A method for gas film hole segmentation and center positioning based on region prompting and human-computer collaborative correction

Abstract

The application discloses a kind of based on regional hint and man-machine collaborative correction air film hole segmentation and center positioning method, to solve the problem of complex surface texture, low contrast, boundary fuzzy and local reflection condition under air film hole automatic segmentation is prone to missing, misclassification and center positioning stability insufficient problem.This method first obtains the image containing air film hole to be handled and establishes original image coordinate system, generates the regional hint of target air film hole according to artificial frame selection or target detection result;Subsequently, the image to be handled and regional hint are input into hint type segmentation model, and the initial mask of target air film hole is obtained;For the missing area and misclassification area in initial mask, respectively, add and erase operation is performed, to obtain the final mask after artificial collaborative correction;Further, the final mask is extracted and boundary contour analysis is carried out, and the reliability of elliptical fitting result is judged according to the number of contour points, elliptical axis length, fitting center position and area consistency, and the elliptical fitting center is used when fitting is reliable, and the mask geometric centroid is used when fitting is unreliable;Finally, the air film hole center coordinates, mask area, equivalent diameter, elliptical long axis, short axis and direction angle are output.The application can improve the accuracy of air film hole segmentation result under complex imaging condition and the reliability of center positioning and geometric parameter extraction.

Description

Technical Field

[0001] This invention relates to the fields of machine vision, image segmentation, and industrial precision inspection technology, specifically to a method for segmenting and centering film vents based on region prompting and human-machine collaborative correction. This method is used to accurately extract film vent regions from surface images of components such as turbine blades of aero-engines and to determine the center position and related geometric parameters of the film vents. Background Technology

[0002] Film cooling holes are important microstructures on the surface of high-temperature components such as turbine blades in aero-engines. They introduce cooling airflow onto the blade surface to form a film layer, thereby reducing the thermal load of the high-temperature combustion gases on the blade substrate. The location, shape, and distribution of the film cooling holes affect the coverage effect of the cooling airflow. Therefore, in turbine blade manufacturing inspection, defect identification, hole location measurement, and subsequent automatic alignment of measurement probes, it is necessary to accurately extract the film cooling hole region from the blade surface image and obtain the center position and related geometric parameters of the film cooling holes.

[0003] Existing methods for processing images of film vents mainly include thresholding, edge detection, region growing, morphological processing, and contour-based target extraction. These methods can achieve certain segmentation results when there is a significant grayscale difference between the film vent and the blade surface, stable lighting conditions, and clear vent boundaries. However, actual blade surfaces typically exhibit complex textures, processing marks, localized reflections, shadows, uneven grayscale, and imaging noise. Film vents may also present tilted, elliptical, or irregular shapes depending on the shooting angle. Under these conditions, traditional image processing methods are prone to problems such as boundary interruptions, background missegmentation, missed segmentation of areas within the vent, and misidentification of adjacent textures. Furthermore, they usually require readjusting the segmentation threshold and processing parameters for different images, making it difficult to guarantee the stability and applicability of the segmentation results.

[0004] With the development of deep learning image segmentation technology, neural network-based automatic segmentation methods are increasingly being applied to industrial target extraction. These methods can improve segmentation capabilities in complex backgrounds by learning image features, but they typically rely on large amounts of high-quality pixel-level labeled data, resulting in high costs for model training and data annotation. When the actual image differs from the training data in terms of lighting, viewpoint, surface texture, or imaging quality, automatic segmentation results may still suffer from omissions, misclassifications, and boundary shifts. Cue-based image segmentation methods can define the target to be segmented based on rectangular regions or feature points, offering advantages in reducing background interference. However, their output is still affected by the sharpness of the target boundary, the range of the region cues, and local image features, making it difficult to guarantee that the initial mask for each air-supported membrane aperture accurately matches the actual aperture boundary.

[0005] In locating the center of the air film aperture, existing methods typically employ the center of the detection frame, the geometric centroid of the mask, or a fitted center using an ellipse. The detection frame center only reflects the location of the candidate region and is insufficient to accurately characterize the actual geometric center of the air film aperture; the geometric centroid of the mask is easily affected by local misclassification or omissions; ellipse fitting can obtain geometric parameters such as the center and major and minor axes using the aperture contour, but when the contour is incomplete, the boundary is distorted, or there are insufficient effective contour points, the fitting result may have significant deviations or even fail. Furthermore, existing technologies usually treat automatic segmentation, manual correction, and center calculation as independent processing steps. After the mask is manually modified, the original center position and geometric parameters may not be promptly invalidated and recalculated based on the correction results, easily leading to inconsistencies between the mask results and the center location results.

