Td can be set as follows: in, This is the noise amplification factor, used to avoid misinterpreting equipment measurement fluctuations as true profile errors. Preferably, Take a value of 2 to 3; when the surface roughness of the area to be processed is large or the optical coherence tomography signal fluctuates strongly. Take 3; when the surface of the area to be processed is relatively flat and the imaging signal is stable, The value is set to 2. This ensures that the preset threshold can both eliminate the influence of detection noise and retain the true depth error region requiring laser reshaping compensation.

[0029] In one specific implementation, the same reference area is scanned continuously at least five times, and the depth map obtained from each scan is acquired; the standard deviation of the depth values ​​from multiple scans is calculated for each pixel location, and the average of the standard deviations of all pixel locations is taken as the depth map. Read the allowable depth deviation value from the product design documents or target contour model as... Determined based on the minimum depth that can be stably removed or compensated in laser processing calibration experiments. Finally, according to: Determine a preset threshold for the current workpiece or the current batch of workpieces. Use this preset threshold to perform a pixel-by-pixel comparison between the current contour image and the target contour image to obtain the contour depth error pixels.

[0030] The existing technology divides the contour depth error pixels into several connected candidate regions based on the connected component labeling algorithm. Specifically, it involves traversing each contour depth error pixel in the current contour map, and when an unlabeled current contour map is encountered, it is used as a seed point of a new connected component.

[0031] Starting from the seed point, use a breadth-first search or depth-first search algorithm to recursively find and label all contour depth error pixels connected to it by the 8-connectivity rule. This process will form an independent, connected candidate region and assign it a unique label. Repeat this process until all contour depth error pixels in the image have been visited and labeled.

[0032] S2: For each candidate region, the optimal incident direction of the shaping beam for processing the candidate region is determined based on the distribution characteristics of the internal profile depth error pixels. The aspherical parameters of the aspherical mirror system used to generate the shaping beam are used as optimization variables to construct the initial population of each candidate region. Furthermore, the logic for determining the optimal incident angle based on the depth difference is as follows: For each candidate region, divide the candidate region into two regions by a straight line passing through the center of the candidate region, calculate the sum of the depth differences of all pixels in the two regions, and calculate the absolute difference of the sum of the depth differences between the two regions. Furthermore, using the center of the candidate region as the rotation center, the dividing line is continuously rotated according to a preset angle step size, and the dividing deviation at different angles is repeatedly calculated. Since the dividing line coincides with the original line direction after rotating 180°, it can be traversed within the range of 0° to 180°. The preset angle step size can be determined according to the angle adjustment accuracy of the laser processing equipment and the pixel resolution of the current contour map; for example, when the equipment angle adjustment accuracy is 1°, the preset angle step size is set to 1°; when it is necessary to improve the search accuracy, a coarse search step size of 5° can be used to determine the angle range with smaller dividing deviations, and then a second search can be performed within this angle range using a fine search step size of 0.5° or 1°.

[0033] After calculating the segmentation deviation of the dividing lines at each angle, the dividing line with the smallest segmentation deviation is selected and determined as the depth bisector. The depth bisector indicates that after the candidate region is divided by the line, the total depth difference between the two sides is the closest, thus representing the balance direction of the depth error distribution within the candidate region.

[0034] Subsequently, an incident plane is constructed based on the depth bisectors. Specifically, a spatial plane is constructed that is perpendicular to the plane containing the current contour map and includes the depth bisectors. This spatial plane is used as the incident plane. Since the incident plane includes both the depth direction and the balance direction of the depth error distribution within the candidate region, the candidate incident directions can be restricted to a spatial range related to the error distribution of the current candidate region, reducing invalid direction searches.

[0035] Multiple candidate incident rays are arranged within the incident plane. Each candidate incident ray passes through the center of a candidate region and has a different included angle relative to the normal of the current contour map. The angular range of the candidate incident rays is determined based on the allowable deflection range of the laser processing head, the workpiece surface occlusion, and the laser focal length constraint; for example, when the maximum allowable deflection angle of the laser processing head relative to the normal is 60°, candidate incident rays can be arranged within the range of 0° to 60°. The angular interval between adjacent candidate incident rays can be set to 0.5° to 5°, preferably consistent with the minimum angle adjustment accuracy of the laser processing head.

