A printing color difference defect detection method and device based on an artificial intelligence system

Abstract

The present application relates to a kind of printing color difference defect detection method based on artificial intelligence system, it is related to printing quality detection technical field, it includes gathering initial image to be inspected and incident light beam information;With initial image to be inspected and template image determine contrast image information and first space coordinate;According to contrast image information and incident light beam information output comparison result, comparison result includes abnormal state;Based on abnormal state, contrast image information is as first light field abnormal feature;Determine spatial incidence corresponding relationship in combination;According to spatial incidence corresponding relationship and the physical mapping of spherical coordinate determines second space coordinate;Second light field abnormal feature is generated;First light field abnormal feature and second light field abnormal feature are compared, and physical reflection difference value is obtained;If physical reflection difference value is greater than color difference tolerance threshold, then output color difference defect detection result.The present application has the effect of reducing detection false alarm rate.

Description

Technical Field

[0001] This invention relates to the field of printing quality inspection technology, and in particular to a method and apparatus for detecting printing color difference defects based on an artificial intelligence system. Background Technology

[0002] In printing quality inspection technology, printing quality inspection is a core quality control link on the production line used to assess printing color difference defects, identify printing flaws and color deviations.

[0003] Current traditional methods for detecting color difference defects in printing use colorimeters or basic machine vision equipment. Typically, an industrial camera is used with a fixed ring light source or strobe light source as illumination to capture images, which are then compared using preset color thresholds to determine pixel overlap. This method meets basic automated defect screening needs and provides quality control.

[0004] When using traditional inspection equipment to inspect printed parts, the light-emitting components of these devices often employ constant global illumination. This fixed lighting easily produces abnormal light field characteristics on the inspected surface, such as localized reflections, shadows, or stray light interference. Because the light-emitting components can only provide constant global illumination, traditional image processing systems cannot eliminate interference at its physical source. This easily leads to misjudging normal physical light field abnormalities as genuine color difference defects, resulting in an extremely high false alarm rate. Summary of the Invention

[0005] To reduce the false alarm rate, this invention provides a method and apparatus for detecting printing color difference defects based on an artificial intelligence system.

[0006] In a first aspect, the present invention provides a printing color difference defect detection method based on an artificial intelligence system, which adopts the following technical solution: A method for detecting color difference defects in printing based on an artificial intelligence system includes: Acquire the initial image of the printed part to be inspected and the information of the incident light beam; The initial image to be inspected is combined with the preset template image to determine the comparison image information and the first spatial coordinates through image difference; the comparison result is output by spatial ray tracing based on the comparison image information and the incident beam information, and the comparison result includes abnormal states; Based on the abnormal state, the information of the comparison image is used as the first light field abnormality feature; the spatial incidence correspondence is determined by combining the incident beam information and the first spatial coordinate correspondence. The second spatial coordinates of the preset standard part surface are determined based on the physical mapping between the spatial incident correspondence and the preset spherical coordinates; The incident beam is physically transmitted to the second spatial coordinates at an equivalent incident angle to generate a second optical field anomalous feature; The first light field anomaly feature is compared with the second light field anomaly feature to obtain the physical reflection difference value; The physical reflection difference value is compared with the preset color difference tolerance threshold. If the physical reflection difference value is greater than the preset color difference tolerance threshold, it is determined that there is a printing color difference defect and the color difference defect detection result is output.

[0007] By employing the above technical solution, based on image difference and incident beam information, the first light field anomaly features and spatial incidence correspondence are extracted through spatial ray tracing. The second spatial coordinates are located on the standard part through physical mapping of spherical coordinates, and the second light field anomaly features are obtained by equivalent angle lighting. This physically reproduces a consistent local lighting environment. Finally, the physical reflection difference values ​​of the two sets of features are compared, and the detection result is output based on the color difference tolerance threshold. Through a cross-validation mechanism that eliminates interference from normal reflections, true minute color differences are accurately identified, and the influence of external stray light is reduced. This reduces the possibility of misjudging normal physical light field anomalies as true color difference defects, thus lowering the false alarm rate.

[0008] Optionally, acquiring the initial image of the printed workpiece to be inspected and the incident beam information includes: Under the preset indoor equipment coordinate system, the initial image of the printed parts to be inspected is obtained by image acquisition. Based on the initial image to be inspected, the pose parameters of the printed part to be inspected are determined. Based on the pose parameters, the bright reflection region in the initial image to be inspected is located, and the optical features contained in the bright reflection region are extracted; the optical features include the shape and contour of the light spot and the brightness and darkness distribution features; The shape and outline of the light spot are matched and compared with a preset geometric deformation mapping table to determine the projection direction of the light source; The distance to the light source is determined by matching and comparing the characteristics of the light and dark distribution with a preset light intensity attenuation benchmark library. By combining the direction of light source projection and the distance to the light source, the absolute physical coordinates of the light source device currently projected onto the surface of the printed workpiece to be inspected can be identified; Call the pre-built 3D spatial calibration data to map the absolute physical coordinates to the absolute physical coordinates of the current light source device; Absolute physical coordinates are used as information for the incident beam.

[0009] By adopting the above technical solution, the shape and brightness features of the bright spot are extracted based on the pose parameters of the part under inspection. The angle and distance coordinates of the light source are locked respectively. Then, the absolute physical coordinates are identified and the calibration data is called. The absolute physical coordinates are mapped as the incident beam information, thereby reducing the situation where abnormal physical light field features are misjudged as color difference defects due to the missing position of the light source.

[0010] Optionally, based on the anomalous state, the specific steps for using the contrast image information as the first light field anomalous feature include: The difference connected components are determined by pixel-level image difference operations by combining the initial image to be inspected and the preset template image; By combining the incident beam information and pose parameters, the theoretical reflection coordinates are calculated through spatial ray tracing. The spatial positions of the differential connected components are compared with the theoretical reflection coordinates, and the overlap verification results are output. When the overlap verification result is less than or equal to the preset threshold, it is determined to be in an abnormal state, and the corresponding comparison image information is used as the first light field abnormal feature based on the abnormal state.

[0011] By adopting the above technical solution, the difference connected components are determined based on image difference operation. The theoretical reflection coordinates are calculated by ray tracing in combination with the incident beam and pose parameters. The spatial positions of the two are then compared. When the overlap is greater than a preset threshold, it is extracted as the first light field anomalous feature, thereby realizing the accurate identification of the physical source of the difference in two-dimensional images.

[0012] Optional methods for determining spatial incident correspondence include: Based on the incident beam information, a three-dimensional incident ray vector is constructed by performing vector difference calculation using the first spatial coordinates. The surface normal vector of the printed part under inspection at the first spatial coordinate is calculated by combining the initial image to be inspected with the pose parameters. Calculate the spatial geometric angle between the three-dimensional incident ray vector and the surface normal vector to obtain the incident angle and azimuth angle of the incident beam at the first spatial coordinate. The spatial incident correspondence is determined by combining the incident angle and the azimuth angle.

[0013] By adopting the above technical solution, a three-dimensional incident ray vector is constructed based on the incident beam information and the first spatial coordinates, and the surface normal vector of the workpiece at the coordinates is calculated; the spatial geometric angle between the two vectors is calculated to extract the incident angle and azimuth angle, and then the spatial incident correspondence is determined, which facilitates the equivalent reproduction of the lighting environment on the standard part.