[0006] Therefore, how to reduce the interference of complex blade surface background on the segmentation of film vents through region prompts, manually and collaboratively correct the missed and mis-segmented regions based on the automatically generated initial mask, and make a reasonable selection between the ellipse fitting center and the mask geometric centroid according to the reliability of contour fitting, while ensuring that the center and geometric parameters are updated synchronously after mask correction, are technical problems that urgently need to be solved in this field. Summary of the Invention

[0007] The purpose of this invention is to provide a method for segmenting and centering film vents based on region prompting and human-computer collaborative correction, in order to solve the problems that existing film vent image processing methods are prone to missing, missegmenting, and boundary shifting of film vents under conditions of complex blade surface texture, low contrast between target and background, blurred vent boundaries, and interference from local reflections or shadows, which further leads to inaccurate results in the centering and geometric parameter extraction of film vents.

[0008] To achieve the above objectives, this invention proposes a method for air film aperture segmentation and center localization based on region prompting and human-machine collaborative correction, comprising the following steps: Step 1: Acquire the image of the air film aperture and establish the original image coordinate system. Acquire the image to be processed containing at least one air film aperture, record the width and height of the image to be processed, and establish the original image coordinate system with the upper left corner of the image as the origin. The horizontal direction is the x-axis direction and the vertical direction is the y-axis direction, so that subsequent area prompts, mask pixels, contour points and air film aperture centers are all uniformly represented in the original image coordinate system.

[0009] Step 2: Generate air film pore region hints. Identify candidate regions containing the target air film pores in the image to be processed, and generate rectangular region hints surrounding the target air film pores. The area corresponding to each air film pore is indicated as follows: in, and These represent the x and y coordinates of the top-left corner of the region prompt in the original image coordinate system, respectively. and These represent the width and height of the region hint, respectively. The region hint can be generated manually by selecting a bounding box, or through target detection, reading existing annotations, or other target localization methods. The boundary of the region hint is located outside the target film vent, ensuring that the target film vent is completely contained within the region hint, while minimizing interference from blade surface texture, adjacent film vents, and other background areas on the segmentation process.

[0010] Step 3: Generate the initial mask for the air film aperture based on region cues. Input the image to be processed into the cue-based segmentation model for image feature encoding to obtain the image features corresponding to the image to be processed; convert the region cues generated in Step 2 into rectangular cues recognizable by the cue-based segmentation model, and input the rectangular cues and image features together into the mask decoding process to obtain the initial mask corresponding to the target air film aperture and its prediction evaluation value. For multiple air film apertures in the same image to be processed, it is preferable to perform image feature encoding only once on the image to be processed, and to reuse the image features for multiple region cues to reduce redundant computation. When the cue-based segmentation model outputs multiple prediction results, the mask with the highest prediction evaluation value is selected as the first mask. Initial mask for each air film aperture The initial mask is then mapped onto the original image coordinate system.

[0011] Step 4: Perform human-machine collaborative correction on the initial mask. The initial mask is overlaid on the image to be processed or a partial image of the target film cooling aperture. Based on the correspondence between the boundary of the initial mask and the actual aperture boundary of the target film cooling aperture, it is determined whether the initial mask needs correction. When the initial mask misses areas that do not cover the actual film cooling aperture, the missed areas are added to the initial mask using an add operation; when the initial mask incorrectly covers the blade background or other non-film cooling aperture areas, the incorrectly covered areas are deleted from the initial mask using an erase operation. Let the added area be... The area to be erased is The final mask after manual collaborative correction Represented as: in, This indicates that the added region will be incorporated into the initial mask. This indicates the removal of the erased area from the mask. During the manual collaborative correction process, the scope of the add or erase operation can be adjusted, and undo, redo, and restore the initial mask are supported. After confirming the correction result, the corrected mask is used as the final mask for the target air film aperture, and the center position and geometric parameters obtained based on the mask before correction are invalidated to avoid inconsistencies between the mask data and the geometric calculation results.

