Granite line processing calibration system and method based on industrial vision

Abstract

The application relates to the fields of industrial visual detection and stone processing calibration, and discloses a granite line processing calibration system and method based on industrial visual detection, which comprises the following steps: obtaining the processing specifications of a granite line, and establishing a segmented standard contour model, wherein the segmented standard contour model comprises an outer boundary reference segment, a groove line reference segment, a chamfer reference segment, an arc transition reference segment and an end surface positioning reference segment; each reference segment is respectively configured with a candidate search area, a deviation type and a calibration trigger condition. The granite line processing calibration system and method based on industrial visual detection can limit the candidate search areas corresponding to the outer boundary reference segment, the groove line reference segment, the chamfer reference segment, the arc transition reference segment and the end surface positioning reference segment through the segmented standard contour model, can reduce the misleading of low-trust visual contours to the calibration of processing equipment, and can improve the reliability of granite line processing deviation identification and the stability of calibration actions.

Description

Technical Field

[0001] This invention relates to the field of industrial visual inspection and stone processing calibration technology, and in particular to a granite line processing calibration system and method based on industrial visual inspection. Background Technology

[0002] In current granite molding processes, manual inspection, template comparison, or images of the processed surface captured by industrial cameras are typically used. The identified actual contour is then compared with a standard contour to determine if there are any processing deviations in the outer boundaries, grooves, chamfers, curved transition lines, and end face positioning lines. When the inspection results show that the actual contour deviates from the standard contour, existing methods usually generate corresponding compensation or adjustment instructions based on the deviation results, thereby calibrating or correcting the processing equipment.

[0003] However, granite lines are subject to interference from natural textures, color bands, polished reflections, watermarks, dust, and localized chipping. Ordinary visual inspection can easily mistake natural texture lines, reflective lines, or edges of localized defects for processing boundaries. The deviations caused by these misidentifications do not necessarily originate from equipment reference offsets or continuous processing errors. Existing technologies typically focus on identifying the contour deviation itself, lacking a mechanism to jointly assess the boundary reliability of the linear response, the continuous length of the deviation, the consistency of the deviation direction, and the correlation between adjacent processing sections before generating calibration parameters. This means that non-equipment-related deviations caused by texture interference, reflective interference, localized chipping, or abnormal clamping of individual pieces can still trigger equipment calibration, leading to erroneous compensation by the processing equipment and affecting the stability of subsequent continuous processing of granite lines. Summary of the Invention

[0004] The technical problem to be solved by the present invention is that the existing technology has the disadvantage of preventing non-equipment visual deviations from accidentally triggering equipment calibration. To this end, we propose a calibration system and method for granite line processing based on industrial visual inspection.

[0005] To achieve the above objectives, this application adopts the following technical solution: a method for calibrating granite line processing based on industrial visual inspection, comprising:

[0006] Obtain the processing specifications of granite lines and establish a segmented standard contour model. The segmented standard contour model includes an outer boundary reference segment, a groove reference segment, a chamfer reference segment, an arc transition reference segment, and an end face positioning reference segment. Each reference segment is configured with a candidate search area, a deviation type, and a calibration trigger condition.

[0007] The diffuse reference image, low-angle oblique light image, and cross-polarization image of the processing surface are collected as three types of images, and the three types of images are registered to the coordinate system of the processing surface.

[0008] Linear responses are extracted within the candidate search area. Based on the three types of images, texture continuity evaluation, concavity / convexity structure evaluation, and reflectivity attenuation evaluation are determined respectively. The boundary confidence of the linear responses is also determined.

[0009] The linear response whose boundary confidence satisfies the corresponding benchmark segment confidence threshold is used as the processing boundary candidate line. The actual processing contour is extracted and the actual processing contour is registered with the segmented standard contour model to obtain the deviation amount, deviation continuity length, deviation direction consistency, intra-segment confidence and adjacent segment association consistency.

[0010] Based on the intra-segment confidence, the deviation continuity length, the deviation direction consistency, and the adjacent segment correlation consistency, the deviation source is identified and calibration trigger qualifications are generated.

[0011] When the source of the deviation is identified as a machine-related processing deviation and the calibration triggering condition is met, the calibration parameter generation result is output; when the source of the deviation is identified as texture interference, reflection interference, local edge chipping, or single-piece clamping abnormality, the calibration parameter suppression result is output.

[0012] After the processing equipment performs calibration based on the calibration parameters, the three types of images are re-acquired, and based on the segmented standard contour model and the candidate search area, the extraction of the actual processing contour, the identification of the source of deviation, and the judgment of the calibration trigger qualification are performed again for verification. The calibration pass result, the equipment reference parameter update result, or the abnormal verification result are output.

[0013] Preferably, the candidate search region is determined based on the contour position and allowable deviation range of each reference segment in the coordinate system of the machining surface, and is extended along the normal direction, contour extension direction or surface transition direction of the corresponding reference segment; the calibration triggering conditions include boundary confidence conditions, deviation continuity length conditions, deviation direction consistency conditions and adjacent segment association conditions.

[0014] Preferably, registering the three types of images to the processing surface coordinate system includes: performing coordinate transformation on the three types of images based on camera calibration parameters, end face positioning reference, outer boundary stable segment or image feature points, and performing local fine registration on the three types of images based on the outer boundary stable segment, end face positioning line or high confidence contour response.

[0015] Preferably, the texture continuity evaluation value is determined based on the main direction of the natural texture in the candidate search region or its neighboring region, the angle between the linear response and the main direction of the natural texture, the continuous length of the linear response, and the gray-level gradient change rate on both sides of the linear response.

[0016] Preferably, the evaluation value of the concave-convex structure is determined based on the difference in the mean gray value, the shadow width, and the gradient peak value of the sampling bands on both sides of the linear response in the low-angle oblique light image; the evaluation value of the reflective attenuation is determined based on the gray value difference or gray value ratio of the corresponding area of ​​the linear response in the diffuse reference image and the cross-polarization image, as well as the contour response retention status.

[0017] Preferably, the boundary confidence level is determined based on the convex-concave structure evaluation value, the degree of alignment between the linear response and the corresponding reference segment, the texture continuity evaluation value, and the reflective interference factor; wherein the convex-concave structure evaluation value and the degree of alignment are positive factors, and the texture continuity evaluation value and the reflective interference factor are negative factors.