[0014] Optionally, methods for determining the second spatial coordinates include: By combining the incident beam information and the first spatial coordinates, the equivalent mapping radius is determined by calculating the spatial Euclidean distance. A virtual spherical coordinate system is established using the reference point of the preset standard part surface as the origin and the equivalent mapping radius. Extract the incident angle and azimuth angle from the spatial incident correspondence and use them as the target polar angle and target azimuth angle; By combining the equivalent mapping radius, the target polar angle, and the target azimuth angle, a transformation calculation from spherical coordinates to rectangular coordinates is performed in this virtual spherical coordinate system to obtain the theoretical rectangular coordinates; The theoretical rectangular coordinates are projected onto the surface of a preset standard part to obtain the actual landing point, and the second spatial coordinates are determined based on the actual landing point.

[0015] By adopting the above technical solution, the equivalent mapping radius is calculated based on the incident beam and the first spatial coordinates, and a virtual spherical coordinate system is established; the incident angle and azimuth angle are extracted and converted into theoretical rectangular coordinates, and finally projected onto the surface of the standard part to determine the actual landing point and the second spatial coordinates, thereby correcting the spatial mapping deviation generated when projecting across the surface.

[0016] Optionally, methods for generating the second optical field anomalous feature include: By combining the spatial incident correspondence with the second spatial coordinates, a three-dimensional spatial inverse deduction is performed to construct a three-dimensional target illumination vector; Determine spatial lighting control commands based on three-dimensional target lighting vectors; Drive the preset directional reproduction light source according to the spatial lighting control command, and output a directional beam of light to project to the second spatial coordinate; During the directional beam projection, synchronous image acquisition is performed on the surface of a preset standard part to obtain an initial reference image; Based on the second spatial coordinates, locate and determine the corresponding physical light spot reflection characteristics in the initial reference image; The physical light spot reflection characteristics are used as the second light field anomaly characteristics.

[0017] By employing the above technical solution, a three-dimensional illumination vector is deduced based on the spatial incidence relationship and the second spatial coordinates, and instructions are generated to drive the ambient light source to project a directional beam and simultaneously acquire a reference image; then, the physical light spot reflection characteristics in the image are located as the second light field anomaly characteristics. This allows for the reproduction of a local illumination environment identical to the anomaly area to be detected.

[0018] Optional methods for determining the results of printing color difference defect detection include: Optical parameters are determined by combining the first and second light field anomaly features. These optical parameters include multi-channel color intensity and reflectance brightness components. Perform difference matrix calculations on the multi-channel color intensity and reflectance components to obtain the spatial spectral deviation. The difference in physical reflectance is determined based on the spatial spectral deviation. The physical reflection difference value is compared with the preset color difference tolerance threshold. When the physical reflection difference value is greater than the preset color difference tolerance threshold, it is determined that there is a printing color difference defect, and the color difference defect detection result is output.

[0019] By employing the above technical solution, multi-channel color and reflectance brightness components are extracted based on the first and second light field anomaly features. A difference matrix calculation is then performed to obtain the spatial spectral deviation. This deviation is then used to confirm the physical reflectance difference value, which is compared with a tolerance threshold to output the detection result. This quantitative calculation of the spectrum and brightness improves the reliability of the output detection result.

[0020] Optional methods for setting the color difference tolerance threshold include: Obtain the basic color parameters and surface roughness parameters of the printed part to be inspected; Based on the surface base color parameters and surface roughness parameters, match the corresponding initial color difference tolerance; The preset ambient light source is turned off, and image acquisition is performed in a dark environment to obtain a pure black baseline image. Calculate the average deviation range of pixel color values ​​in a pure black reference image and use it as a device background noise parameter; The actual color difference tolerance is determined by summing the background noise parameters of the equipment with the initial color difference tolerance. The actual color difference tolerance is used as the color difference tolerance threshold.

[0021] By adopting the above technical solution, based on the initial color difference tolerance matching the surface base color and roughness, a pure black image is then acquired in a dark environment to calculate the device background noise parameter; finally, this noise parameter is used to perform numerical compensation on the initial tolerance, outputting the color difference tolerance threshold. This achieves dynamic and accurate calibration of the color difference tolerance threshold.

[0022] Optional methods for determining actual color difference tolerance include: Get the device operating temperature during the current image acquisition process; Calculate the temperature difference between the operating temperature of the equipment and the preset equipment calibration temperature; Determine whether the temperature difference is less than the preset temperature drift compensation threshold; When the temperature difference is less than the preset temperature drift compensation threshold, the background noise parameter of the equipment and the initial color difference tolerance are directly added together to determine the actual color difference tolerance. When the temperature difference is not less than the preset temperature drift compensation threshold, the corresponding amplification factor is determined based on the temperature difference, and the amplification factor is used as the hardware error relaxation factor. The inherent error tolerance of the equipment is determined by multiplying the background noise parameters of the equipment with the hardware error relaxation factor. The actual color difference tolerance is determined by summing the inherent error tolerance of the equipment with the initial color difference tolerance.

[0023] By employing the above technical solution, the difference between the operating temperature and the calibrated temperature of the computing device is obtained. Based on the comparison between this difference and the temperature drift compensation threshold, a hardware error relaxation coefficient is introduced. Finally, combined with background noise and the initial color difference tolerance, the actual color difference tolerance is calculated. This effectively reduces the surge in noise in the underlying image and improves the ability of the actual color difference tolerance to correct for hardware temperature rise.

[0024] Secondly, the present invention provides a printing color difference defect detection device for an artificial intelligence system, which adopts the following technical solution: A printing color difference defect detection device based on an artificial intelligence system includes a memory and a processor. The memory stores a computer program that can be loaded and executed by the processor, as described in any of the first aspects, for a printing color difference defect detection method based on an artificial intelligence system.

[0025] By adopting the above technical solution, through the collaborative interaction of the processor and memory, the core algorithm programs such as spatial ray tracing, three-dimensional coordinate mapping and physical lighting reproduction are automatically loaded and executed. The multi-dimensional light field cross-verification mechanism is solidified into the underlying operating logic of the device, thereby building a detection closed loop to prevent false judgment at the physical hardware level. Finally, the color difference defect detection result is output, which reduces the drawback of traditional vision devices lacking a real physical verification mechanism, which easily misjudge normal physical light field abnormal features as real color difference defects, resulting in a very high false alarm rate.

[0026] In summary, the present invention has at least one of the following beneficial technical effects: 1. By physically reproducing the same local lighting environment on standard parts through spatial ray tracing and spherical coordinate mapping, a cross-validation mechanism was constructed, which realized the essential distinction between physical reflection and true color difference. This effectively reduced the large number of false alarms caused by workshop stray light and material reflection interference in traditional vision systems, thereby reducing the situation of misjudging normal physical light field abnormal features as true color difference defects and reducing the false alarm rate. 2. By monitoring the equipment operating temperature and background noise parameters, a temperature drift compensation threshold and a hardware error relaxation coefficient are introduced to numerically compensate for the color difference threshold. This effectively avoids the detection fluctuations caused by the surge in noise in the underlying image due to continuous heat generation during equipment operation, reduces the problem of increased misjudgment caused by equipment temperature drift in industrial environments, and improves the long-term stability of the system under high-load operation. 3. Based on the surface roughness of the printed parts and the basic color, the color difference tolerance is adaptively matched, and the calculation of multi-channel color intensity and reflectance brightness is performed to obtain the spatial spectral deviation. The visual color difference is converted into an optical quantitative index, which realizes the accurate capture of minute color difference defects and improves the high stability of automated quality control on modern high-end printing production lines. Attached Figure Description