[0012] Step 5: Extract the effective connected components and boundary contours of the final mask. The final mask is binarized and connected component analysis is performed to remove isolated pixels or invalid connected regions unrelated to the target air film aperture, thus determining the effective connected components corresponding to the target air film aperture. Preferably, the effective connected components of the target air film aperture are determined from the final mask based on the area of ​​the connected components, the distance between the connected components and the region hint center, and the overlap relationship between the connected components and the region hint. Boundary tracking is performed on the effective connected components to obtain a set of air film aperture contour points arranged in spatial order, providing a data foundation for subsequent ellipse fitting and geometric parameter calculation.

[0013] Step 6: Determine the center of the air film aperture based on the reliability of the ellipse fitting. First, calculate the geometric centroid of the mask based on the effective pixels of the final mask. Let the first... The set of pixels in the effective masking area of ​​each air hole is The number of effective pixels is Then its geometric centroid for: Subsequently, an ellipse fitting is performed on the set of air film aperture contour points obtained in step 5 to obtain the ellipse fitting center, major axis, minor axis, and principal axis direction angle. The reliability of the ellipse fitting result is judged based on at least one of the following indicators: the number of contour points, the effectiveness of the major and minor axes, the ratio of the major and minor axes, whether the fitting center is located inside the final mask, and the consistency between the area of ​​the fitted ellipse and the area of ​​the mask.

[0014] When the ellipse fitting result meets the preset reliability conditions, the center of the ellipse fitting is taken as the final center of the air film aperture; when the number of contour points is insufficient, the fitting axis length is abnormal, the fitting center is located outside the final mask, or the area consistency does not meet the preset conditions, the ellipse fitting result is determined to be unreliable, and the geometric centroid of the mask is taken as the final center of the air film aperture. Therefore, the... The final center of each air film pore Represented as: in, The center of the ellipse fitting is The centroid of the mask geometry.

[0015] Step 7: Output the center and geometric parameters of the film membrane aperture. Calculate the mask area of ​​the film membrane aperture based on the final mask. And calculate the equivalent diameter based on the mask area. : Simultaneously, the final center coordinates, center positioning method, mask area, equivalent diameter, major axis, minor axis, and principal axis direction angle of the air film aperture are output. For air film apertures that have undergone manual collaborative correction, the contour is re-extracted and all center and geometric parameters are recalculated based on the corrected final mask to ensure that the output results correspond to the final mask.

[0016] The beneficial effects of this invention are as follows: This invention limits the processing range of the target film pore by region-based prompts, reducing interference from complex blade surface textures, adjacent film pores, and non-target backgrounds on the segmentation process; obtaining an initial mask for the film pore through prompt-based segmentation, and manually co-correcting missed and misclassified areas in the initial mask using add and erase operations, improves the completeness and accuracy of film pore boundary extraction; by assessing the reliability of the ellipse fitting results and adaptively selecting between the ellipse fitting center and the mask's geometric centroid, it avoids center positioning failures caused by contour defects, boundary distortions, or fitting anomalies; by invalidating the original geometric results and recalculating the center and related parameters after mask correction, it ensures consistency between the final mask, center position, and geometric parameters, thereby improving the reliability of film pore segmentation and center positioning results under complex imaging conditions. Attached Figure Description

[0017] Figure 1 This is a flowchart of the overall process for the air film pore segmentation and center positioning method based on region prompting and human-machine collaborative correction according to the present invention.

[0018] Figure 2 This is a schematic diagram illustrating the process of generating the initial mask for the air film aperture based on region prompts according to the present invention.

[0019] Figure 3 This is a schematic diagram of the human-machine collaborative correction process of the air film aperture mask according to the present invention.

[0020] Figure 4 This is a schematic diagram of the process of locating the center of the air film pore and extracting geometric parameters according to the present invention.

[0021] Figure 5 This is a schematic diagram showing the segmentation and centering results of different air film pore images according to the present invention. Detailed Implementation

[0022] To make the technical solution, technical features, and beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the following embodiments are only used to illustrate the technical solution of the present invention and are not intended to limit the scope of protection of the present invention. Equivalent substitutions and adjustments made by those skilled in the art to the processing steps, parameter settings, or implementation methods without departing from the technical concept of the present invention should all fall within the scope of protection of the present invention.

[0023] This embodiment takes the image of film venting holes on the surface of aero-engine turbine blades as the processing object. It uses a region-based prompting image segmentation model to generate an initial mask for the film venting holes. The initial mask is then collaboratively corrected through manual addition and erasure operations. Finally, the contour extraction, ellipse fitting, center reliability judgment, and geometric parameter calculation of the film venting holes are completed based on the corrected final mask.