[0018] Preferably, when the actual machining contour is registered with the segmented standard contour model, the actual outer boundary, actual groove line, actual chamfer line, actual arc transition line and actual end face positioning line are respectively registered with the corresponding reference segment, and the intra-segment confidence is determined based on the boundary confidence of the machining boundary candidate lines participating in the registration.

[0019] Preferably, the adjacent segment correlation consistency is determined based on the deviation direction, deviation trend, and machining geometry relationship between adjacent reference segments; when multiple reference segments are offset in the same direction as a whole, the adjacent segment correlation consistency is used to identify single-piece clamping abnormalities; when the deviation of adjacent reference segments meets the corresponding machining geometry relationship, the adjacent segment correlation consistency is used to identify equipment-related machining deviations.

[0020] Preferably, the verification includes: when the verification deviation converges along the direction pointed to by the calibration parameter generation result and enters the allowable deviation range, outputting the calibration pass result; when the verification deviation is still located in the region within the segment with a confidence level higher than a preset threshold and maintains the same direction of change, outputting the device reference parameter update result; when the verification deviation shifts to the region within the segment with a confidence level lower than a preset threshold, the deviation direction changes abruptly, or the verification result is inconsistent with the calibration trigger qualification, outputting the abnormal verification result and suppressing secondary calibration.

[0021] A calibration system for granite line processing based on industrial visual inspection includes: a standard contour modeling unit, an image acquisition unit, an image registration unit, a line response extraction and evaluation unit, a boundary confidence determination unit, an actual processed contour extraction unit, a segmented deviation analysis unit, a deviation source identification unit, a calibration trigger qualification judgment unit, a calibration output unit, and a verification and update unit. The standard contour modeling unit is used to establish the segmented standard contour model; the image acquisition unit is used to acquire the three types of images; the image registration unit is used to register the three types of images; and the line response extraction and evaluation unit is used to extract the line response and determine the texture continuity evaluation quantity and the concavity / convexity. The system includes structural evaluation metrics and reflectivity attenuation evaluation metrics. The boundary confidence determination unit is used to determine the boundary confidence. The actual processing contour extraction unit is used to extract the actual processing contour. The segmented deviation analysis unit is used to obtain the deviation amount, deviation continuity length, deviation direction consistency, intra-segment confidence, and adjacent segment correlation consistency. The deviation source identification unit is used to identify the deviation source. The calibration trigger qualification judgment unit is used to generate the calibration trigger qualification. The calibration output unit is used to output the calibration parameter generation result or calibration parameter suppression result. The verification update unit is used to output the calibration pass result, equipment reference parameter update result, or anomaly verification result.

[0022] The technical effects and advantages of this invention are as follows:

[0023] In this invention, a segmented standard contour model is used to define candidate search areas corresponding to the outer boundary reference segment, groove reference segment, chamfer reference segment, arc transition reference segment, and end face positioning reference segment. A diffuse reference map, a low-angle oblique light map, and a cross-polarization map are used to form texture continuity evaluation, concave-convex structure evaluation, and reflectivity attenuation evaluation, respectively, and the boundary confidence of the linear response is determined accordingly. After extracting the actual processed contour and completing segmented registration, the source of deviation is identified by combining intra-segment confidence, deviation continuity length, deviation direction consistency, and adjacent segment correlation consistency. Calibration parameter generation results are output only when the deviation source is equipment-related processing deviation and meets the calibration triggering conditions. Calibration parameter suppression results are output when the deviation source is texture interference, reflectivity interference, local edge chipping, or single-piece clamping anomalies. The same rules are used for verification after calibration. Therefore, the misleading effect of low-confidence visual contours on processing equipment calibration can be reduced, improving the reliability of granite line processing deviation identification and the stability of calibration actions. Attached Figure Description

[0024] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts:

[0025] Figure 1This is a schematic diagram of the system structure of the present invention;

[0026] Figure 2 This is a schematic diagram of the multi-optical path acquisition and registration method of the present invention;

[0027] Figure 3 This is a flowchart illustrating the sources of deviation and calibration triggering in this invention.

[0028] Figure 4 This is a flowchart of the calibration output and verification closed loop of the present invention. Detailed Implementation

[0029] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0030] Reference Figure 1-4 As shown, the present invention provides a technical solution: a granite line processing calibration system and method based on industrial vision inspection, which is applicable to the processing inspection and calibration of granite lines after cutting, grooving, chamfering, grinding, polishing or composite molding.

[0031] Granite moldings can be used for door frames, window frames, baseboards, decorative trims, irregularly shaped stone moldings, or stone moldings with grooves, chamfers, or curved transition surfaces. Granite molding surfaces typically exhibit natural textures, color bands, polished reflections, watermarks, dust, localized chipping, and clamping deviations. Conventional image edge detection can easily misidentify natural texture lines or reflective boundaries as processed contours, potentially causing miscalibration of processing equipment.

[0032] This invention uses multi-path image evaluation, segmented standard contour constraints, boundary confidence judgment, deviation source identification, and calibration trigger qualification judgment to ensure the reliability of visual inspection results before triggering processing calibration, thereby reducing the misleading effect of low-confidence visual contours on processing calibration.

[0033] The system includes a conveying and positioning unit for transporting the granite line to be inspected to the inspection station and positioning and holding the granite line at the inspection station. The conveying and positioning unit includes a conveyor roller, lateral limiting components, clamping components, end face positioning components, support components, and a positioning sensor. The conveyor roller is used to carry and transport the granite line, the lateral limiting components are used to limit the lateral swing of the granite line, the clamping components are used to limit the vertical vibration of the granite line during image acquisition, the end face positioning components are used to form the end positioning reference of the granite line, and the positioning sensor is used to detect whether the granite line has entered the inspection area and sends an acquisition trigger signal to the image acquisition unit after the granite line reaches the preset inspection position.

[0034] The image acquisition unit is used to acquire multi-path images of the same processed surface of granite lines. The image acquisition unit includes an industrial camera, a diffuse light source, a low-angle oblique light source, and a cross-polarized light source. The industrial camera can be positioned above the processed surface of the granite line, or, depending on the cross-sectional shape of the granite line, above or to the side of the processed surface. The diffuse light source is used to obtain the overall grayscale distribution, color variations, and natural texture of the processed surface of the granite line. The low-angle oblique light source is used to enhance the unevenness response of grooves, chamfers, curved transition lines, and local chipped edges on the processed surface. The cross-polarized light source is used to reduce the interference of polished surfaces, watermarks, or localized high-brightness reflections on image recognition.