[0027] Figure 1 This is a flowchart of a printing color difference defect detection method based on an artificial intelligence system. Detailed Implementation

[0028] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0029] A method for detecting color difference defects in printing based on an artificial intelligence system first acquires an initial image of the printed part to be inspected and information about the incident light beam. Then, the initial image to be inspected is compared with a preset template image to determine the comparison image information, and spatial ray tracing is used to verify and locate the first spatial coordinates. Based on the abnormal state, the comparison image information is directly extracted as a first light field anomaly feature. Simultaneously, the spatial incidence correspondence is determined by calculating and deducing the incident light beam information and the first spatial coordinates. Based on this, the system accurately maps the anomaly landing point to a preset standard part surface according to a preset spherical coordinate physical mapping algorithm, establishing a second spatial coordinate. Next, the incident light beam is physically transmitted to the second spatial coordinate at an equivalent incident angle, thereby exciting a second light field anomaly feature on the standard part surface for equivalent reference. The first and second light field anomaly features are then compared using a low-level matrix comparison to obtain a pure physical reflection difference value. Finally, the physical reflection difference value is rigorously compared with a preset color difference tolerance threshold to effectively filter out physical reflection interference caused by complex illumination angles, thereby reducing the false alarm rate.

[0030] Reference Figure 1 This invention discloses a printing color difference defect detection method based on an artificial intelligence system, which includes: S100: Acquires the initial image of the printed material to be inspected and the information of the incident beam.

[0031] The initial image to be inspected refers to the basic image data acquired at the beginning of the inspection process, reflecting the initial optical state of the surface of the printed workpiece to be inspected. The incident beam information refers to the absolute spatial coordinate information of the external light source that illuminates the surface of the printed workpiece to be inspected and causes it to produce high-brightness reflection.

[0032] By triggering an industrial camera set at the beginning of the production line to perform a shooting action, the initial image to be inspected is directly acquired. Then, the light spot contour and brightness distribution characteristics in the high-brightness reflection area of ​​the initial image to be inspected are calculated in reverse to obtain the absolute coordinate parameters of the light source and use them as the information of the incident beam.

[0033] High-brightness reflection areas refer to local image features in the initial image to be inspected that exhibit grayscale saturation or surges due to direct illumination from an external light source. Reverse deduction calculation refers to a mathematical algorithm that derives the position of a three-dimensional light source based on the physical shape of the light spot.

[0034] S101: Combine the initial image to be inspected with the preset template image to determine the comparison image information and the first spatial coordinates through image difference; output the comparison result by spatial ray tracing based on the comparison image information and the incident beam information.

[0035] The template image refers to the reference image data used to characterize the printed part under ideal conditions, obtained by performing a pre-storage operation before the inspection task begins. The comparison image information refers to the difference feature information between the initial image to be inspected and the preset template image. The first spatial coordinates refer to the absolute three-dimensional position coordinates of the suspected abnormal area on the surface of the printed part to be inspected.

[0036] The comparison result refers to the qualitative judgment output after optical imaging verification based on the comparison image information. The comparison result includes abnormal and normal states. An abnormal state refers to the data characterization of a local area on the surface of the printed part under inspection that actually deviates from the preset standard benchmark in optical characteristics after excluding reasonable physical reflection interference. A normal state refers to the defect-free data characterization of the surface difference area in the comparison image information that has been confirmed by spatial ray tracing as reasonable physical reflection interference, thus excluding real printing defects.

[0037] By performing differential calculations on the initial image to be inspected and the preset template image, comparison image information is obtained. Then, the geometric center pixel or boundary centroid of the differential connected domain in the comparison image information is extracted as the position reference to determine the first spatial coordinates. The actual abnormal position represented by the first spatial coordinates and the theoretical reflection characteristics derived based on the incident beam information are compared with the overlap degree of the algorithm results of spatial ray tracing. The Boolean judgment result output by the overlap degree verification is used as the comparison result.

[0038] Reflection characteristics refer to the theoretical spatial coordinates of the expected bright reflections produced by an external light beam on the surface of the printed workpiece under inspection, derived from the laws of optical reflection. Spatial ray tracing is a geometric algorithm that uses the absolute spatial coordinates of the incident light beam and the three-dimensional surface normal vector of the printed workpiece under inspection, based on the physical law that the angle of incidence equals the angle of reflection, to calculate the expected landing point of the light beam. Overlap verification is the process of calculating the spatial Euclidean distance or intersection-union ratio between the actual abnormal location and the theoretical reflection location to determine whether the actual abnormal patch is caused by purely physical reflection.

[0039] S102: Based on the abnormal state, the information of the comparison image is used as the first light field abnormal feature, and the spatial incident correspondence is determined by combining the incident beam information and the first spatial coordinate correspondence.

[0040] Among them, the first light field anomaly feature refers to the specific abnormal image feature data characterizing the actual printing defects on the surface of the printed workpiece under inspection. The first light field anomaly feature specifically includes visual features such as actual abnormal patches caused by local ink loss, exposed underlying material, or uneven color development. The spatial incident correspondence refers to the three-dimensional geometric direction parameters between the position of the incident light beam and the surface of the printed workpiece under inspection.

[0041] Based on the logical premise of establishing an abnormal state, the corresponding abnormal connected component image block in the comparison image information is extracted and directly assigned values ​​to determine the first light field anomaly feature. The spatial incidence correspondence is obtained by performing spatial vector difference calculation between the coordinates contained in the incident beam information and the first spatial coordinates.

[0042] S103: Determine the second spatial coordinates of the preset standard part surface based on the physical mapping between the spatial incident correspondence and the preset spherical coordinates.

[0043] Among them, the physical mapping of spherical coordinates refers to the coordinate transformation rule that converts spatial geometric angles into coordinates of physical points on a three-dimensional surface. Using spherical mapping can extremely accurately match the physical nature of light rays propagating radially in three-dimensional space, thereby achieving a lossless equivalent transfer of lighting angles.

[0044] A standard part refers to a defect-free physical reference object that belongs to the same batch, material, surface roughness, and base color as the printed part to be inspected. It is obtained by pre-placing the defect-free physical reference object in the inspection device and completing the calibration. The second spatial coordinates refer to the position coordinates of the actual corresponding points formed on the surface of the standard part after the original light field is physically mapped and cloned.

[0045] The second spatial coordinates are obtained by combining the mapping parameters to perform geometric deduction of the theoretical landing point and extending along a specific ray incident direction to converge to the surface of the actual physical entity through three-dimensional geometric calculation.

[0046] S104: Physically transmit the incident beam to the second spatial coordinates at an equivalent incident angle to generate a second optical field anomalous feature.

[0047] The equivalent incident angle refers to a three-dimensional lighting angle that maintains absolute geometric consistency with the actual light-receiving angle on the surface of the printed workpiece under inspection. Physical transmission refers to the action of projecting a real directional light beam onto the target's physical coordinates without relying on pure software image rendering. Second light field anomaly characteristics refer to the light spot reflection characteristics generated at the second spatial coordinates after equivalent physical lighting. A directional reproduction light source refers to a high-precision, controlled hardware entity, independent of conventional background lighting, specifically designed to accurately reproduce a light beam incident at a specific angle in three-dimensional space. Examples of directional reproduction light sources include directional spotlights with two-axis servo-driven pan-tilt heads, or programmable digital micromirror (DMD) projection devices.

