LED lamp bead packaging detection method and system

Abstract

This invention relates to the field of LED chip packaging inspection technology, specifically to an LED chip packaging inspection method and system. Addressing the problem that transparent colloids, cup wall reflections, chip emitting surfaces, and gold wire reflections all affect image response, leading to frequent misjudgments and missed detections of packaging defects, the method acquires a coaxial diffuse positioning image and establishes the packaging area coordinates. Under these coordinates, coaxial diffuse images, left and right oblique incidence dark field images, and low-current self-emissive images are acquired. Based on the regional response benchmark, abnormal connected components are extracted to generate a packaging optical path response cross-verification matrix. Then, based on the response consistency of the same suspected defect across the dark field image, self-emissive image, and packaging area, defect cross-verification objects are generated and matched with surface scattering, internal attenuation, contour deviation, overflow continuity, and gold wire occlusion rules. The method outputs the defect type, the packaging area, the contour mask, and the severity, suitable for LED chip packaging quality inspection.

Description

Technical Field

[0001] This invention relates to the field of LED chip packaging and testing technology, specifically to an LED chip packaging and testing method and system. Background Technology

[0002] LED chip packaging inspection is typically used for quality checks after chip dispensing, curing, fluorescent glue encapsulation, or transparent colloid encapsulation. Current methods often involve using industrial cameras to capture images of the chip's appearance and combining this with fixed light sources, dark lighting, image segmentation, threshold determination, or model recognition to inspect the encapsulation colloid, cup edge, chip light-emitting area, gold wire area, and overflow area to determine if there are any abnormalities such as bubbles, cracks, insufficient glue, excessive glue, leakage, overflow, surface foreign matter, scratches, or gold wire obstruction.

[0003] In actual testing, the transparent colloid of LED beads, reflections from the cup wall, the chip's light-emitting surface, the phosphor layer, and gold wire reflections all simultaneously affect the image response. The same suspected anomaly can manifest significantly differently under different lighting conditions and in different packaging areas. Existing methods, which primarily rely on a single image, a single optical path response, or ordinary defect labels for judgment, often struggle to adequately distinguish between differences in surface scattering, internal attenuation, colloid contour deviation, overflowing continuous areas, and gold wire obstruction. Furthermore, when the response is uncertain, there is a lack of clear criteria for re-judgment, easily leading to misjudgments or missed detections. Therefore, a new LED bead packaging inspection method and system are needed to address these issues. Summary of the Invention

[0004] To solve the above problems, the present invention provides the following technical solution: a method for testing LED bead packaging, comprising:

[0005] S1. Obtain the coaxial diffuse positioning image of the LED bead to be tested, identify the cup edge, chip light-emitting area, gold wire channel and outer edge of adhesive surface, and establish the coordinates of the packaging area containing multiple packaging areas;

[0006] S2. Acquire multi-optical-path images in the coordinates of the encapsulation area. The multi-optical-path images include a coaxial diffuse image, a left oblique incident dark field image, a right oblique incident dark field image, and a low-current self-emissive image.

[0007] S3. Based on the regional response reference corresponding to the coordinates of the encapsulation region, extract abnormal connected components from the multi-optical path image as suspected defects. Generate an encapsulation optical path response verification matrix according to the optical path number, encapsulation region number, and the abnormal connected component position, response intensity, centroid coordinates, and self-emission attenuation value of the suspected defects.

[0008] S4. Based on the response consistency of the same suspected defect in the encapsulated optical path response mutual verification matrix among the dark field image, the self-illuminating image and the encapsulated area, generate a defect mutual verification object.

[0009] S5. Match the defect verification object with a preset optical path response rule that includes multiple response types to determine the defect type, the encapsulation area where the defect is located, the contour mask, and the severity.

[0010] Furthermore, the multiple packaging regions include a chip window region, a gold wire channel region, a colloid center region, a colloid edge region, a cup wall region, and an overflow inspection region; establishing the packaging region coordinates containing multiple packaging regions includes:

[0011] Determine the center of the cup rim based on the described cup rim boundary;

[0012] The center of the chip's light-emitting area is determined based on the chip's light-emitting area;

[0013] A positioning reference is determined from the center of the cup rim, the center of the chip light-emitting area, and the outer edge of the adhesive surface, and the image coordinates of the LED bead to be tested are normalized according to the positioning reference.