[0035] The image acquisition unit acquires a diffuse reference image, a low-angle oblique light image, and a cross-polarization image at the same detection position. The diffuse reference image is used to calculate the texture continuity evaluation value, the low-angle oblique light image is used to calculate the concave-convex structure evaluation value, and the cross-polarization image is used to calculate the reflectivity attenuation evaluation value and contour response retention status. All three types of images correspond to the same processing surface area. The acquisition time interval is controlled within the range where the granite line position is stable to reduce image shift caused by transportation or vibration.

[0036] The standard contour modeling unit establishes a segmented standard contour model based on the processing specifications of granite lines;

[0037] The segmented standard contour model includes an outer boundary reference segment, a groove reference segment, a chamfer reference segment, an arc transition reference segment, and an end face positioning reference segment. The outer boundary reference segment characterizes the target position of the outer edge of the granite line. The groove reference segment characterizes the target state of the groove location, width, and depth. The chamfer reference segment characterizes the chamfer edge position and angle. The arc transition reference segment characterizes the target contour trend of the curved transition area. The end face positioning reference segment characterizes the positioning reference at the end of the granite line.

[0038] Each reference segment is configured with a candidate search area, an allowable deviation range, a main deviation type, and a calibration trigger condition. The candidate search area refers to the image search range formed by expanding to both sides around the corresponding reference segment. The allowable deviation range is used to indicate the deviation range that the reference segment is allowed to have under qualified processing conditions. The main deviation type is used to limit the types of deviations that the reference segment participates in the judgment. The calibration trigger condition is used to limit whether the visual deviation of the reference segment can trigger the generation of calibration parameters.

[0039] The candidate search area for the outer boundary reference segment extends inward and outward along the outer boundary normal, with the main deviation type being lateral offset. The candidate search area for the groove reference segment is formed around the target groove line and groove bottom area, with the main deviation types being groove line position deviation and groove depth deviation. The candidate search area for the chamfer reference segment is formed around the target chamfer edge and chamfer surface, with the main deviation type being angular deviation. The candidate search area for the arc transition reference segment is formed around the target arc transition curve, with the main deviation type being surface transition deviation. The candidate search area for the end face positioning reference segment is formed around the end positioning edge of the line, with the main deviation types being end face positioning deviation and clamping offset.

[0040] The image registration unit is used to register the diffuse reference image, low-angle oblique light image, and cross-polarization image to the same processing surface coordinate system. The image registration unit can perform coordinate transformations based on camera calibration parameters, detection station coordinates, end-face positioning reference, outer boundary stable segments, or image feature points. After registration, the same processing surface area has corresponding coordinates in the three types of images. The system can calculate texture continuity evaluation, concavity / convexity structure evaluation, and reflectivity attenuation evaluation for the same linear response within the same candidate search area.

[0041] Considering the potential for minor vibrations in the granite lines during transport and positioning, the image registration unit performs local fine registration based on the high-confidence contour response within the stable outer boundary segment, end face positioning line, or candidate search area after coordinate transformation. Local fine registration can employ sub-pixel-level alignment methods based on contour mutual information, phase correlation, or feature point matching to reduce dynamic ghosting introduced by temporal flicker acquisition in multi-path images, ensuring sub-pixel-level correspondence of the same linear response in the diffuse reference image, low-angle oblique light image, and cross-polarization image.

[0042] The candidate search region determination unit is used to determine the candidate search regions corresponding to each reference segment in the registered image based on the segmented standard contour model. The unit first maps the segmented standard contour model to the machining surface coordinate system, and then determines the search width based on the allowable deviation range and machining tolerance of each reference segment. For the outer boundary reference segment, the candidate search region extends along the boundary normal. For the groove reference segment, the candidate search region covers the groove edge, groove bottom line, and groove wall transition area. For the chamfer reference segment, the candidate search region covers the chamfer edge and the chamfer surface intersection area. For the arc transition reference segment, the candidate search region covers the transition zones on both sides of the arc. For the end face positioning reference segment, the candidate search region covers the end face positioning line and its adjacent area. By limiting the candidate search region, the number of natural texture lines and non-machining linear responses entering subsequent judgments can be reduced.

[0043] The linear response extraction and evaluation unit is used to identify and extract linear responses within each candidate search region, and calculates three types of evaluation metrics for the same linear response. The linear response extraction and evaluation unit calculates a texture continuity evaluation metric based on a diffuse reference map. The texture continuity evaluation metric is used to represent the degree of similarity between the linear response and the natural texture of the granite.

[0044] The linear response extraction and evaluation unit can first statistically analyze the main directions of natural textures in the candidate search region or its neighboring regions, and then calculate the angle between the target linear response and the main directions of natural textures. Statistical analysis of the main directions of natural granite textures can be achieved using texture analysis algorithms such as gray-level co-occurrence matrix, gradient orientation histogram, or Fourier transform spectrum analysis, extracting the angle with the highest response energy in the candidate search region or its neighboring regions as the main directions of natural textures.

[0045] If the angle between the linear response and the main direction of the natural texture is less than or equal to 15°, the continuous pixel length of the linear response is greater than 50 pixels, and the rate of change of gray-level gradient on both sides of the normal of the linear response is lower than a preset smoothness threshold, then the linear response is considered to have a high texture continuity evaluation score. If the angle between the linear response and the main direction of the natural texture is greater than 15°, and the rate of change of gray-level gradient on both sides of its normal is higher than a preset abrupt change threshold, then the linear response is considered to not conform to the characteristics of natural texture.

[0046] The linear response extraction and evaluation unit calculates the convex-concave structure evaluation score based on a low-angle oblique light image. This unit can obtain the mean grayscale value, brightness-darkness difference, shadow width, and gradient peak value within a preset width range on both sides of the linear response. If a stable brightness-darkness difference exists on both sides of the linear response, and the shadow width is continuously distributed along the direction of the linear response, while the gradient peak value is concentrated near the linear response, then the convex-concave structure evaluation score of the linear response is high. If the linear response does not form a stable brightness-darkness transition in the low-angle oblique light image, or only exhibits color texture changes, then the convex-concave structure evaluation score is low.