[0048] The equivalent incident angle is obtained by extracting the actual spatial tilt and deflection parameters of the surface of the printed part to be inspected. Then, a directional reproduction light source arranged in the inspection environment is driven to output a directional beam for physical transmission. The directional beam is physically transmitted to the second spatial coordinate of the preset standard part surface at the equivalent incident angle, exciting the solid optical response to obtain the light spot feature, which is used as the second light field anomaly feature.

[0049] Ambient light sources refer to physical external lighting devices that emit controlled light, perform physical conduction, and output light beams.

[0050] S105: Compare the first light field anomaly feature with the second light field anomaly feature to obtain the physical reflection difference value.

[0051] Among them, the physical reflection difference value refers to the specific numerical difference in the pixel feature space of the abnormal features of the printed part under inspection and the reflection features of the standard part under the same physical lighting.

[0052] The physical reflection difference value is obtained by performing matrix operations and comparisons on the first and second light field anomaly features.

[0053] Matrix operations and comparisons refer to the underlying mathematical calculation process of subtracting the image data matrix contained in the first light field anomaly feature from the reference image data matrix contained in the second light field anomaly feature and taking the absolute value.

[0054] S106: Compare the physical reflection difference value with the preset color difference tolerance threshold. If the physical reflection difference value is greater than the preset color difference tolerance threshold, it is determined that there is a printing color difference defect and the color difference defect detection result is output.

[0055] The color difference tolerance threshold refers to the maximum allowable fluctuation of normal physical color difference by the system. Printing color difference defects refer to actual printing flaws after excluding normal physical reflection interference. Actual printing flaws include uneven ink distribution or missed printing. The color difference defect detection result is the final judgment information that clearly indicates the presence of actual printing flaws in the inspected printed part and the specific degree of deviation.

[0056] The color difference tolerance threshold is determined by combining the material parameters of the part under inspection with the equipment's noise floor to perform numerical compensation calculations and pre-store the data. The physical reflection difference value is compared with the preset color difference tolerance threshold. When the physical reflection difference value is significantly greater than the preset threshold, a qualitative judgment of a genuine defect is directly generated, establishing a printing color difference defect. Finally, the established printing color difference defect and its corresponding physical reflection difference value are integrated and output to generate the color difference defect detection result.

[0057] Numerical compensation calculation refers to the underlying mathematical process of correcting and relaxing the initial allowable color difference value of the material parameters of the workpiece under inspection based on the equipment noise floor.

[0058] To further ensure the rationality of the printing color difference defect detection method of the artificial intelligence system, it is necessary to analyze and calculate the initial image of the printed part to be inspected and the incident light beam information separately. The specific steps are explained in detail below.

[0059] The initial image of the printed part to be inspected and the information of the incident light beam are acquired in the following steps: S200: Under the preset indoor equipment coordinate system, the initial image of the printed material to be inspected is obtained by image acquisition.

[0060] The indoor equipment coordinate system refers to a three-dimensional spatial reference datum established based on the detection device and its physical environment. The initial image to be inspected refers to the basic image data reflecting the initial optical state of the surface of the printed part to be inspected under the preset indoor equipment coordinate system.

[0061] The indoor equipment coordinate system is determined by performing the equipment's factory calibration. Then, based on the indoor equipment coordinate system, an industrial camera is used to acquire the initial image of the printed parts to be inspected.

[0062] Equipment factory calibration refers to the high-precision mapping and calibration process of the relative positional relationship between the camera and mechanical structure of the testing device under standard conditions.

[0063] S201: Analyze the initial image to be inspected to determine the pose parameters of the printed part to be inspected.

[0064] Among them, the pose parameter refers to the real-time physical position and spatial deflection angle data of the printed part to be inspected in the preset indoor equipment coordinate system.

[0065] The pose parameters are obtained by extracting the edge contour or feature points of the printed part to be inspected from the initial image to be inspected.

[0066] S202: Based on the pose parameters, locate the bright reflection area in the initial image to be inspected, and extract the optical features contained in the bright reflection area, including the shape and contour of the light spot and the brightness distribution features.

[0067] In this context, a bright reflection region refers to a locally connected region in the initial image to be inspected where the grayscale value surges or becomes saturated due to illumination from an external ambient light source. Optical features refer to the data set showing the physical form and brightness variation patterns of the bright reflection region. The shape and contour of the light spot refer to the external geometric boundary features of the bright reflection region on the two-dimensional image plane. The brightness distribution features refer to the gradient variation data showing the decrease in pixel grayscale value from the central high-brightness extreme value towards the surrounding edges within the bright reflection region.

[0068] By extracting real-time physical position and spatial deflection angle data from the pose parameters and combining them with perspective projection mapping performed by the camera, the three-dimensional physical boundary of the printed part to be inspected is transformed into a two-dimensional pixel coordinate system to eliminate background interference, thereby accurately delineating the effective projection plane in the initial image to be inspected. Subsequently, within the effective projection plane, brightness and grid lines are set to perform grayscale threshold segmentation, filtering out discrete bright pixels. Pixel clustering is then performed on spatially adjacent bright pixels, connecting and stitching them together to locate the bright reflection area. Next, an edge detection algorithm is called to perform boundary extraction on the bright reflection area to obtain the shape contour of the light spot. The grayscale difference of each pixel within the bright reflection area is extracted and calculated to perform grayscale gradient calculation to obtain the brightness and darkness distribution features. Finally, the optical features are obtained by summarizing these features.

[0069] S203: Match and compare the shape and outline of the light spot with the preset geometric deformation mapping table to determine the projection direction of the light source.

[0070] The geometric deformation mapping table is a reference model of the distortion morphology of the light spot projection under different incident angles. The geometric deformation mapping table is obtained by pre-storing it within the system. The light source projection direction refers to the specific three-dimensional spatial tilt angle and orientation data when the external light-emitting device illuminates the surface of the printed workpiece under inspection.

[0071] By performing pre-simulated real lighting or physical optics simulation, a pre-defined geometric deformation mapping table is established. Then, the extracted light spot shape contour is compared with the standard deformation model at different angles in the pre-defined geometric deformation mapping table to deduce the actual spatial tilt posture of the light and establish it as the projection direction of the light source.

[0072] S204: Match and compare the light and dark distribution characteristics with the preset light intensity attenuation benchmark library to determine the distance of the light source.

[0073] The light intensity attenuation reference library refers to the pre-stored parameters showing the gradient change in brightness as transmission distance decreases. The light intensity attenuation reference library is determined by calculating using the optical attenuation law. The light source distance refers to the physical straight-line distance between the external light-emitting point and the reflective point on the surface of the printed workpiece under inspection.

[0074] The distance to the light source is obtained by mathematically fitting the gray-scale decrease slope from the center to the edge of the light spot, as reflected in the characteristics of light and dark distribution, with the physical theoretical curve in the preset light intensity attenuation benchmark library.

[0075] S205: Combine the light source projection direction and the distance to the light source to identify the absolute physical coordinates of the light source device currently projecting onto the surface of the printed workpiece to be inspected.

[0076] Absolute physical coordinates refer to the precise position data of the light-emitting fixture in real three-dimensional space. The light source device refers to the physical lighting hardware arranged around the printed workpiece to be inspected and currently emitting light.