[0036] For each profile depth error pixel within the candidate region, a depth difference line segment is constructed based on the current depth value and the target depth value at that pixel. One end of the depth difference line segment is located at the depth position corresponding to the current profile, and the other end is located at the depth position corresponding to the target profile. Its length represents the amount of depth error that needs to be corrected at that pixel. The projection length of each depth difference line segment in the direction of each candidate incident ray is calculated, and all projection lengths corresponding to the same candidate incident ray are accumulated to obtain the total projection length corresponding to the candidate incident ray.

[0037] The larger the total projected length, the better the match between the candidate incident ray and the main depth error direction within the candidate region. This means that when the laser beam acts along this direction, it is more effective in covering areas with large depth errors. Therefore, after calculating the total projected length of all candidate incident rays, the candidate incident ray with the largest total projected length is selected as the target ray, and the opposite direction of this target ray is determined as the optimal incident direction.

[0038] Furthermore, the angle between the optimal incident direction and the normal of the current contour map is determined as the optimal incident angle. When the candidate region is located in a curved surface region, a weighted average normal can be calculated based on the local surface normal of each contour depth error pixel within the candidate region, and the angle between the optimal incident direction and the weighted average normal is determined as the optimal incident angle, where the weighting coefficient is the contour depth error value of the corresponding pixel. This ensures that the determined optimal incident angle can simultaneously adapt to the surface pose and depth error distribution of the candidate region, improving the incident matching during conformal laser beam processing.

[0039] Furthermore, the logic for projecting the depth difference along the optimal incident direction to generate a deviation projection map with energy distribution characteristics is as follows: Using the optimal incident direction as the projection axis, the profile depth error is mapped to the optimal incident direction to obtain the error projection value of each error pixel. A blank grayscale image is created based on the current contour map. The error projection values ​​of each projection deviation point are normalized and converted to grayscale to generate a deviation projection map of the standard incident space contour.

[0040] Furthermore, for each aspherical parameter, a random number is generated within its corresponding preset range. All random values ​​generated by the aspherical parameters constitute an individual, and multiple individuals are generated to form the initial population. The aspherical parameters can be selected according to the specific aspherical mirror system. The vertex curvature, conic coefficient, and higher-order aspherical coefficients of each aspherical optical element are used as optimization parameters, while the parameters of the remaining aspherical mirror system are fixed. The preset range of each aspherical parameter is determined based on the historical set of valid aspherical parameters and the system's allowed parameter range. The maximum and minimum values ​​of the historical valid aspherical parameters are obtained, and an interval is determined between the maximum and minimum values. The system's allowed interval is obtained, and the intersection of the two intervals is taken as the preset range.

[0041] The aspherical mirror system mentioned is existing technology. Refer to "Design and Optimization of Aspherical Mirror System for Gaussian Beam Shaping into Flat-Top Beam" in Journal of Huanggang Normal University, Vol. 43, No. 3, 2023, or "Particle Swarm Optimization Design Method for Gaussian Beam Shaping Aspherical Lens" in Infrared and Laser Engineering, Vol. 46, No. 12, 2017, and will not be elaborated here.

[0042] S3: Based on the aspherical mirror system, determine the shaping beam corresponding to each individual in the initial population, and construct optimization index by combining the distribution of profile depth error pixels within the error candidate region; The calculated deviation projection map and energy distribution map are registered, and the light intensity of the shaped beam is normalized and converted to grayscale to generate the energy distribution map. Texture features of the gray-level co-occurrence matrix of the deviation projection map and the energy distribution map are extracted. Texture feature vectors are constructed from the texture features of the gray-level co-occurrence matrix of the deviation projection map and the energy distribution map. The texture features include homogeneity, contrast and entropy. The cosine similarity of the texture feature vectors of the deviation projection map and the shaping beam is calculated as a shape optimization index. The larger this index is, the higher the similarity between the deviation projection map and the shaping beam energy distribution map in terms of texture structure, grayscale changes and overall distribution pattern. This indicates that the shaping beam generated by the current aspherical parameters is more suitable for the current depth error candidate region in terms of shape.

[0043] The edge pixels of the deviation projection map and the energy distribution map are extracted based on the Sobel operator and the corresponding edge pixel set is formed. The intersection-union ratio of the edge pixel sets of the deviation projection map and the energy distribution map is used as the boundary optimization index. The larger the index, the higher the degree of overlap between the boundary of the energy distribution map of the shaping beam and the boundary of the deviation projection map. This indicates that the effective range of the current shaping beam matches the boundary of the target error region better, which can reduce the risk of beam spillover or under-coverage of the boundary.