[0047] The linear response extraction and evaluation unit calculates a reflectance attenuation evaluation value based on a cross-polarization diagram. This value indicates whether the linear response primarily originates from reflections, watermarks, or localized high brightness on a polished surface. The unit calculates the first average grayscale value of the corresponding region in the diffuse reference diagram and the second average grayscale value in the cross-polarization diagram. The grayscale difference or ratio between the first and second average grayscale values ​​is used as a quantitative indicator of the brightness reduction. Simultaneously, it determines whether the contour response of the linear response is preserved after the brightness reduction.

[0048] If the brightness decreases to a preset reflectivity attenuation threshold, and the outline response of the linear response disappears or significantly weakens after the brightness decreases, then the linear response is determined to have strong reflectivity interference characteristics. If the linear response still retains a continuous outline response after the brightness decreases, and there is a concave-convex structure support in the low-angle oblique light image, then the linear response is determined to be closer to the actual processing boundary.

[0049] The boundary confidence determination unit is used to determine the boundary confidence of the linear response based on texture continuity evaluation, convexity / concave structure evaluation, and reflectivity attenuation evaluation. Boundary confidence indicates the reliability of the linear response as an actual processed contour in subsequent segment registration and deviation assessment. Boundary confidence can be determined using a rule table, threshold judgment, or weighted scoring method.

[0050] Boundary confidence is determined by using the convexity / concave structure evaluation as a positive factor, and the texture continuity evaluation and reflective interference as negative factors, combined with the degree of positional fit between the linear response and the corresponding reference segment. Positional fit indicates the degree to which the linear response is located within the corresponding candidate search region and is close to the corresponding standard reference segment. If the linear response has a high convexity / concave structure evaluation, a high degree of positional fit, and retains the contour response in the cross-polarization diagram, the boundary confidence of the linear response is increased. If the linear response has a high texture continuity evaluation, a low convexity / concave structure evaluation, and the contour response disappears after reflective attenuation, the boundary confidence of the linear response is decreased.

[0051] Boundary credibility is expressed as an evaluation score, or boundary credibility score. The calculation formula is expressed as:

[0052] ;

[0053] in, For evaluating concave-convex structures, For the degree of fit, As a measure of texture continuity, It is a reflective interference factor; These are preset weighting coefficients. A higher evaluation score for the convex / concave structure, and a linear response closer to the corresponding baseline segment, results in a higher boundary confidence score. Conversely, a higher evaluation score for texture continuity, and a larger reflection interference factor, results in a lower boundary confidence score. When the boundary confidence score reaches the confidence threshold of the corresponding baseline segment, the linear response is marked as a candidate processing boundary line; when the boundary confidence score does not reach the confidence threshold of the corresponding baseline segment, the linear response is marked as a texture interference line, a reflection interference line, or a line to be verified.

[0054] The actual machining contour extraction unit is used to extract the actual machining contour based on boundary confidence. The actual machining contour includes the actual outer boundary, actual groove line, actual chamfer line, actual arc transition line, and actual end face positioning line. The actual machining contour extraction unit only uses linear responses that meet the corresponding benchmark segment confidence threshold as machining boundary candidate lines. It connects, fits, and removes outliers for the machining boundary candidate lines based on the line segment direction, endpoint distance, continuous length, and geometric constraints of the corresponding benchmark segment. For linear responses with low confidence, the actual machining contour extraction unit does not treat them as the main components of the actual machining contour, but retains their position, evaluation value, and labeling status for deviation source identification and verification.

[0055] The segmented deviation analysis unit is used to segment and register the actual machined contour with the corresponding reference segments in the segmented standard contour model, and obtain the deviation amount, deviation continuity length, deviation direction consistency, intra-segment reliability, and adjacent segment correlation consistency for each reference segment. The deviation amount of the outer boundary reference segment can be the lateral offset of the actual outer boundary relative to the standard outer boundary. The deviation amount of the groove reference segment can be the actual groove position deviation or groove depth deviation. The deviation amount of the chamfer reference segment can be the angle difference between the actual chamfer direction and the standard chamfer direction. The deviation amount of the arc transition reference segment can be the local distance difference between the actual arc transition line and the standard arc transition line. The deviation amount of the end face positioning reference segment can be the overall offset of the actual end face positioning line relative to the standard end face positioning line.

[0056] Deviation continuity length indicates the length of time a deviation continuously exists along the corresponding reference segment. If the deviation occurs only within a short distance and the preceding and following areas quickly return to normal, the deviation continuity length is short. If the deviation persists along the corresponding reference segment, the deviation continuity length is long. Deviation direction consistency indicates whether the deviation directions of multiple sampling points within the same reference segment are consistent. If multiple sampling points all shift to the same side, the deviation direction consistency is high. If the shift directions of multiple sampling points alternate, the deviation direction consistency is low. Intra-segment confidence can be obtained by combining the boundary confidence of multiple linear responses participating in registration within the reference segment. Adjacent segment correlation consistency indicates whether the deviation of the current reference segment and the deviations of adjacent reference segments conform to the geometric relationship of granite line processing.

[0057] If both the outer boundary reference segment and the groove reference segment show lateral offset in the same direction, and the trends of their offset changes are similar, then they are considered to be correlated and consistent. If the groove reference segment shows a groove depth deviation, and the chamfer reference segment or the arc transition reference segment shows a corresponding morphological change, then the groove deviation is considered to be correlated and consistent with the deviation of the adjacent segment. If the end face positioning reference segment experiences an overall offset, and multiple machining reference segments also shift in the same direction, then this state is considered to meet the correlation characteristics of single-piece clamping anomalies. If a reference segment shows a sudden deviation, but the adjacent reference segments do not show a corresponding change, then the deviation is more likely to originate from local edge chipping or image misidentification.

[0058] The deviation source identification unit is used to identify the source of deviation based on the reliability within a segment, the continuous length of the deviation, the consistency of the deviation direction, and the consistency of the association between adjacent segments. Deviation sources include at least texture interference, reflection interference, local edge chipping, single-piece clamping abnormalities, and equipment-related processing deviations.