[0077] By combining the projection direction and distance of the light source to perform inverse calculation, a vector point is determined in three-dimensional space, and the coordinates corresponding to the vector point are used as absolute physical coordinates.

[0078] S206: Use absolute physical coordinates as information for the incident beam.

[0079] The data encapsulation and assignment operations are performed by obtaining the absolute physical coordinates, and the absolute physical coordinates are directly used as the incident beam information of the preceding basic parameters.

[0080] To further ensure the rationality of the printing color difference defect detection method of the artificial intelligence system, it is necessary to analyze and calculate the specific steps of using the comparison image information as the first light field anomaly feature separately. The specific steps are explained in detail below.

[0081] The initial image of the printed part to be inspected and the information of the incident light beam are acquired in the following steps: S300: Combine the initial image to be inspected and the preset template image to determine the difference connected components through pixel-level image difference operation.

[0082] Among them, the difference connected region refers to the local image region where there is a numerical difference between the initial image to be inspected and the preset template image.

[0083] By combining the initial image to be inspected and the preset template image, pixel image difference operations are performed to obtain the difference connected components, which are convenient for subsequent use.

[0084] S301: Combine the incident beam information and pose parameters to calculate the theoretical reflection coordinates through spatial ray tracing.

[0085] The theoretical reflection coordinates refer to the specific spatial location data of the reflective spots produced by the incident light beam on the surface of the printed workpiece to be inspected.

[0086] By extracting the absolute spatial coordinates of the light source from the incident beam information and the tilt posture of the printed part under inspection from the pose parameters, and using the physical optical law that the incident angle equals the reflection angle, spatial ray tracing calculations are performed to simulate the actual illumination and reflection path of light in space, thereby deriving the theoretical reflection coordinates.

[0087] S302: Compare the spatial positions of the differential connected components with the theoretical reflection coordinates, and output the overlap verification results.

[0088] The overlap verification result refers to the spatial matching value between the actual location of the abnormal patch in the image and the calculated theoretical reflection point. Spatial location comparison refers to the mathematical verification process of calculating the intersection of the physical distance or area between the actual abnormal patch area and the theoretical reflection point.

[0089] By extracting local image regions with numerical differences in the differential connected components and the specific spatial location data of reflective patches with theoretical reflection coordinates, the cross-union ratio algorithm for evaluating area overlap by measuring straight-line distance is used to verify the spatial location comparison of the two in a unified coordinate system, thereby calculating and outputting the overlap verification result.

[0090] S303: When the overlap verification result is less than or equal to the preset threshold, it is determined to be in an abnormal state, and the corresponding comparison image information is used as the first light field abnormal feature based on the abnormal state.

[0091] The preset threshold refers to the spatial distance limit used to distinguish between the overlapping areas of regular printing defects and high-brightness reflections. An abnormal state is a high-risk processing state in which the image difference area falls exactly at the theoretical high-brightness reflection position. Since it is impossible to distinguish whether it is pure physical reflection or a real defect hidden under the reflection based on the two-dimensional basic image alone, it must be triggered by subsequent cross-validation.

[0092] By verifying the overlap results, when the spatial distance between the actual patch and the theoretical reflective point is greater than a preset threshold, it indicates that the image difference is not caused by external light reflection and is directly judged as a regular printing defect. When the spatial distance is less than or equal to the preset threshold, it indicates that the two spatial positions are highly overlapping, and the system judges that it has entered an abnormal state. Under this premise, the information of this comparison image is directly extracted as the first light field anomaly feature so that it can be sent to the subsequent standard part equal lighting comparison process to completely remove the reflective interference through physical reproduction.

[0093] The determination of spatial incident correspondence follows these steps: S400: Based on the incident beam information, a three-dimensional incident ray vector is constructed by performing vector difference calculation through the first spatial coordinates.

[0094] Among them, the three-dimensional incident ray vector refers to three-dimensional data that represents the location of defects on the surface of the printed workpiece from the luminous point and has a clear spatial direction.

[0095] The three-dimensional incident ray vector is obtained by calculating the coordinates of the incident ray information and the first spatial coordinates.

[0096] S401: Combine the initial image to be inspected with the pose parameters to calculate the surface normal vector of the printed part to be inspected at the first spatial coordinate.

[0097] Among them, the surface normal vector refers to the three-dimensional spatial geometric entity that is perpendicular to the local plane of the printed workpiece at the first spatial coordinate. It includes both the absolute vertical spatial orientation and specific unit distance data.

[0098] By extracting surface contour features from the initial image to be inspected and combining them with the spatial tilt state in the pose parameters, the direction and distance are determined by normal calculation at the first spatial coordinates, and the direction and distance are used as the surface normal vector.

[0099] S402: Calculate the spatial geometric angle between the three-dimensional incident ray vector and the surface normal vector to obtain the incident angle and azimuth angle of the incident beam at the first spatial coordinate.

[0100] Among them, the spatial geometric angle refers to the relative angle formed by the three-dimensional incident ray vector and the surface normal vector in three-dimensional space. The incident angle refers to the absolute solid angle formed directly by the three-dimensional incident ray vector and the surface normal vector in three-dimensional space, representing the degree of oblique illumination of the light. The azimuth angle refers to the plane angle formed between the projection direction of the three-dimensional incident ray vector on the tangent plane corresponding to the surface of the printed workpiece and the projection direction of the reference axis of the three-dimensional coordinate system to which the first spatial coordinate belongs on the tangent plane.

[0101] The incident angle is directly extracted by calculating the spatial geometric angle formed by the three-dimensional incident ray vector and the surface normal vector in three-dimensional space. At the same time, the azimuth angle is determined by synchronously projecting the three-dimensional incident ray vector and the reference axis of the three-dimensional coordinate system to which the first spatial coordinate belongs onto a local tangent plane on the surface of the printed workpiece to be inspected, and calculating the plane angle formed between the two projection directions.

[0102] S403: Determine the spatial incident correspondence by combining the incident angle and azimuth angle.

[0103] In this step, the spatial incident correspondence specifically refers to the set of three-dimensional geometric direction parameter data formed by logically coupling independent incident angles and azimuth angles.

[0104] By extracting the incident angle values ​​representing the degree of tilt of the solid and the azimuth angle values ​​representing the deflection direction on the local tangent plane, and treating them as independent and complementary spatial dimension parameters, vectorization and data encapsulation are performed to generate a set of directional data sets that form a spatial incident correspondence.

[0105] To further ensure the rationality of the printing color difference defect detection method of the artificial intelligence system, it is necessary to analyze and calculate the method for determining the second spatial coordinates separately, which will be explained in detail through the following steps.

[0106] The method for determining the second spatial coordinates is as follows: S500: Combining the incident beam information and the first spatial coordinates, the equivalent mapping radius is determined by calculating the spatial Euclidean distance.

[0107] The equivalent mapping radius refers to the spatial parameter that characterizes the physical distance between the light source point and the surface reflection point.

[0108] The equivalent mapping radius is obtained by performing spatial Euclidean distance calculation by combining the incident beam information and the first spatial coordinates.

[0109] S501: Establish a virtual spherical coordinate system by taking the reference point of the preset standard part surface as the origin and combining it with the equivalent mapping radius.

[0110] The reference point on the surface of the standard part refers to the initial coordinate position selected on the pre-defined surface of the standard part for spatial positioning. The virtual spherical coordinate system refers to a three-dimensional spatial reference frame constructed with the reference point on the surface of the standard part as the center.