[0024] like Figure 1 As shown, the air film pore segmentation and center positioning method described in this embodiment includes the following steps.

[0025] Step 1: Acquire the image of the air film vents to be processed. Use an industrial camera or other vision imaging device to acquire an image of the blade surface containing one or more air film vents, and read the width and height of the image. Establish an original image coordinate system with the top left corner of the image as the origin. Subsequent region hints, masks, outlines, and center positions are all uniformly represented in this coordinate system. When the image is scaled in the processing interface, the interactive operation position is converted to the original image coordinates according to the scaling relationship between the displayed image and the original image, avoiding coordinate deviations between region hints, masks, and center positions caused by image scaling.

[0026] Step 2: Generating prompts for the air film pore area. For example... Figure 2 (a) and Figure 2 As shown in (b), the region where the target air-film pore is located is determined in the image to be processed, and a rectangular region hint surrounding the target air-film pore is generated. The region hint should completely contain the target air-film pore, and retain an appropriate amount of background area outside the boundary of the air-film pore, so that the segmentation model can simultaneously acquire image features of the target and the adjacent background. In this embodiment, the region hint is generated by manual bounding box selection. The operator drags in the image to form a rectangular box, and after completing the bounding box selection, records the position and size of the rectangular box in the original image coordinate system. In other embodiments, the region hint can also be automatically generated by the target detection model, or read from existing annotation files and historical processing results. When using the target detection model to generate the region hint, the detection box can be appropriately expanded outward to ensure that the air-film pore is completely contained within the region hint.

[0027] Step 3: Generate the initial mask for the film air vents. For example... Figure 2As shown, the image to be processed is input into a prompting segmentation model for image feature extraction, and the rectangular region of the target air-supported membrane aperture is prompted into the model to obtain the initial mask corresponding to the target air-supported membrane aperture. When the prompting segmentation model generates multiple candidate masks, the mask with the highest evaluation is automatically selected as the initial mask based on the prediction evaluation result given by the model. The processing interface only displays the finally selected initial mask. For multiple air-supported membrane apertures in the same image, it is preferable to perform feature extraction only once on the original image and save the obtained image features. The image features can be reused for prompting different air-supported membrane aperture regions. Only prompting processing and mask generation need to be re-executed, thereby reducing redundant calculations and improving the efficiency of continuous processing of multiple air-supported membrane apertures. Figure 2 As shown in (c), the initial mask is represented by a binary image, where the foreground region corresponds to the target air film aperture, and the background region corresponds to the non-air film aperture region. Figure 2 As shown in (d), the initial mask can also be superimposed on the original image in a semi-transparent manner to observe the correspondence between the mask boundary and the actual air film aperture boundary.

[0028] Step 4: Human-machine collaborative correction of the initial mask. After obtaining the initial mask, determine whether it completely covers the target film cooling aperture and whether it incorrectly includes the blade background, reflective areas, shadow areas, or other non-target areas. When the boundary of the initial mask is basically consistent with the actual film cooling aperture boundary, the initial mask can be directly confirmed as the final mask. When the initial mask has omissions, misclassifications, or local boundary deviations, it enters the manual fine-tuning process. Figure 3 As shown, manual fine-tuning mainly includes addition and erasure operations. For areas not covered by the initial mask but actually belonging to the air film aperture, the addition operation adds them to the mask; for areas incorrectly covered by the initial mask but actually belonging to the background, the erasure operation removes them from the mask. Addition and erasure operations can use a circular action area with adjustable size. The operator moves the action area within the local image of the air film aperture, continuously modifying the mask pixels at the corresponding positions. During manual fine-tuning, functions such as undo, redo, and restore the initial mask can be set to adjust the correction process. After the operator confirms the fine-tuning results, the current mask is saved as the final mask. When the mask is added, erased, or restored, the center position and geometric parameters calculated based on the original mask are marked as invalid and recalculated after the final mask is confirmed, thus ensuring that the mask result is consistent with the center positioning result.