[0059] When the deviation is located in a low-confidence region, lacks supporting evaluation metrics for uneven structure, and the deviation direction is consistent with or close to the main direction of the natural texture, the deviation source identification unit determines the deviation as texture interference. In this case, even if there is a clear linear response in the image, the linear response will not be used as a processing contour deviation to trigger calibration;

[0060] When the brightness of the deviation area decreases significantly in the cross-polarization image, and the contour response disappears or weakens significantly after the brightness decreases, the deviation source identification unit determines the deviation as reflective interference. In this case, the system can output a re-inspection prompt, or re-acquire the image after cleaning, purging, or changing the illumination angle;

[0061] When the deviation manifests as a short-distance abrupt change and does not extend continuously along the standard contour direction, while the consistency between adjacent segments is low, the deviation source identification unit determines the deviation as a local edge breakage. In this case, the system can record the location, length, and reference segment of the local defect, but will not use this local edge breakage as the basis for overall equipment calibration.

[0062] When the end face positioning reference segment shifts, and the outer boundary reference segment, groove reference segment, chamfer reference segment, or arc transition reference segment shifts in the same direction relative to their respective reference segments, the deviation source identification unit determines this state as a single-piece clamping abnormality. In this case, the system outputs a re-clamping prompt to avoid mistaking a single-piece clamping abnormality for a machining equipment reference shift;

[0063] When a certain machining reference segment or multiple related machining reference segments exhibit continuous unidirectional deviations within a high-confidence region, and the consistency of the deviation's continuous length, deviation direction, and adjacent segment correlation all meet the corresponding conditions, the deviation source identification unit determines the deviation as an equipment-related machining deviation. Equipment-related machining deviations may include clamping reference offset, conveying reference offset, machining depth offset, machining head posture offset, or machining path offset.

[0064] The calibration trigger qualification judgment unit generates calibration trigger qualifications based on the deviation source identification results. Calibration trigger qualifications indicate whether the deviation obtained from visual inspection allows for triggering processing calibration. If the deviation source is an equipment-related processing deviation, and the deviation amount of the corresponding reference segment exceeds the allowable deviation range, while the continuous length of the deviation reaches a set length, the consistency of the deviation direction reaches a set ratio, the intra-segment confidence reaches the corresponding threshold, and the correlation consistency between adjacent segments satisfies the corresponding processing relationship, then the calibration trigger qualification judgment unit generates a judgment result indicating that calibration trigger qualification is met. If the deviation source is texture interference, reflection interference, local edge chipping, or single-piece clamping abnormality, then the calibration trigger qualification judgment unit generates a judgment result indicating that calibration trigger qualification is not met.

[0065] Calibration trigger eligibility can employ a multi-condition admission rule. This rule includes reliability admission criteria, continuity admission criteria, directional consistency admission criteria, and adjacent segment association admission criteria. The reliability admission criteria determine whether the reliability within a segment reaches a threshold. The continuity admission criteria determine whether the continuous length of the deviation reaches a threshold. The directional consistency admission criteria determine whether the deviation directions of multiple sampling points within the same reference segment are consistent. The adjacent segment association admission criteria determine whether the current deviation conforms to the machining geometry relationship with the deviations of adjacent reference segments. Only when all of the above admission criteria are met will the system output a judgment result indicating that calibration trigger eligibility is met.

[0066] The calibration output unit outputs calibration parameter generation results or calibration parameter suppression results based on calibration trigger eligibility. When calibration trigger eligibility is met, the calibration output unit outputs the corresponding calibration parameter generation results based on the reference segment to which the deviation belongs. When the outer boundary reference segment corresponds to a lateral offset, it outputs the clamping position correction amount or the feed compensation amount. When the groove reference segment corresponds to a groove position deviation or groove depth deviation, it outputs the machining depth compensation amount. When the chamfer reference segment corresponds to an angle deviation, it outputs the machining head posture correction amount. When the arc transition reference segment corresponds to a curved surface transition deviation, it outputs the local machining path correction amount. When the source of the deviation is determined to be a single-piece clamping abnormality, such as an offset of the end face positioning reference segment causing multiple machining segments to move in the same direction, the calibration output unit outputs the calibration parameter suppression results, but does not output the equipment reference calibration parameters, only outputting a re-clamping prompt.

[0067] When calibration triggering is not qualified, the calibration output unit outputs the calibration parameter suppression result. For texture interference or reflection interference, a re-inspection prompt is output. For local edge chipping, a local defect record is output. For single-piece clamping abnormalities, a re-clamping prompt is output. The calibration parameter suppression result is used to prevent low-confidence visual contours, local defects, or clamping abnormalities from triggering incorrect calibration of the processing equipment.

[0068] The verification and update unit is used to re-acquire multi-path images of the same processing surface after calibration, and to verify them using the same segmented standard contour model, the same candidate search region, the same boundary confidence determination rules, and the same calibration trigger qualification judgment rules. The verification and update unit verifies the change in deviation before and after verification against the original calibration trigger qualification. If the verification deviation decreases in the direction indicated by the calibration parameters and enters the allowable deviation range, the verification and update unit outputs a calibration pass result. If the verification deviation is still in the high confidence region and the deviation direction remains the same, the verification and update unit outputs the updated equipment reference parameters. If the verification deviation shifts to the low confidence region, the deviation direction changes abruptly, or the verification result is inconsistent with the original calibration trigger qualification, the verification and update unit suppresses secondary calibration and outputs an abnormal verification result.

[0069] The system executes closed-loop processing based on the anomaly verification results: when the anomaly verification result indicates that the current visual feedback is unstable, the control conveying and positioning unit suspends the current line conveying and sends a pause processing command to the processing equipment; when the anomaly verification result indicates that the current line has local edge chipping or clamping abnormalities, the system marks the abnormal state of the current granite line and outputs a manual verification prompt; when the anomaly verification result occurs consecutively or exceeds a preset number, the alarm module issues an audible and visual alarm and triggers manual intervention. Through these measures, the system prevents the processing equipment from continuing to perform erroneous calibrations under unstable visual feedback.

[0070] The data recording unit is used to record the processing specifications, segmented standard contour model, acquired images, linear response evaluation quantity, boundary confidence, actual processed contour, segmented deviation results, deviation sources, calibration trigger qualifications, calibration output results, anomaly review results, and review closed-loop handling results for each granite line. The data recording unit can use the above data for quality traceability, equipment reference parameter adjustment, statistical analysis of texture characteristics of materials in the same batch, or optimization of subsequent detection thresholds.

[0071] This embodiment also provides a method for calibrating granite line processing based on industrial visual inspection. This method can be executed by the system described above, or by a control device capable of image acquisition, image processing, deviation judgment, and calibration result output.