[0044] The Pearson correlation coefficient between the deviation projection map and the energy distribution map is calculated as an energy optimization index. The larger the index, the more consistent the pixel grayscale change trend is between the energy distribution map and the deviation projection map, indicating that the energy distribution of the shaping beam can better match the energy compensation requirements of the profile depth error pixels.

[0045] S4: Select individuals from the initial population based on the optimization index, and divide the selected individuals into sets of shape-dominant individuals, boundary-dominant individuals, and energy-dominant individuals. A conventional genetic algorithm is a global optimization algorithm that simulates the processes of natural selection and genetic evolution. It typically encodes the parameters of the problem to be optimized as individuals and randomly generates multiple individuals to form an initial population. Then, it evaluates the quality of each individual according to a preset fitness function, retains individuals with higher fitness through selection, and generates offspring by exchanging genetic information between parent individuals through crossover. Mutation is then used to randomly perturb some offspring to increase population diversity. Through multiple rounds of iterative processes of "fitness evaluation—selection—crossover—mutation—population update," the population gradually converges towards a better solution region, ultimately obtaining a better combination of parameters that satisfies the optimization objective. This application improves the crossover operation; the remaining steps all employ a conventional genetic algorithm.

[0046] The selected bodies are sorted in descending order of shape optimization index, boundary optimization index, and energy optimization index, respectively, to obtain the corresponding shape sorting result, boundary sorting result, and energy sorting result. For each individual, compare its ranking in the shape ranking, boundary ranking, and energy ranking results from largest to smallest. If its ranking in the shape ranking result is the lowest, define the individual as a shape-dominant individual and form a shape-dominant individual set. If its ranking in the boundary ranking result is the lowest, define the individual as a boundary-dominant individual and form a boundary-dominant individual set. If its ranking in the energy ranking result is the lowest, define the individual as an energy-dominant individual and form a boundary-dominant individual set.

[0047] Shape optimization, boundary optimization, and energy optimization indices reflect the degree of matching between the shaping beam and the depth error candidate region in terms of overall shape, action boundary, and energy distribution, respectively. These three indices represent different optimization directions, and simply using a comprehensive score can easily lead to the averaging of individuals excelling in one aspect. By ranking the offspring population used for crossover according to each of the three indices and comparing the relative rankings of the same individual in the three rankings, the individual's most prominent strength can be identified. If an individual ranks lowest in the shape ranking, it indicates relatively better shape matching and is suitable as a shape-dominant individual for crossover. If it ranks lowest in the boundary ranking, it indicates relatively better boundary fit and is suitable as a boundary-dominant individual. If it ranks lowest in the energy ranking, it indicates relatively better energy distribution matching and is suitable as an energy-dominant individual. This allows for the preservation of the advantageous attributes of different individuals before crossover, enabling the subsequent weighted crossover of the three parent generations to simultaneously introduce three types of advantageous information: shape, boundary, and energy. This avoids the loss of certain types of advantageous information due to the selection of parent generations based solely on a single comprehensive fitness, and increases the probability of generating offspring that simultaneously meet the requirements of shape matching, boundary matching, and energy matching.

[0048] S5: Select one individual from each of the shape-dominant individual set, the boundary-dominant individual set, and the energy-dominant individual set to form several sets of cross-parent combinations; S6: Based on the updatable weight coefficients, perform three-parent weighted cross-crossing on each parent in each cross-parent combination to obtain cross-offspring. Adaptively update the weight coefficients in the corresponding cross-parent combination according to the improvement of each cross-offspring in the optimization index; obtain multiple cross-offspring and classify the cross-offspring into the second group. Please see Figure 2 , Figure 2 This is a schematic diagram of the three-parent generation weighted crossover process of the present invention.