[0014] The multiple encapsulation regions are divided into normalized image coordinates to generate the coordinates of the encapsulation regions.

[0015] Furthermore, the low-current self-emissive image is acquired after the external light source is turned off; when acquiring the low-current self-emissive image, the LED bead to be tested is driven in a short pulse mode, and the short pulse driving timing is synchronized with the camera exposure timing; the driving current in the short pulse mode is lower than the preset temperature rise influence current, which is determined by the device calibration parameters, and is used to make the low-current self-emissive image characterize the blocking, scattering and attenuation of the light emission path by the encapsulating colloid.

[0016] Furthermore, the encapsulated optical path response verification matrix uses optical path numbers as matrix rows and encapsulation region numbers as matrix columns; the optical path numbers include at least coaxial diffuse optical path, left oblique incident dark field optical path, right oblique incident dark field optical path, and low-current self-emissive optical path; the matrix units of the encapsulated optical path response verification matrix are used to store at least one abnormal response record, each abnormal response record including the optical path number, encapsulation region number, abnormal connected component position, response intensity, centroid coordinates, edge gradient direction, and color channel ratio under the corresponding optical path; wherein, the abnormal response record corresponding to the low-current self-emissive optical path also includes the self-emissive attenuation value.

[0017] Furthermore, the step of generating a defect verification object based on the response consistency of the same suspected defect in the encapsulated optical path response verification matrix across the dark field image, the self-illuminating image, and the encapsulated region includes:

[0018] Based on the overlapping relationships of abnormal connected domain positions, the proximity relationships of centroid coordinates, the consistency relationships of encapsulation region numbers, and the changing trends of response intensity in the abnormal response records, abnormal response records belonging to the same suspected defect are grouped into a set of abnormal response records for mutual verification.

[0019] The response consistency type is determined based on the set of abnormal response records participating in the mutual authentication;

[0020] Based on the suspected defect, the encapsulation region number, the set of abnormal response records participating in the mutual verification, the response consistency type, the candidate response type, and the severity, the defect mutual verification object is generated.

[0021] Furthermore, the preset optical path response rules include surface scattering rules, internal attenuation rules, contour deviation rules, overflow continuity rules, and gold line occlusion rules.

[0022] The surface scattering rule is used to determine the corresponding defect verification object as surface contamination or scratch when the response intensity of the suspected defect in the left oblique incident dark field image and the right oblique incident dark field image reaches the surface scattering threshold and the self-luminescence attenuation value is lower than the internal defect judgment threshold.

[0023] The internal attenuation rule is used to determine the corresponding defect verification object as a bubble or crack when the centroid coordinates of the suspected defect in the left oblique incident dark field image and the right oblique incident dark field image form a relative offset related to the corresponding incident direction, and the self-luminous attenuation value reaches the internal defect judgment threshold.

[0024] The contour deviation rule is used to determine the corresponding defect verification object as excess glue, insufficient glue, or leakage glue when the abnormal connected domain position in the edge region of the colloid coincides with the outer edge of the glue surface, and the contour deviation range or the self-luminous radial attenuation range reaches the corresponding benchmark.

[0025] The overflow continuity rule is used to identify the corresponding defect verification object as glue overflow when a continuous high-response area is formed between the overflow inspection area and the cup wall area.

[0026] The gold wire occlusion rule is used to identify the corresponding defect verification object as a gold wire occlusion anomaly when the gold wire channel region forms a linear high reflectivity or occlusion contour in the coaxial diffuse image, forms a local linear attenuation in the low current self-luminous image, and the dark field scattering response does not meet the surface scattering rule.

[0027] Furthermore, the corresponding benchmarks for the surface scattering threshold, the internal defect judgment threshold, the contour deviation range, and the self-luminous radial attenuation range are determined based on at least one of the following: qualified LED bead samples from the same batch, the non-abnormal reference area of ​​the current LED bead to be tested, or the equipment calibration parameters.

[0028] Furthermore, if the same defect verification object simultaneously satisfies multiple preset optical path response rules, or does not satisfy any preset optical path response rule but its response intensity reaches the threshold for re-judgment, then the defect verification object is marked as a defect verification object to be re-judged; supplementary acquisition is performed on the encapsulation area corresponding to the defect verification object to be re-judged, and the defect type is re-determined after updating the encapsulation optical path response verification matrix based on the image obtained from the supplementary acquisition; the supplementary acquisition includes at least one of adjusting the oblique incidence angle, adjusting the self-illumination exposure time, or reducing the local acquisition window.