[0072] S101. Obtain the target width, target height, groove position, groove depth, chamfer angle, arc transition radius, end face positioning datum, and allowable deviation range of the granite line to be inspected. Establish a segmented standard contour model based on the above processing specifications. The segmented standard contour model includes an outer boundary datum segment, a groove datum segment, a chamfer datum segment, an arc transition datum segment, and an end face positioning datum segment. Each datum segment is configured with a candidate search area, allowable deviation range, main deviation type, and calibration trigger condition.

[0073] S102. The granite line is fed into the inspection station via a conveyor roller. Lateral limiting components, clamping components, and end-face positioning components position and hold the granite line. Once the position sensor detects that the granite line has reached the preset inspection position, it generates a data acquisition trigger signal. If the granite line has not reached the preset inspection position, image acquisition is stopped or a positioning error message is output.

[0074] S103. A diffuse reference image is acquired using an industrial camera and a diffuse light source; a low-angle oblique light image is acquired using an industrial camera and a low-angle oblique light source; and a cross-polarization image is acquired using an industrial camera and a cross-polarization light source. The diffuse reference image reflects the surface texture and overall grayscale distribution of the granite lines. The low-angle oblique light image enhances the texture corresponding to the processing contours. The cross-polarization image reduces the interference of reflections and watermarks on the linear response.

[0075] S104. Based on the camera calibration parameters, end-face positioning reference, outer boundary stable section, or detection station coordinates, register the diffuse reference image, low-angle oblique light image, and cross-polarization image to the same processing surface coordinate system. After registration, the same processing surface area has corresponding coordinates in the three types of images. Considering that granite lines may experience minor vibrations during transport and positioning, after coordinate transformation, local fine registration can be performed based on the high-confidence contour response within the outer boundary stable section, end-face positioning line, or candidate search area. Local fine registration can employ sub-pixel level alignment methods based on contour mutual information, phase correlation, or feature point matching to reduce dynamic ghosting introduced by temporal flicker acquisition in multi-optical path images, ensuring that the same linear response achieves sub-pixel level correspondence in the diffuse reference image, low-angle oblique light image, and cross-polarization image.

[0076] S105. Map the segmented standard contour model to the machining surface coordinate system, and determine the candidate search areas based on the allowable deviation range and machining tolerance of each reference segment. The candidate search area for the outer boundary reference segment extends along the boundary normal. The candidate search area for the groove reference segment covers the groove edge, groove bottom line, and groove wall transition area. The candidate search area for the chamfer reference segment covers the chamfer edge and the chamfer surface junction area. The candidate search area for the arc transition reference segment covers the transition zones on both sides of the target arc. The candidate search area for the end face positioning reference segment covers the end face positioning line and adjacent positioning areas.

[0077] S106. Extract linear responses within each candidate search region. Linear responses are obtained through edge detection, grayscale gradient extraction, line segment connection, local extremum tracking, or curve fitting. Linear responses can be natural texture lines, reflective lines, groove lines, chamfer lines, curved transition lines, outer boundary lines, or locally chipped edges. In this step, the system does not directly identify linear responses as processing boundaries, but rather uses them as objects for subsequent evaluation.

[0078] S107. Calculate the texture continuity evaluation value of the linear response based on the diffuse reference map. Specifically, first, statistically analyze the main direction of the natural granite texture in the candidate search area or its neighboring area, and then calculate the angle between the linear response and the main direction of the natural texture, the continuous length of the linear response, and the gray-level gradient stability on both sides of the linear response. The main direction of the natural granite texture is statistically analyzed using texture analysis algorithms such as gray-level co-occurrence matrix, gradient direction histogram, or Fourier transform spectrum analysis, extracting the angle with the highest response energy in the candidate search area or its neighboring area as the main direction of the natural texture. If the angle between the linear response and the main direction of the natural texture is less than or equal to 15°, the continuous pixel length of the linear response is greater than 50 pixels, and the gray-level gradient change rate on both sides of the normal direction of the linear response is lower than a preset smoothness threshold, then the linear response is considered to have a high texture continuity evaluation value. If the angle between the linear response and the main direction of the natural texture is greater than 15°, and the gray-level gradient change rate on both sides of its normal direction is higher than a preset abrupt change threshold, then the linear response is considered not to conform to the characteristics of natural texture. The angle, length, and threshold mentioned above are only one option in this embodiment and can be adjusted according to the specifications of the granite lines, image resolution, and processing accuracy requirements.

[0079] S108. Calculate the concave-convex structure evaluation value of the linear response based on the low-angle oblique light image. Specifically, sampling bands are set on both sides of the linear response, and the gray-scale mean difference, shadow width, and gradient peak value of the sampling bands on both sides are calculated. If there is a stable brightness difference on both sides of the linear response, the shadow width is continuously distributed along the direction of the linear response, and the gradient peak value is concentrated near the linear response, then the concave-convex structure evaluation value is high. If the linear response is obvious only in the diffuse reference image, but there is no stable brightness change in the low-angle oblique light image, then the concave-convex structure evaluation value is low.

[0080] S109. Calculate the reflectivity attenuation evaluation based on the diffuse reference map and the cross-polarization map. Specifically, calculate the first average gray value of the corresponding region of the linear response in the diffuse reference map and the second average gray value in the cross-polarization map. Use the gray value difference or gray value ratio between the first and second average gray values ​​as a quantitative indicator of the brightness reduction. At the same time, determine whether the contour response of the linear response is retained after the brightness reduction. If the brightness reduction reaches the preset reflectivity attenuation threshold, and the contour response of the linear response disappears or is significantly weakened after the brightness reduction, the linear response is judged to have strong reflectivity interference characteristics. If the linear response still retains a continuous contour response after the brightness reduction, and there is concave-convex structure support in the low-angle oblique light map, the linear response is judged to be closer to the actual processing boundary.