[0029] Step 5: Extract the effective air film pore area and boundary contour. For example... Figure 4As shown in (a), connected component analysis is performed on the final mask to remove isolated noise points, non-target small regions, and mask portions unrelated to the target air film aperture. When only one effective connected region exists in the final mask, that region is directly identified as the target air film aperture region. When multiple connected regions exist, the target region can be determined based on conditions such as region area, distance between the region and the center of the prompt box, and overlap between the region and the prompt box. In this embodiment, the connected region corresponding to the prompt center and meeting the area requirements is preferred. When the prompt center is not located within any connected region, the connected region closest to the prompt center or with the largest area can be selected as the target air film aperture region. After determining the effective air film aperture region, its boundary is extracted to obtain the air film aperture contour. To reduce the impact of a small number of burrs and isolated pixels on subsequent processing, hole filling, isolated point removal, or boundary smoothing can be performed before contour extraction.

[0030] Step 6: Perform ellipse fitting and reliability assessment. For example... Figure 4 As shown in (b), an ellipse is fitted based on the extracted air film aperture contour to obtain parameters such as the ellipse center, major axis, minor axis, and orientation angle. To avoid abnormal fitting caused by contour defects, boundary distortion, or insufficient effective contour points, this embodiment performs a reliability judgment on the ellipse fitting result. The reliability judgment can comprehensively consider the number of contour points, whether the major and minor axes are effective, whether the ratio of the major and minor axes is reasonable, whether the fitting center is located inside the final mask, and whether the area of ​​the fitted ellipse is consistent with the area of ​​the mask. When the above conditions meet the preset requirements, the ellipse fitting result is determined to be reliable; when the number of contour points is insufficient, the axis length is abnormal, the fitting center is located outside the mask, or the fitting result differs greatly from the actual mask, the ellipse fitting result is determined to be unreliable.

[0031] Step 7: Determine the center of the film aperture. While performing ellipse fitting, calculate the geometric centroid of the mask based on the final effective mask. When the ellipse fitting result is reliable, the center of the ellipse fitting is used as the final center of the film aperture. This center can better reflect the elliptical shape of the film aperture opening under tilted imaging conditions. When the ellipse fitting result is unreliable, the geometric centroid of the mask is used as the final center of the film aperture to avoid significant shifts in the center position or positioning failures caused by abnormal fitting results. Figure 4 As shown in (c), the determined final center, air film pore outline and fitted ellipse are superimposed and displayed in the original image, and the center positioning method used is indicated.

[0032] Step 8: Calculate and output geometric parameters. Based on the final mask and fitting results, calculate and output parameters such as the center coordinates of the air film aperture, mask area, equivalent diameter, major axis of the ellipse, minor axis of the ellipse, and orientation angle. When the image has been spatially calibrated, the area, axis length, and equivalent diameter in pixels can be converted to actual dimensions based on the pixel size. For manually corrected air film apertures, all geometric parameters are recalculated based on the corrected final mask. The processing results may also include air film aperture number, region indication location, initial mask, final mask, ellipse fitting reliability, center positioning method, and whether manual correction was performed. The above results can be saved by overlaying images, data tables, or structured files for subsequent result verification, air film aperture position measurement, measurement probe alignment, or automated detection. Figure 5 As shown, for air-supported holes with clear boundaries and significant differences between the target and background, the initial mask generated by the cue-based segmentation model can cover the target area relatively completely. In this case, the initial mask can be directly used as the final mask for center localization. For air-supported holes with local reflections, shadows, blurred boundaries, or complex texture interference, the initial mask may miss or misclassify them. After manually adding and erasing the initial mask, a final mask that better matches the actual hole boundary can be obtained, and the center and geometric parameters can be recalculated based on the final mask. When there are multiple air-supported holes in the same image, region cue can be created for each air-supported hole in sequence, and the initial mask generation, manual correction, and center localization can be performed separately. Each air-supported hole saves its region cue, mask state, and geometric parameters, and the correction operation for one air-supported hole will not affect the processing results of other air-supported holes.

[0033] The above description is merely a preferred embodiment of the present invention. Without departing from the inventive concept, region prompts can also employ point prompts, multiple point prompts, or a combination of rectangular and point prompts; the prompt-based segmentation model can be replaced with other image segmentation models capable of generating target masks based on spatial prompts; the connected component selection method, ellipse fitting reliability conditions, and result storage format can also be adjusted according to actual application requirements. All equivalent substitutions, modifications, or improvements described above should fall within the protection scope of the present invention.