[0049] Furthermore, based on the improvement of each crossover offspring in terms of optimization metrics, the logic for adaptively updating the weight coefficients in the corresponding crossover parent combination is as follows: Based on the updatable weight coefficients, a three-parent weighted cross is performed on each parent in each cross parent combination to obtain cross offspring. The weight coefficients in the corresponding cross parent combination are adaptively updated according to the improvement of each cross offspring in the optimization index. Multiple cross offspring are obtained and the cross offspring are assigned to the second group. Each time, a crossover parent combination is selected for a three-parent weighted crossover. In the first weighted crossover, all weight coefficients are initialized to be equal, and the sum of the weight coefficients is 1. Weighted crossover is performed according to the initialized weight coefficients to obtain crossover offspring. The optimization index of the crossover parents is weighted using the initialized weight coefficients to obtain a comprehensive optimization index. For each optimization index, the comprehensive optimization index is subtracted from the optimization index of the offspring, and the ratio of the result to the maximum value of the corresponding optimization index in the initial population is calculated to obtain the improvement amount of the optimization index. The improvement amount of the optimization index is compared with 0. If it is greater than 0, 0 is set as the penalty term for the optimization index. If it is not greater than 0, a penalty factor for the optimization index is preset, and the penalty term is set as the product of the optimization index and the penalty factor. The improvement amount of the optimization index is added to the penalty terms of the other optimization indices except for the optimization index to serve as the feedback term for the optimization index.

[0050] For each three-parent weighted cross except the first three-parent weighted cross, the average of all feedback terms before that three-parent weighted cross is obtained, which is called the average feedback term of that three-parent weighted cross. The feedback term of each optimization index in the previous three-parent weighted cross is obtained, and the average of the feedback term of the previous three-parent weighted cross and the average feedback term of that three-parent weighted cross is calculated. This average is then substituted into the Sigmoid function, and the result is multiplied by the weight coefficient of the optimization index in the previous three-parent weighted cross, which is used as the feedback weight coefficient of that three-parent weighted cross. The feedback weight coefficients of all optimization indices in that three-parent weighted cross are normalized to obtain the weight coefficient of that three-parent weighted cross.

[0051] In each weighted crossover, a shape-dominant individual vector, a boundary-dominant individual vector, and a shape-energy-dominant individual vector are selected from the offspring population used for crossover as the crossover parent. In the first weighted crossover, each selected individual vector is weighted according to initialized weights to obtain the crossover offspring. The optimization index of the crossover parent is then weighted using the initialized weights to obtain the comprehensive optimization index. In the three-parent weighted crossover, shape-dominant, boundary-dominant, and energy-dominant individuals represent three different optimization directions. Since the crossover offspring are generated jointly by the three parents according to their weights, an offspring cannot be compared to any single parent. Instead, the optimization indexes of the three parents should be weighted according to the current crossover weights to obtain a reference value for the comprehensive optimization index of this crossover.

[0052] For each optimization index, the optimization index of the offspring is subtracted from the comprehensive optimization index, and the result is compared with the maximum value of the corresponding optimization index in the initial population to obtain the improvement of the optimization index. Dividing the difference by the maximum value of the corresponding optimization index in the initial population is to unify the scale of different indexes. Because the numerical ranges of shape optimization index, boundary optimization index and energy optimization index may be different, direct comparison will lead to the index with a larger numerical scale dominating the feedback result. After normalization, the improvement of the three indexes is comparable.

[0053] For each optimization indicator, the improvement amount of the optimization indicator is compared with 0. If it is greater than 0, then 0 is set as the penalty term for the optimization indicator. If it is not greater than 0, then the penalty factor of the optimization indicator is preset, and the penalty term is set as the product of the optimization indicator and the penalty factor. For each optimization metric, first determine whether its improvement amount is greater than 0. If it is greater than 0, it means that the metric has not degenerated and no penalty is set; if it is less than or equal to 0, it means that the metric has not improved relative to the parent weighted reference value. Multiply the improvement amount by the preset penalty factor to obtain the penalty term for the metric.

[0054] The penalty term here is essentially a degradation inhibition term.

[0055] Since the improvement amount corresponding to the unimproved indicator is negative, and multiplying it by the penalty factor still results in a negative value, adding a feedback item will reduce the final feedback value; this avoids the following situation: The offspring shape index improved, but the boundary or energy deteriorated significantly, and it was still mistakenly considered a valid crossover result. Similarly, the boundary feedback term and energy feedback term also introduce penalty terms for the other two indicators, so that the feedback of any indicator is constrained by the degradation of other indicators. The improvement amount of the optimization indicator is added to the penalty terms of the other optimization indicators, and this is used as the feedback term for the optimization indicator.