[0081] S110. The boundary confidence of a linear response is determined by combining the texture continuity evaluation, the convexity / concave structure evaluation, and the reflectivity attenuation evaluation. If the linear response is located within the candidate search region of the corresponding benchmark segment, the convexity / concave structure evaluation reaches the structure threshold of the corresponding benchmark segment, and the contour response is still retained in the cross-polarization diagram, then its boundary confidence is increased. If the linear response has a high texture continuity evaluation, a low convexity / concave structure evaluation, and the contour response disappears after reflectivity attenuation, then its boundary confidence is decreased. Boundary confidence can be determined using a rule table method and a weighted scoring method, with a boundary confidence score. The calculation formula is expressed as:

[0082] ;

[0083] in, For evaluating concave-convex structures, For the degree of fit, As a measure of texture continuity, It is a reflective interference factor; The weighting coefficients are preset. When the boundary confidence score reaches the confidence threshold of the corresponding benchmark segment, the linear response is marked as a candidate line for processing boundary; when the boundary confidence score does not reach the confidence threshold of the corresponding benchmark segment, the linear response is marked as a texture interference line, a reflective interference line, or a line to be verified.

[0084] S111. Only linear responses whose boundary confidence meets the corresponding benchmark segment confidence threshold are used as candidate processing boundary lines. These candidate lines are then connected, fitted, and outlier removed to obtain the actual processing contour. The actual processing contour includes the actual outer boundary, actual groove line, actual chamfer line, actual arc transition line, and actual end face positioning line. For linear responses with low boundary confidence or abrupt changes in direction, the system does not use them as the primary basis for the actual processing contour, but retains their evaluation results.

[0085] S112. Register the actual outer boundary with the outer boundary reference segment, register the actual groove line with the groove line reference segment, register the actual chamfer line with the chamfer reference segment, register the actual arc transition line with the arc transition reference segment, and register the actual end face positioning line with the end face positioning reference segment. After registration, the system obtains the deviation amount, continuous length of deviation, consistency of deviation direction, intra-segment reliability, and consistency of association between adjacent segments for each reference segment.

[0086] S113. Identify the source of deviation based on intra-segment confidence, continuous length of deviation, consistency of deviation direction, and consistency of association between adjacent segments. If the deviation is located in a low-confidence area, lacks support from concave and convex structures, and extends along the natural texture direction, it is determined to be texture interference. If the brightness of the deviation area decreases significantly and the contour response disappears in the cross-polarization diagram, it is determined to be reflective interference. If the deviation manifests as a short-distance abrupt change and is not continuous along the standard contour direction, it is determined to be local edge chipping. If the end face positioning reference segment shifts, and multiple processing reference segments move in the same direction as a whole, it is determined to be a single-piece clamping anomaly. If there are continuous deviations in the same direction within a high-confidence area, and adjacent reference segments satisfy the corresponding processing relationship, it is determined to be an equipment-related processing deviation.

[0087] S114. When the source of the deviation is equipment-related processing deviation, and the deviation amount exceeds the allowable deviation range, the continuous length of the deviation reaches the set length, the consistency of the deviation direction reaches the set ratio, the intra-segment reliability reaches the corresponding threshold, and the correlation consistency between adjacent segments meets the corresponding processing relationship, the system generates a result that is qualified to trigger calibration. When the source of the deviation is texture interference, reflection interference, local edge chipping, or single-piece clamping abnormality, the system generates a result that is not qualified to trigger calibration.

[0088] S115. When the system generates a result that qualifies for calibration triggering, it outputs the corresponding calibration parameters based on the reference segment to which the deviation belongs. The lateral offset of the outer boundary reference segment corresponds to the clamping position correction or feed compensation. The groove position deviation or groove depth deviation of the groove reference segment corresponds to the machining depth compensation. The angular deviation of the chamfer reference segment corresponds to the machining head posture correction. The curved transition deviation of the arc transition reference segment corresponds to the local machining path correction. When the system generates a result that does not qualify for calibration triggering, it outputs the calibration parameter suppression result. Texture interference or reflection interference corresponds to a re-inspection prompt, local edge chipping corresponds to a local defect record, and single-piece clamping abnormality corresponds to a re-clamping prompt. When the source of the deviation is determined to be a single-piece clamping abnormality, such as the end face positioning reference segment shifting and causing multiple machining segments to move in the same direction, the system does not output the equipment reference calibration parameters.

[0089] S116. If the processing equipment performs calibration according to the calibration parameters, the system will re-acquire the diffuse reference map, low-angle oblique light map, and cross-polarization map of the same processing surface of the granite line after calibration, and will use the same segmented standard contour model, the same candidate search area, the same boundary confidence determination rule, and the same calibration trigger qualification judgment rule for verification. The system will verify the change in deviation before and after verification against the original calibration trigger qualification. If the verification deviation converges along the direction of the calibration parameters and enters the allowable deviation range, the calibration pass result will be output. If the verification deviation is still in the high confidence region and continues to change in the same direction, the equipment reference parameter update result will be output. If the verification deviation shifts to the low confidence region, the deviation direction jumps, or the verification result is inconsistent with the original calibration trigger qualification, the secondary calibration will be suppressed, and the abnormal verification result will be output. The system executes closed-loop processing based on the anomaly verification results: when the anomaly verification result indicates that the current visual feedback is unstable, the control conveying and positioning unit suspends the current line conveying and sends a pause processing command to the processing equipment; when the anomaly verification result indicates that the current line has local edge chipping or clamping abnormalities, the system marks the abnormal state of the current granite line and outputs a manual verification prompt; when the anomaly verification result occurs consecutively or exceeds a preset number, the alarm module issues an audible and visual alarm and triggers manual intervention. Through these measures, the system prevents the processing equipment from continuing to perform erroneous calibrations under unstable visual feedback.

[0090] Using the above method, the linear responses in the granite line processing image undergo sequential processes before participating in calibration judgment: candidate search region constraint, multi-path evaluation, boundary confidence determination, actual processing contour extraction, segmented deviation analysis, deviation source identification, and calibration trigger qualification judgment. This method ensures that the visual inspection result only triggers calibration parameter generation when it is determined to be a high-confidence equipment-related processing deviation. It outputs calibration parameter suppression results when there is texture interference, reflection interference, local edge chipping, or single-piece clamping abnormalities, thereby reducing the risk of erroneous calibration and improving the stability of granite line processing calibration.