[0111] The reference points on the surface of the standard parts are obtained by pre-calibrating and extracting the standard parts. Then, a virtual spherical coordinate system is obtained by performing spatial architecture deduction with the reference points on the surface of the standard parts as the origin and in combination with the equivalent mapping radius.

[0112] S502: Extract the incident angle and azimuth angle from the spatial incident correspondence and use them as the target polar angle and target azimuth angle.

[0113] The target polar angle refers to the zenith angle parameter, which is the angle between the virtual ray vector and the vertical normal in the virtual spherical coordinate system. The target azimuth angle refers to the longitude angle parameter, which is the angle between the virtual ray vector on the equatorial projection plane and the reference axis.

[0114] By extracting the incident angle value representing the three-dimensional tilt of the light in the spatial incident correspondence, and performing an assignment operation without changing its absolute value, the value is mapped and established as the target polar angle in the virtual spherical coordinate system. At the same time, the azimuth angle value representing the plane deflection direction of the light is extracted, and the value is assigned without changing its absolute value, and the value is mapped and established as the target azimuth angle in the virtual spherical coordinate system.

[0115] S503: Combining the equivalent mapping radius, target polar angle, and target azimuth angle, perform the conversion calculation from spherical coordinates to rectangular coordinates in this virtual spherical coordinate system to obtain the theoretical rectangular coordinates.

[0116] The conversion from spherical coordinates to rectangular coordinates involves using basic trigonometric functions to map distance and angle parameters from a virtual spherical coordinate system to specific coordinates on a three-dimensional rectangular coordinate system. Theoretical rectangular coordinates refer to the positional data representing the expected point of impact of light rays in a three-dimensional rectangular coordinate system.

[0117] By extracting the equivalent mapping radius, target polar angle, and target azimuth angle, the theoretical rectangular coordinates are determined by performing a transformation calculation from spherical coordinates to rectangular coordinates in a virtual spherical coordinate system.

[0118] S504: Project the theoretical rectangular coordinates onto the surface of the preset standard part to obtain the actual landing point, and determine the second spatial coordinates based on the actual landing point.

[0119] The actual landing point refers to the contact point position corresponding to the theoretical rectangular coordinates mapped vertically onto the real physical surface.

[0120] The calculated theoretical rectangular coordinates are projected onto the surface of a pre-defined standard part along a vertical spatial path. During this process, the intersection calculation of the spatial ray and the mesh surface is performed using the pre-stored 3D surface contour model of the standard part (3D CAD topography data imported from the equipment factory). This allows the system to automatically adapt to whether the standard part surface is planar or curved, thus ensuring precise landing and fitting to obtain the actual landing point. The spatial coordinates of the actual landing point are then extracted as the second spatial coordinates.

[0121] To further ensure the rationality of the printing color difference defect detection method of the artificial intelligence system, it is necessary to analyze and calculate the generation method of the second light field anomaly feature separately, which will be explained in detail through the following steps.

[0122] The method for generating the second light field anomaly feature is as follows: S600: Combine spatial incident correspondence with second spatial coordinates to perform three-dimensional spatial inverse deduction and construct a three-dimensional target illumination vector.

[0123] Among them, the three-dimensional target illumination vector refers to the directional distance in three-dimensional space that represents the expected luminous direction.

[0124] By combining the spatial incident correspondence with the second spatial coordinates and calculating the three-dimensional geometry of the light source's position and lighting direction along the light path in reverse, a three-dimensional target illumination vector is constructed.

[0125] S601: Determine spatial lighting control commands based on three-dimensional target lighting vectors.

[0126] Among them, spatial lighting control commands refer to the underlying hardware communication signal data used to adjust and control the working status of lighting equipment.

[0127] Spatial lighting control instructions are obtained by compiling instructions based on the three-dimensional target lighting vector.

[0128] S602: Drives a preset directional reproduction light source according to the spatial lighting control command, and outputs a directional beam to project to the second spatial coordinate.

[0129] By receiving spatial lighting control commands from the underlying system, the system precisely drives the servo motor or micromirror array inside the directional reproduction light source, controls its output to be a directional beam that is completely equivalent to the original incident light path, and accurately projects the beam onto the second spatial coordinates on the surface of the standard part. This serves as a dedicated reproduction tool, forcibly exciting physical reflection patches on the surface of the standard part for equivalent comparison, and identifying the second light field anomaly characteristics from the physical reflection patches.

[0130] S603: During the projection of the directional beam, synchronous image acquisition is performed on the surface of a preset standard part to obtain an initial reference image.

[0131] The initial reference image refers to the basic image data that reflects the optical response of the surface of a standard part, obtained under specific physical orientation lighting conditions.

[0132] By continuously and stably illuminating the surface of a preset standard part with a directional beam, the vision sensor is controlled to align with the target area to trigger exposure and capture an initial reference image.

[0133] S604: Based on the second spatial coordinates, locate and determine the corresponding physical light spot reflection characteristics in the initial reference image.

[0134] Among them, physical light spot reflection characteristics refer to the pure visual characteristics presented by light reflection at a specific coordinate position.

[0135] By inversely projecting the region of the second spatial coordinates in the initial reference image into the two-dimensional pixel matrix of the initial reference image, a specific region is delineated. Subsequently, a continuous set of pixels within this specific region is directly extracted and established as a physical light spot reflection feature.

[0136] To further ensure the rationality of the printing color difference defect detection method of the artificial intelligence system, it is necessary to analyze and calculate the judgment method of printing color difference defect detection results separately, which will be explained in detail through the following steps.

[0137] The method for determining the results of printing color difference defect detection is as follows: S700: Optical parameters are determined by combining the first and second light field anomaly features. The optical parameters include multi-channel color intensity and reflectance brightness components.

[0138] Among them, optical parameters refer to the set of values ​​characterizing the difference in light field characteristics between the abnormal area of ​​the printed part under inspection and the equivalent mapping area of ​​the standard part. Multi-channel color intensity refers to the pixel grayscale value quantized in the red, green, and blue color channels of the image. Reflectance brightness component refers to the numerical representation of light intensity and brightness after stripping away color information.

[0139] By directly extracting pixel data from the first and second light field anomaly features, feature parsing and comparison are performed to calculate and determine the optical parameters. Then, by performing orthogonal separation of color and brightness on the acquired optical parameters, multi-channel color intensity and reflectance brightness components are extracted.

[0140] S701: Performs difference matrix calculations on multi-channel color intensity and reflectance brightness components to obtain spatial spectral deviation.

[0141] Among them, spatial spectral deviation refers to a quantitative indicator that characterizes the degree of deviation between the actual physical reflection state of the printed part under inspection and the defect-free standard benchmark in the spectral dimension.

[0142] The multi-channel color intensity and reflectance luminance components of the printed parts to be inspected and the standard parts are arranged into corresponding data matrices. Then, the color and luminance values ​​at the same position in the two sets of data matrices are directly subtracted one-to-one to calculate the true difference between the two, and this difference is established as the spatial spectral deviation.

[0143] S702: Confirm the physical reflectance difference value based on the spatial spectral deviation.

[0144] By extracting the spatial spectral deviation and multiplying it by the corresponding influence weight, a comprehensive value is calculated to intuitively measure the overall error between actual lighting and theoretical comprehensive values. This overall error between lighting is then used as the physical reflection difference value.

[0145] S703: Compare the physical reflection difference value with the preset color difference tolerance threshold.