[0056] Furthermore, the logic for adaptively updating the weight coefficients of the contribution ratios of each parent generation in the next round of crossover based on the improvement of the corresponding indicators of each offspring generation is as follows: For each three-parent generation weighted crossover except for the first three-parent generation weighted crossover, the average value of all feedback terms before the current three-parent generation weighted crossover is obtained, which is called the average feedback term of the current three-parent generation weighted crossover. The feedback term of each optimization indicator in the previous three-parent generation weighted crossover is obtained, and the average value of the feedback term of the previous three-parent generation weighted crossover and the average feedback term of the current three-parent generation weighted crossover is calculated and substituted into the Sigmoid function. The result is multiplied by the weight coefficient of the optimization indicator in the previous three-parent generation weighted crossover, which is used as the feedback weight coefficient of the current three-parent generation weighted crossover. The feedback weight coefficients of all optimization indicators in the current three-parent generation weighted crossover are normalized to obtain the weight coefficient of the current three-parent generation weighted crossover.

[0057] By comparing the various optimization indicators of the cross-generation with the weighted comprehensive indicators of the current cross-generation, and normalizing the improvement amount, the feedback results in the three directions of shape, boundary, and energy are made comparable. At the same time, by setting a penalty term to suppress the degradation of other indicators, the effectiveness of the offspring is not mistakenly judged due to the improvement of a single indicator. This allows each cross-generation result to adjust the contribution ratio of each parent generation in the next round, forming an adaptive optimization closed loop of "cross-evaluation-feedback-re-cross". This improves the comprehensive adaptability of the shaping beam in terms of shape matching, boundary matching, and energy matching during the optimization of aspherical parameters.

[0058] In each weighted crossover, a shape advantage individual vector, a boundary advantage individual vector, and a shape energy individual vector are selected from the offspring population used for crossover as the crossover parent. In the first weighted crossover, each selected individual vector is weighted according to the initialized weights to obtain the crossover offspring. The optimization index of the crossover parent is weighted using the initialized weights to obtain the comprehensive optimization index. For each optimization index, the comprehensive optimization index is subtracted from the optimization index of the offspring, and the ratio of the result to the maximum value of the corresponding optimization index in the initial population is calculated to obtain the improvement of the optimization index. The improvement of the optimization index is compared with 0. If it is greater than 0, 0 is set as the penalty term for the optimization index. If it is not greater than 0, a penalty factor for the optimization index is preset, and the penalty term is set as the product of the optimization index and the penalty factor. The improvement of the optimization index is added to the penalty terms of the other optimization indices except for the optimization index to obtain the feedback term for the optimization index.

[0059] By selecting shape-dominant, boundary-dominant, and energy-dominant individual vectors as crossover parents and performing weighted crossover according to initial weights, the crossover offspring can simultaneously integrate three types of advantage information. By comparing the optimized indices of the crossover offspring with the weighted comprehensive optimized indices of the crossover parents, it can be determined whether the crossover has actually improved the overall level relative to the parents. Dividing by the maximum value of the corresponding optimized indices in the initial population can eliminate numerical scale differences between different optimized indices, making the improvements in shape, boundary, and energy indices comparable. By setting penalty terms for unimproved indices and adding the penalty terms of the remaining optimized indices to the feedback term of the current optimized indices, it can be prevented that offspring that only improve in a single index while significantly deteriorating in other indices are still misjudged as valid crossover results. Therefore, each weighted crossover can generate feedback information that can be used to adjust the contribution ratio of subsequent parents, transforming the crossover process from fixed-weight fusion into a closed-loop process of adaptive correction based on offspring performance, thereby improving the comprehensive optimization capability of the shaping beam in terms of shape matching, boundary matching, and energy matching.

[0060] Please see Figure 3 , Figure 3 This is a schematic diagram illustrating the adaptive update of the weight coefficients in this invention.

[0061] S7: After performing mutation on the crossover offspring, update the initial population. Iterate through steps S4 to S7 until the termination condition is met. When the termination condition is met, determine the optimal individual in the second population and incident the shaping beam corresponding to the optimal individual along the optimal incident direction of the corresponding candidate region to the candidate region for laser beam conformal processing.

[0062] Furthermore, the logic for determining the optimal individual is as follows: among the offspring population that meets the termination condition, the offspring population is sorted non-dominated based on the optimization index to obtain multiple non-dominated layers. In the first non-dominated layer, the individual with the largest energy optimization index is selected as the optimal individual.