[0091] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A method for calibrating granite line processing based on industrial visual inspection, characterized in that, include: Obtain the processing specifications of granite lines and establish a segmented standard contour model. The segmented standard contour model includes an outer boundary reference segment, a groove reference segment, a chamfer reference segment, an arc transition reference segment, and an end face positioning reference segment. Each reference segment is configured with a candidate search area, a deviation type, and a calibration trigger condition. The diffuse reference image, low-angle oblique light image, and cross-polarization image of the processing surface are collected as three types of images, and the three types of images are registered to the coordinate system of the processing surface. Linear responses are extracted within the candidate search area. Based on the three types of images, texture continuity evaluation, concavity / convexity structure evaluation, and reflectivity attenuation evaluation are determined respectively. The boundary confidence of the linear responses is also determined. The linear response whose boundary confidence satisfies the corresponding benchmark segment confidence threshold is used as the processing boundary candidate line. The actual processing contour is extracted and the actual processing contour is registered with the segmented standard contour model to obtain the deviation amount, deviation continuity length, deviation direction consistency, intra-segment confidence and adjacent segment association consistency. Based on the intra-segment confidence, the deviation continuity length, the deviation direction consistency, and the adjacent segment correlation consistency, the deviation source is identified and calibration trigger qualifications are generated. When the source of the deviation is identified as a machine-related processing deviation and the calibration triggering condition is met, the calibration parameter generation result is output; when the source of the deviation is identified as texture interference, reflection interference, local edge chipping, or single-piece clamping abnormality, the calibration parameter suppression result is output. After the processing equipment performs calibration based on the calibration parameters, the three types of images are re-acquired, and based on the segmented standard contour model and the candidate search area, the extraction of the actual processing contour, the identification of the source of deviation, and the judgment of the calibration trigger qualification are performed again for verification. The calibration pass result, the equipment reference parameter update result, or the abnormal verification result are output.

2. The method for calibrating granite line processing based on industrial visual inspection according to claim 1, characterized in that, The candidate search region is determined based on the contour position and allowable deviation range of each reference segment in the coordinate system of the machining surface, and is extended along the normal direction, contour extension direction or surface transition direction of the corresponding reference segment; the calibration triggering conditions include boundary confidence conditions, deviation continuity length conditions, deviation direction consistency conditions and adjacent segment association conditions.

3. The method for calibrating granite line processing based on industrial visual inspection according to claim 1, characterized in that, Registering the three types of images to the processing surface coordinate system includes: performing coordinate transformation on the three types of images based on camera calibration parameters, end face positioning reference, outer boundary stable segment or image feature points, and performing local fine registration on the three types of images based on the outer boundary stable segment, end face positioning line or high confidence contour response.

4. The method for calibrating granite line processing based on industrial visual inspection according to claim 1, characterized in that, The texture continuity evaluation metric is determined based on the main direction of the natural texture in the candidate search region or its neighboring region, the angle between the linear response and the main direction of the natural texture, the continuous length of the linear response, and the gray-level gradient change rate on both sides of the linear response.

5. The method for calibrating granite line processing based on industrial visual inspection according to claim 1, characterized in that, The evaluation metric for the concave-convex structure is determined based on the difference in grayscale mean, shadow width, and gradient peak value of the sampling bands on both sides of the linear response in the low-angle oblique light image; the evaluation metric for reflective attenuation is determined based on the grayscale difference or grayscale ratio of the corresponding region of the linear response in the diffuse reference image and the cross-polarization image, as well as the contour response retention status.

6. The method for calibrating granite line processing based on industrial visual inspection according to claim 1, characterized in that, The boundary confidence level is determined based on the convex-concave structure evaluation value, the degree of alignment between the linear response and the corresponding reference segment, the texture continuity evaluation value, and the reflective interference factor; wherein, the convex-concave structure evaluation value and the degree of alignment are positive factors, and the texture continuity evaluation value and the reflective interference factor are negative factors.

7. The method for calibrating granite line processing based on industrial visual inspection according to claim 6, characterized in that, When registering the actual machining contour with the segmented standard contour model, the actual outer boundary, actual groove line, actual chamfer line, actual arc transition line and actual end face positioning line are respectively registered with the corresponding reference segment, and the intra-segment confidence is determined based on the boundary confidence of the machining boundary candidate lines participating in the registration.

8. The method for calibrating granite line processing based on industrial visual inspection according to claim 1, characterized in that, The adjacent segment correlation consistency is determined based on the deviation direction, deviation trend, and machining geometry relationship between adjacent reference segments; when multiple reference segments are offset in the same direction as a whole, the adjacent segment correlation consistency is used to identify single-piece clamping abnormalities; when the deviation of adjacent reference segments meets the corresponding machining geometry relationship, the adjacent segment correlation consistency is used to identify equipment-related machining deviations.

9. The method for calibrating granite line processing based on industrial visual inspection according to claim 1, characterized in that, The verification includes: when the verification deviation converges along the direction pointed to by the calibration parameter generation result and enters the allowable deviation range, outputting the calibration pass result; when the verification deviation is still located in the region within the segment with a confidence level higher than a preset threshold and maintains the same direction of change, outputting the device reference parameter update result; when the verification deviation shifts to the region within the segment with a confidence level lower than a preset threshold, the deviation direction changes abruptly, or the verification result is inconsistent with the calibration trigger qualification, outputting the abnormal verification result and suppressing secondary calibration.

10. A calibration system for granite line processing based on industrial visual inspection, characterized in that, A method for calibrating granite line processing based on industrial visual inspection, as described in any one of claims 1 to 9, comprises: a standard contour modeling unit, an image acquisition unit, an image registration unit, a linear response extraction and evaluation unit, a boundary confidence determination unit, an actual processing contour extraction unit, a segmented deviation analysis unit, a deviation source identification unit, a calibration trigger qualification judgment unit, a calibration output unit, and a verification and update unit; wherein, the standard contour modeling unit is used to establish the segmented standard contour model, the image acquisition unit is used to acquire the three types of images, the image registration unit is used to register the three types of images, and the linear response extraction and evaluation unit is used to extract the linear response and determine the texture. The system includes a continuity evaluation quantity, an uneven structure evaluation quantity, and a reflectivity attenuation evaluation quantity. The boundary confidence determination unit is used to determine the boundary confidence. The actual processing contour extraction unit is used to extract the actual processing contour. The segmented deviation analysis unit is used to obtain the deviation amount, deviation continuity length, deviation direction consistency, intra-segment confidence, and adjacent segment correlation consistency. The deviation source identification unit is used to identify the deviation source. The calibration trigger qualification judgment unit is used to generate the calibration trigger qualification. The calibration output unit is used to output the calibration parameter generation result or calibration parameter suppression result. The verification update unit is used to output the calibration pass result, equipment reference parameter update result, or anomaly verification result.

Images