[0146] This is achieved by performing a numerical comparison between the physical reflectance difference value and the color difference tolerance threshold.

[0147] Numerical comparison operation refers to the Boolean logic operation process in the system's underlying logic unit to determine whether the physical reflection difference value exceeds the floating range of the color difference tolerance threshold.

[0148] S704: When the physical reflection difference value is greater than the preset color difference tolerance threshold, it is determined that there is a printing color difference defect and the color difference defect detection result is output.

[0149] Printing color difference defects are identified by performing a judgment operation when the physical reflectance difference value exceeds the color difference tolerance threshold. The color difference defect detection results are output by performing a report generation action on the identified defect data.

[0150] To further ensure the rationality of the printing color difference defect detection method of the artificial intelligence system, it is necessary to analyze and calculate the method for setting the color difference tolerance threshold separately, which will be explained in detail through the following steps.

[0151] The method for setting the color difference tolerance threshold is as follows: S800: Obtain the basic surface color parameters and surface roughness parameters of the printed part to be inspected.

[0152] Among them, the surface base color parameter refers to the value that characterizes the color reference of the surface of the printed part to be inspected. The surface roughness parameter refers to the physical value that characterizes the degree of microscopic unevenness of the surface of the printed part to be inspected.

[0153] The surface base color parameters and surface roughness parameters are obtained by performing data reading.

[0154] S801: Match the corresponding initial color difference tolerance based on the surface base color parameters and surface roughness parameters.

[0155] The initial chromatic difference tolerance refers to the initial value of the range of chromatic difference fluctuations allowed under ideal imaging conditions.

[0156] The initial color difference tolerance is obtained by performing a logical query process based on the surface base color parameters and surface roughness parameters in a preset database to find the corresponding empirical deviation threshold according to different material colors and roughness levels.

[0157] The pre-set database refers to a structured data set that is pre-stored in the system storage medium and contains a multi-dimensional mapping relationship between the basic colors, roughness grades and empirical deviation thresholds of various printing materials.

[0158] S802: Controls the ambient light source to turn off, performs image acquisition in a dark environment, and obtains a pure black reference image.

[0159] The "no-light environment" refers to a controlled darkroom state where external stray light is shielded and the internal lighting source of the detection device is extinguished. The pure black reference image refers to the basic image data reflecting the dark current state of the imaging system, acquired under conditions of no active light source illumination.

[0160] A dark environment is obtained by controlling the ambient light source to turn off, and a pure black reference image is obtained by performing image acquisition in the dark environment.

[0161] S803: Calculates the average deviation range of pixel color values ​​in a pure black reference image and uses it as a device background noise parameter.

[0162] Among them, the device background noise parameter refers to the quantitative value of imaging interference caused by hardware characteristics.

[0163] The device background noise parameters are obtained by calculating the degree of dispersion of the gray values ​​of each pixel in a pure black reference image relative to the absolute zero value.

[0164] S804: Combine the device background noise parameters with the initial color difference tolerance and sum them to determine the actual color difference tolerance.

[0165] Among them, the actual color difference tolerance refers to the color difference benchmark used for the final judgment after background noise correction.

[0166] By extracting the background noise parameters of the device and adding them directly to the preset initial color difference tolerance, the original color difference judgment standard is adjusted to complete the numerical compensation, thereby obtaining the actual color difference tolerance.

[0167] S805: Use the actual color difference tolerance as the color difference tolerance threshold.

[0168] The actual color difference tolerance value is written into the global decision register as the color difference tolerance threshold.

[0169] To further ensure the rationality of the printing color difference defect detection method of the artificial intelligence system, it is necessary to analyze and calculate the method for determining the actual color difference tolerance separately, which will be explained in detail through the following steps.

[0170] The method for determining the actual color difference tolerance is as follows: S900: Get the device operating temperature during the current image acquisition.

[0171] Among them, the device operating temperature refers to the real-time thermal state value of the image acquisition hardware environment when performing image acquisition tasks.

[0172] The operating temperature of the device is obtained by performing a sampling operation on a thermistor on the heat sink of an industrial camera.

[0173] S901: Calculate the temperature difference between the operating temperature of the equipment and the preset equipment calibration temperature.

[0174] The equipment calibration temperature refers to the ambient temperature value during the initial accuracy calibration of the image acquisition hardware in a controlled laboratory environment. The temperature difference refers to the numerical difference between the real-time operating temperature and the factory calibration temperature.

[0175] The temperature difference is obtained by subtracting the equipment's operating temperature from its preset calibration temperature.

[0176] S902: Determine whether the temperature difference is less than the preset temperature drift compensation threshold.

[0177] Among them, the temperature drift compensation threshold refers to a pre-set critical value used to determine whether temperature drift significantly affects the hardware imaging accuracy.

[0178] By performing a numerical comparison operation, it is determined whether the temperature difference is less than the preset temperature drift compensation threshold, which facilitates the selection of different compensation paths in the future.

[0179] S903: When the temperature difference is less than the preset temperature drift compensation threshold, the background noise parameter of the equipment and the initial color difference tolerance are added together to determine the actual color difference tolerance.

[0180] When the detected temperature difference is less than the preset temperature drift compensation threshold, the system is determined to be in a thermally stable state. At this time, the background noise parameter extracted by the direct device is added to the initial color difference tolerance to complete the linear correction, and the corrected result is determined as the actual color difference tolerance.

[0181] S904: When the temperature difference is not less than the preset temperature drift compensation threshold, the corresponding amplification factor is determined based on the temperature difference, and the amplification factor is used as the hardware error relaxation factor.

[0182] Here, the magnification factor refers to the proportional compensation value used to quantify the thermal noise of the photosensitive component as temperature increases. The hardware error relaxation factor refers to the dynamic adjustment weight value used to compensate for the decrease in imaging accuracy of the underlying camera caused by high-temperature environments.

[0183] By inputting the temperature difference into a pre-established mapping model and performing a matching calculation, the corresponding amplification factor is directly output, and this amplification factor is used as a hardware error relaxation factor. The mapping model refers to a table storing different temperature differences and their corresponding amplification factors, which is pre-established in a database during the calibration and testing phase. The larger the input temperature difference, the larger the corresponding amplification factor.

[0184] S905: Calculate the inherent error tolerance of the equipment by multiplying the background noise parameters of the equipment with the hardware error relaxation factor.

[0185] Among them, the inherent error tolerance of the equipment refers to the comprehensive hardware error allowable value under the current environmental temperature drift interference.

[0186] The inherent error tolerance of the device is obtained by multiplying the device background noise parameters with the hardware error relaxation factor, which facilitates subsequent deep error fusion.

[0187] S906: Summing the inherent error tolerance of the equipment with the initial color difference tolerance determines the actual color difference tolerance.

[0188] By extracting the inherent error tolerance of the equipment after temperature drift correction, and adding it directly to the preset initial color difference tolerance, the original color difference judgment standard is adjusted, and the adjusted color difference judgment standard is used as the actual color difference tolerance.

[0189] Based on the same inventive concept, this invention provides a printing color difference defect detection device based on an artificial intelligence system, including a memory and a processor. The memory stores a computer program that can be loaded and executed by the processor as described above for a printing color difference defect detection method based on an artificial intelligence system.