[0063] Shape optimization index, boundary optimization index, and energy optimization index reflect the degree of matching between the shaping beam and the candidate region of profile depth error in terms of overall shape, action boundary, and energy distribution, respectively. The relationship between these three indices is not such that a larger single index can completely replace the relationships between the others. Therefore, a non-dominated sorting of the offspring population is first performed to select individuals in the first non-dominated layer that have a comprehensive advantage among multiple optimization indices and are not completely dominated by other individuals. This avoids a significant decrease in shape matching or boundary coverage performance due to selection based solely on a single index. Furthermore, since the ultimate goal of laser beam conformal processing is to effectively compensate the processing area according to the profile depth error distribution, a larger energy optimization index indicates a more consistent intensity distribution of the shaping beam with the pixel grayscale variation trend of the deviation projection map, which better meets the energy compensation requirements of the error region. Therefore, in the first non-dominated layer, the individual with the largest energy optimization index is further selected as the optimal individual. This ensures that the shaping beam with the most suitable energy distribution for the current candidate region is obtained first, while maintaining good shape and boundary matching. This reduces the risk of under-processing or over-processing in certain areas and improves the correction accuracy of laser beam conformal processing.

[0064] Please see Figure 4 The present invention further provides a laser beam conformal processing detection system based on optical coherence tomography, wherein the aforementioned laser beam conformal processing detection method based on optical coherence tomography specifically includes: The error localization module is used to acquire the current contour map of the area to be processed through optical coherence tomography, compare the current contour map with the target contour map to obtain the contour depth error pixels, divide the contour depth error pixels into regions, and obtain several connected candidate regions. The beam analysis module determines the optimal incident direction of the shaping beam for processing the candidate region based on the distribution characteristics of the internal profile depth error pixels for each candidate region. The aspherical parameters of the aspherical mirror system used to generate the shaping beam are used as optimization variables to construct the initial population of each candidate region. The index construction module is used to determine the shaping beam corresponding to each individual in the initial population based on the aspherical mirror system, and to construct optimization indexes by combining the distribution of profile depth error pixels within the error candidate region. The dominance segmentation module is used to select individuals from the initial population based on optimization indicators, and to divide the selected individuals into sets of shape-dominant individuals, boundary-dominant individuals, and energy-dominant individuals. The combination cross module is used to select one individual from each of the shape advantage individual set, the boundary advantage individual set, and the energy advantage individual set to form several sets of cross parent combinations; The cross-feedback module is used to perform three-parent weighted cross-feeding on each parent in each cross-parent combination based on updatable weight coefficients to obtain cross-offspring. Based on the improvement of each cross-offspring in the optimization index, the weight coefficients in the corresponding cross-parent combination are adaptively updated; multiple cross-offspring are obtained and the cross-offspring are assigned to the second group. The beam determination module is used to update the initial population after performing mutation operations on the crossover offspring, iteratively execute the index construction module and the crossover feedback module until the termination condition is met. When the termination condition is met, the optimal individual is determined in the second population, and the shaping beam corresponding to the optimal individual is incident on the candidate region along the optimal incident direction of the corresponding candidate region for laser beam conformal processing.