Claims

1. A method for air film aperture segmentation and center localization based on region prompting and human-machine collaborative correction, characterized in that, Includes the following steps: Step 1: Obtain an image to be processed containing at least one air film pore, and establish the original image coordinate system of the image to be processed; Step 2: Determine the region where the target air film pore is located in the image to be processed, and generate a region hint that completely surrounds the target air film pore; Step 3: Input the image to be processed into the prompting segmentation model for image feature extraction, and input the region prompt as a segmentation prompt into the prompting segmentation model to obtain the initial binary mask of the target air film pore; Step 4: Overlay the initial binary mask onto the image to be processed. In response to the mask addition command for the missed area and the mask erasure command for the misclassified area, update the initial binary mask pixel by pixel to obtain the final mask of the target air film aperture. If no mask correction instruction is received, the initial binary mask will be used as the final mask. Step 5: Perform connected component analysis on the final mask and determine the effective connected components corresponding to the target air film apertures. Extract the boundary contours of the effective connected components, perform ellipse fitting on the boundary contours, and calculate the geometric centroid of the final mask. Step 6: Determine whether the ellipse fitting result is reliable according to the preset fitting reliability conditions. If the ellipse fitting result is reliable, determine the ellipse fitting center as the target air film aperture center. If the ellipse fitting result is unreliable, determine the geometric centroid of the final mask as the target air film aperture center. Step 7: Output the center coordinates and geometric parameters of the target air film hole based on the final mask and the center of the target air film hole; when the final mask is modified, the center coordinates and geometric parameters obtained based on the mask before modification become invalid, and the effective connected component extraction, boundary contour extraction, center positioning and geometric parameter calculation are re-performed based on the modified final mask.

2. The method for air film aperture segmentation and center positioning based on region prompting and human-machine collaborative correction according to claim 1, characterized in that: The region hint is a rectangular region hint, generated by at least one of the following methods: manual selection, target detection model output, or reading of existing region annotations; when the image to be processed is scaled or partially cropped, the coordinates of the region hint in the displayed image are transformed to the original image coordinate system, so that the region hint, the initial binary mask, the final mask, the boundary contour, and the center of the target air film hole correspond to the same original image coordinate system.

3. The method for air film aperture segmentation and center positioning based on region prompting and human-machine collaborative correction according to claim 1, characterized in that: The cue-based segmentation model includes image feature encoding, region cue encoding, and mask decoding processes. For multiple target air holes in the same image to be processed, image feature extraction is performed only once on the image to be processed, and the obtained image features are cached. The region cue corresponding to each target air hole reuses the image features. When the cue-based segmentation model generates multiple candidate masks, the candidate mask with the highest prediction evaluation value is selected as the initial binary mask based on the prediction evaluation value corresponding to each candidate mask.

4. The method for air film aperture segmentation and center positioning based on region prompting and human-machine collaborative correction according to claim 1, characterized in that: The mask addition instruction is used to set the mask pixels in the specified action area as the foreground of the air film aperture, and the mask erasure instruction is used to set the mask pixels in the specified action area as the background; the human-machine collaborative correction also includes at least one of the following operations: action area size adjustment, undo, redo, and restore the initial binary mask, and the corresponding mask state is recorded after each mask modification.

5. The method for air film aperture segmentation and center positioning based on region prompting and human-machine collaborative correction according to claim 1, characterized in that: When multiple connected regions exist in the final mask, the effective connected region corresponding to the target air film aperture is determined based on at least one of the following: the area of ​​the connected region, whether the connected region contains the region hint center, the distance between the centroid of the connected region and the region hint center, and the overlap relationship between the connected region and the region hint. The preset fitting reliability conditions include at least one of the following: the number of boundary contour points, the validity of the major and minor axes of the fitted ellipse, the ratio of the major axis to the minor axis, whether the ellipse fitting center is located inside the final mask, the consistency between the area of ​​the fitted ellipse and the area of ​​the final mask, and the fitting residual of the boundary contour relative to the fitted ellipse.

6. The method for air film aperture segmentation and center positioning based on region prompting and human-machine collaborative correction according to claim 1, characterized in that: The geometric parameters include at least one of the following: final mask area, equivalent diameter, major axis of the fitted ellipse, minor axis of the fitted ellipse, and orientation angle of the fitted ellipse; the output results also include the target air film hole number, area prompt position, center positioning method, ellipse fitting reliability status, and whether the final mask has undergone human-machine collaborative correction.

Images