[0190] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0191] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for detecting printing color difference defects based on an artificial intelligence system, characterized in that, include: Acquire the initial image of the printed part to be inspected and the information of the incident light beam; The initial image to be inspected is combined with the preset template image to determine the comparison image information and the first spatial coordinates through image difference; the comparison result is output by spatial ray tracing based on the comparison image information and the incident beam information, and the comparison result includes abnormal states; In abnormal conditions, the information from the contrasting image is used as the first light field anomaly feature; the spatial incidence correspondence is determined by combining the incident beam information and the first spatial coordinate correspondence. The second spatial coordinates of the preset standard part surface are determined based on the physical mapping between the spatial incident correspondence and the preset spherical coordinates; The incident beam is physically transmitted to the second spatial coordinates at an equivalent incident angle to generate a second optical field anomalous feature; The first light field anomaly feature is compared with the second light field anomaly feature to obtain the physical reflection difference value; The physical reflection difference value is compared with the preset color difference tolerance threshold. If the physical reflection difference value is greater than the preset color difference tolerance threshold, it is determined that there is a printing color difference defect and the color difference defect detection result is output.

2. The printing color difference defect detection method based on an artificial intelligence system according to claim 1, characterized in that, The initial image of the printed part to be inspected and the information of the incident light beam are acquired, including: Under the preset indoor equipment coordinate system, the initial image of the printed parts to be inspected is obtained by image acquisition. Incident beam information refers to the external light source that illuminates the surface of the printed workpiece under inspection and produces a bright reflection. Based on the initial image to be inspected, the pose parameters of the printed part to be inspected are determined. Based on the pose parameters, the bright reflection region in the initial image to be inspected is located, and the optical features contained in the bright reflection region are extracted; the optical features include the shape and contour of the light spot and the brightness and darkness distribution features; The shape and outline of the light spot are matched and compared with a preset geometric deformation mapping table to determine the projection direction of the light source; The distance to the light source is determined by matching and comparing the characteristics of the light and dark distribution with a preset light intensity attenuation benchmark library. By combining the direction of light source projection and the distance to the light source, the absolute physical coordinates of the light source device currently projected onto the surface of the printed workpiece to be inspected can be identified; Absolute physical coordinates are used as information for the incident beam.

3. A printing color difference defect detection method based on an artificial intelligence system according to claim 2, characterized in that, Based on the anomalous state, the specific steps for using the comparative image information as the first light field anomalous feature include: The difference connected components are determined by pixel-level image difference operations by combining the initial image to be inspected and the preset template image; By combining the incident beam information and pose parameters, the theoretical reflection coordinates are calculated through spatial ray tracing. The spatial positions of the differential connected components are compared with the theoretical reflection coordinates, and the overlap verification results are output. When the overlap verification result is less than or equal to the preset threshold, it is determined to be in an abnormal state, and the corresponding comparison image information is used as the first light field abnormal feature based on the abnormal state.

4. A printing color difference defect detection method based on an artificial intelligence system according to claim 2, characterized in that, Methods for determining spatial incident correspondence include: Based on the incident beam information, a three-dimensional incident ray vector is constructed by performing vector difference calculation using the first spatial coordinates. The surface normal vector of the printed part under inspection at the first spatial coordinate is calculated by combining the initial image to be inspected with the pose parameters. Calculate the spatial geometric angle between the three-dimensional incident ray vector and the surface normal vector to obtain the incident angle and azimuth angle of the incident beam at the first spatial coordinate. The spatial incident correspondence is determined by combining the incident angle and the azimuth angle.

5. A printing color difference defect detection method based on an artificial intelligence system according to claim 1, characterized in that, Methods for determining the second spatial coordinates include: By combining the incident beam information and the first spatial coordinates, the equivalent mapping radius is determined by calculating the spatial Euclidean distance. A virtual spherical coordinate system is established using the reference point of the preset standard part surface as the origin and the equivalent mapping radius. Extract the incident angle and azimuth angle from the spatial incident correspondence and use them as the target polar angle and target azimuth angle; By combining the equivalent mapping radius, the target polar angle, and the target azimuth angle, a transformation calculation from spherical coordinates to rectangular coordinates is performed in this virtual spherical coordinate system to obtain the theoretical rectangular coordinates; The theoretical rectangular coordinates are projected onto the surface of a preset standard part to obtain the actual landing point, and the second spatial coordinates are determined based on the actual landing point.

6. A printing color difference defect detection method based on an artificial intelligence system according to claim 1, characterized in that, Methods for generating the second light field anomalous feature include: By combining the spatial incident correspondence with the second spatial coordinates, a three-dimensional spatial inverse deduction is performed to construct a three-dimensional target illumination vector; Determine spatial lighting control commands based on three-dimensional target lighting vectors; Drive the preset directional reproduction light source according to the spatial lighting control command, and output a directional beam of light to project to the second spatial coordinate; During the directional beam projection, synchronous image acquisition is performed on the surface of a preset standard part to obtain an initial reference image; Based on the second spatial coordinates, locate and determine the corresponding physical light spot reflection characteristics in the initial reference image; The physical light spot reflection characteristics are used as the second light field anomaly characteristics.

7. A printing color difference defect detection method based on an artificial intelligence system according to claim 1, characterized in that, Methods for determining the results of printing color difference defect detection include: Optical parameters are determined by combining the first and second light field anomaly features. These optical parameters include multi-channel color intensity and reflectance brightness components. Perform difference matrix calculations on the multi-channel color intensity and reflectance components to obtain the spatial spectral deviation. The difference in physical reflectance is determined based on the spatial spectral deviation. The physical reflection difference value is compared with the preset color difference tolerance threshold. When the physical reflection difference value is greater than the preset color difference tolerance threshold, it is determined that there is a printing color difference defect, and the color difference defect detection result is output.

8. A printing color difference defect detection method based on an artificial intelligence system according to claim 7, characterized in that, Methods for setting the color difference tolerance threshold include: Obtain the basic color parameters and surface roughness parameters of the printed part to be inspected; Based on the surface base color parameters and surface roughness parameters, match the corresponding initial color difference tolerance; The preset ambient light source is turned off, and image acquisition is performed in a dark environment to obtain a pure black baseline image. Calculate the average deviation range of pixel color values ​​in a pure black reference image and use it as a device background noise parameter; The actual color difference tolerance is determined by summing the background noise parameters of the equipment with the initial color difference tolerance. The actual color difference tolerance is used as the color difference tolerance threshold.

9. A printing color difference defect detection method based on an artificial intelligence system according to claim 8, characterized in that, Methods for determining actual color difference tolerance include: Get the device operating temperature during the current image acquisition process; Calculate the temperature difference between the operating temperature of the equipment and the preset equipment calibration temperature; Determine whether the temperature difference is less than the preset temperature drift compensation threshold; When the temperature difference is less than the preset temperature drift compensation threshold, the background noise parameter of the equipment and the initial color difference tolerance are directly added together to determine the actual color difference tolerance. When the temperature difference is not less than the preset temperature drift compensation threshold, the corresponding amplification factor is determined based on the temperature difference, and the amplification factor is used as the hardware error relaxation factor. The inherent error tolerance of the equipment is determined by multiplying the background noise parameters of the equipment with the hardware error relaxation factor. The actual color difference tolerance is determined by summing the inherent error tolerance of the equipment with the initial color difference tolerance.

10. A printing color difference defect detection device based on an artificial intelligence system, characterized in that, The invention includes a memory and a processor, wherein the memory stores a computer program that can be loaded and executed by the processor, as described in any one of claims 1 to 9, and is a method for detecting printing color difference defects based on an artificial intelligence system.

Images