[0065] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0066] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0067] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0068] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that cannot be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A laser beam conformal processing detection method based on optical coherence tomography, characterized in that, The specific steps include: S1: Obtain the current contour map of the area to be processed through optical coherence tomography, compare the current contour map with the target contour map to obtain contour depth error pixels, divide the contour depth error pixels into regions, and obtain several connected candidate regions. S2: For each candidate region, the optimal incident direction of the shaping beam for processing the candidate region is determined based on the distribution characteristics of the internal profile depth error pixels. The aspherical parameters of the aspherical mirror system used to generate the shaping beam are used as optimization variables to construct the initial population of each candidate region. The logic for determining the optimal incident angle based on the depth difference is as follows: For each candidate region, the candidate region is divided into two regions by a straight line passing through the midpoint of the candidate region. The sum of the depth differences of all pixels in the two regions is calculated, and the absolute difference of the sum of the depth differences of the two regions is calculated and defined as the segmentation deviation. The line is rotated continuously and the calculation is repeated. After traversing one cycle, the straight line with the smallest segmentation deviation is selected. An incident plane is constructed based on this straight line. The incident plane is a plane that is perpendicular to the current contour map plane and contains the depth bisector. Multiple candidate incident rays are arranged in the plane. The total projection length of all depth difference line segments on each candidate incident ray is counted. The ray with the largest total projection length is selected and its reverse direction is determined as the optimal incident direction. S3: Based on the aspherical mirror system, determine the shaping beam corresponding to each individual in the initial population, and construct optimization index by combining the distribution of profile depth error pixels within the error candidate region; For each aspherical parameter, a random number is generated within its corresponding preset range. All random values ​​generated by the aspherical parameters constitute an individual, and multiple individuals are generated to form the initial population. The calculated deviation projection map and energy distribution map are registered, and the light intensity of the shaped beam is normalized and converted to grayscale to generate the energy distribution map. Texture features are extracted from the gray-level co-occurrence matrices of the deviation projection map and the energy distribution map to construct texture feature vectors. The texture features include homogeneity, contrast, and entropy. The cosine similarity between the texture feature vectors of the deviation projection map and the shaping beam is calculated as a shape optimization index. The edge pixels of the deviation projection map and the energy distribution map are extracted based on the Sobel operator and the corresponding edge pixel set is formed. The intersection-union ratio of the edge pixel sets of the deviation projection map and the energy distribution map is used as the boundary optimization index. Calculate the Pearson correlation coefficient between the deviation projection map and the energy distribution map as an energy optimization index; S4: Select individuals from the initial population based on the optimization index, and divide the selected individuals into sets of shape-dominant individuals, boundary-dominant individuals, and energy-dominant individuals. S5: Select one individual from each of the shape-dominant individual set, the boundary-dominant individual set, and the energy-dominant individual set to form several sets of cross-parent combinations; S6: Based on the updatable weight coefficients, perform three-parent weighted cross-crossing on each parent in each cross-parent combination to obtain cross-offspring. Adaptively update the weight coefficients in the corresponding cross-parent combination according to the improvement of each cross-offspring in the optimization index; obtain multiple cross-offspring and classify the cross-offspring into the second group. S7: After performing mutation on the crossover offspring, update the initial population. Iterate through steps S4 to S7 until the termination condition is met. When the termination condition is met, determine the optimal individual in the second population and incident the shaping beam corresponding to the optimal individual along the optimal incident direction of the corresponding candidate region to the candidate region for laser beam conformal processing.

2. The laser beam conformal processing detection method based on optical coherence tomography according to claim 1, characterized in that: Comparing the current contour image with the target contour image includes: The current contour map is a two-dimensional topography map containing depth information; Perform spatial registration between the current contour and the target contour so that they are located in the same coordinate system; Based on the depth difference between the current contour and the target contour at the corresponding positions, determine the contour depth error value of each pixel; The profile depth error value is compared with a preset threshold, and pixels that are greater than the preset threshold are identified as profile depth error pixels.

3. The laser beam conformal processing detection method based on optical coherence tomography according to claim 2, characterized in that: The logic for projecting the depth difference along the optimal incident direction to generate a deviation projection map with energy distribution characteristics is as follows: Using the optimal incident direction as the projection axis, the profile depth error is mapped to the optimal incident direction to obtain the error projection value of each error pixel. A blank grayscale image is created based on the current contour map. The error projection values ​​of each projection deviation point are normalized and converted to grayscale to generate a deviation projection map of the standard incident space contour.

4. The laser beam conformal processing detection method based on optical coherence tomography according to claim 1, characterized in that: The logic for dividing the selected individuals into sets of shape-advantageous individuals, boundary-advantageous individuals, and energy-advantageous individuals is as follows: selection refers to the selection operation in a multi-objective genetic algorithm; The selected bodies are sorted in descending order of shape optimization index, boundary optimization index, and energy optimization index, respectively, to obtain the corresponding shape sorting result, boundary sorting result, and energy sorting result. For each individual, compare its ranking in the shape ranking, boundary ranking, and energy ranking results from largest to smallest. If its ranking in the shape ranking result is the lowest, define the individual as a shape-dominant individual and form a shape-dominant individual set. If its ranking in the boundary ranking result is the lowest, define the individual as a boundary-dominant individual and form a boundary-dominant individual set. If its ranking in the energy ranking result is the lowest, define the individual as an energy-dominant individual and form a boundary-dominant individual set.

5. The laser beam conformal processing detection method based on optical coherence tomography according to claim 4, characterized in that: The logic for adaptively updating the weight coefficients in the corresponding crossover parent combination based on the improvement of each crossover offspring in terms of optimization metrics is as follows: Based on the updatable weight coefficients, a three-parent weighted cross is performed on each parent in each cross parent combination to obtain cross offspring. The weight coefficients in the corresponding cross parent combination are adaptively updated according to the improvement of each cross offspring in the optimization index. Multiple cross offspring are obtained and the cross offspring are assigned to the second group. Each time, a crossover parent combination is selected for a three-parent weighted crossover. In the first weighted crossover, all weight coefficients are initialized to be equal, and the sum of the weight coefficients is 1. The weighted crossover is performed according to the initialized weight coefficients to obtain the crossover offspring. The optimization index of the crossover parent is weighted using the initialized weight coefficients to obtain the comprehensive optimization index. For each optimization index, the comprehensive optimization index is subtracted from the optimization index of the offspring, and the ratio of the result to the maximum value of the corresponding optimization index in the initial population is calculated to obtain the improvement amount of the optimization index. The improvement amount of the optimization index is compared with 0. If it is greater than 0, 0 is set as the penalty term for the optimization index. If it is not greater than 0, the penalty factor of the optimization index is preset, and the penalty term is set as the product of the optimization index and the penalty factor. The improvement amount of the optimization index is added to the penalty terms of the other optimization indices except for the optimization index as the feedback term of the optimization index. For each three-parent weighted cross except the first three-parent weighted cross, the average of all feedback terms before that three-parent weighted cross is obtained, which is called the average feedback term of that three-parent weighted cross. The feedback term of each optimization index in the previous three-parent weighted cross is obtained, and the average of the feedback term of the previous three-parent weighted cross and the average feedback term of that three-parent weighted cross is calculated. This average is then substituted into the Sigmoid function, and the result is multiplied by the weight coefficient of the optimization index in the previous three-parent weighted cross, which is used as the feedback weight coefficient of that three-parent weighted cross. The feedback weight coefficients of all optimization indices in that three-parent weighted cross are normalized to obtain the weight coefficient of that three-parent weighted cross.

6. The laser beam conformal processing detection method based on optical coherence tomography according to claim 1, characterized in that: The logic for determining the optimal individual is as follows: In the offspring population that meets the termination condition, the offspring population is sorted non-dominated based on the optimization index to obtain multiple non-dominated layers. In the first non-dominated layer, the individual with the largest energy optimization index is selected as the optimal individual.

7. A laser beam conformal processing and inspection system based on optical coherence tomography, characterized in that: The system is used to implement the laser beam conformal processing detection method based on optical coherence tomography as described in any one of claims 1-6, specifically including: The error localization module is used to acquire the current contour map of the area to be processed through optical coherence tomography, compare the current contour map with the target contour map to obtain the contour depth error pixels, divide the contour depth error pixels into regions, and obtain several connected candidate regions. The beam analysis module determines the optimal incident direction of the shaping beam for processing the candidate region based on the distribution characteristics of the internal profile depth error pixels for each candidate region. The aspherical parameters of the aspherical mirror system used to generate the shaping beam are used as optimization variables to construct the initial population of each candidate region. The index construction module is used to determine the shaping beam corresponding to each individual in the initial population based on the aspherical mirror system, and to construct optimization indexes by combining the distribution of profile depth error pixels within the error candidate region. The dominance segmentation module is used to select individuals from the initial population based on optimization indicators, and to divide the selected individuals into sets of shape-dominant individuals, boundary-dominant individuals, and energy-dominant individuals. The combination cross module is used to select one individual from each of the shape advantage individual set, the boundary advantage individual set, and the energy advantage individual set to form several sets of cross parent combinations; The cross-feedback module is used to perform three-parent weighted cross-feeding on each parent in each cross-parent combination based on updatable weight coefficients to obtain cross-offspring. Based on the improvement of each cross-offspring in the optimization index, the weight coefficients in the corresponding cross-parent combination are adaptively updated; multiple cross-offspring are obtained and the cross-offspring are assigned to the second group. The beam determination module is used to update the initial population after performing mutation operations on the crossover offspring, iteratively execute the index construction module and the crossover feedback module until the termination condition is met. When the termination condition is met, the optimal individual is determined in the second population, and the shaping beam corresponding to the optimal individual is incident on the candidate region along the optimal incident direction of the corresponding candidate region for laser beam conformal processing.

Images