[0029] The present invention also provides an LED chip packaging inspection system for implementing the above-mentioned LED chip packaging inspection method, comprising:

[0030] The packaging area coordinate establishment module is used to acquire the coaxial diffuse positioning image of the LED bead to be tested, identify the cup edge, chip light-emitting area, gold wire channel and outer edge of adhesive surface, and establish the packaging area coordinates containing multiple packaging areas;

[0031] A multi-optical-path image acquisition module is used to acquire multi-optical-path images under the coordinates of the encapsulation area. The multi-optical-path images include a coaxial diffuse image, a left oblique incident dark field image, a right oblique incident dark field image, and a low-current self-emissive image.

[0032] The mutual verification matrix generation module is used to extract abnormal connected components as suspected defects from the multi-optical path image based on the regional response reference corresponding to the coordinates of the encapsulation area, and generate an encapsulation optical path response mutual verification matrix according to the optical path number, encapsulation area number, and the abnormal connected component position, response intensity, centroid coordinates and self-emission attenuation value of the suspected defect.

[0033] The mutual verification object generation module is used to generate a defect mutual verification object based on the response consistency of the same suspected defect in the encapsulated optical path response mutual verification matrix among the dark field image, the self-illuminating image and the encapsulated region.

[0034] The defect result determination module is used to match the defect verification object with a preset optical path response rule containing multiple response types to determine the defect type, the encapsulation area where the defect is located, the contour mask, and the severity.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] By establishing a mutual verification matrix of packaged optical path response and correlating the response relationships of the same suspected defect in coaxial diffuse images, left and right oblique incident dark field images, low current self-emissive images, and different packaged regions, the detection process no longer relies solely on a single image or a single optical path response for judgment. This reduces the possibility of cup wall reflection, transparent colloid curvature, chip light-emitting surface, and phosphor layer color differences being misjudged as packaged defects.

[0037] By generating defect verification objects and matching them with preset optical path response rules for surface scattering, internal attenuation, contour deviation, overflow continuity, and gold wire obstruction, surface contamination or scratches, bubbles or cracks, excess glue, insufficient glue, leakage glue, overflow glue, and gold wire obstruction anomalies can be distinguished according to their respective optical response relationships, thereby reducing confusion between different types of packaging anomalies.

[0038] When a defect verification object simultaneously meets multiple preset optical path response rules, or fails to meet any preset optical path response rule but its response intensity reaches the threshold for re-judgment, the defect verification object is marked as a defect verification object to be re-judged. Supplementary acquisition is then performed on the corresponding encapsulation area to re-determine the defect type. This prevents suspected defects with uncertain responses from directly entering the final judgment, thereby reducing misjudgments and omissions caused by single acquisition or single rule matching. Attached Figure Description

[0039] Figure 1 This invention provides a schematic diagram of the steps involved in the packaging and testing of LED beads.

[0040] Figure 2 The overall flowchart of an LED bead packaging and testing method provided by the present invention;

[0041] Figure 3 This invention provides an overview diagram of the multi-path response cross-verification of an LED bead packaging testing method.

[0042] Figure 4 Comparison of defect detection effects of an LED bead packaging inspection method provided by the present invention;

[0043] Figure 5 A defect type identification confusion matrix diagram for an LED lamp bead packaging inspection method provided by the present invention;

[0044] Figure 6 This is a schematic diagram of the structure of an LED bead packaging and testing system provided by the present invention. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] To enable those skilled in the art to better understand the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0047] Please see Figures 1 to 6 , Figure 1 This invention provides a schematic diagram of the steps involved in the packaging and testing of LED beads. Figure 2 The overall flowchart of an LED bead packaging and testing method provided by the present invention; Figure 3 This invention provides an overview diagram of the multi-path response cross-verification of an LED bead packaging testing method. Figure 4 Comparison of defect detection effects of an LED bead packaging inspection method provided by the present invention; Figure 5 A defect type identification confusion matrix diagram for an LED lamp bead packaging inspection method provided by the present invention; Figure 6 This is a schematic diagram of an LED chip packaging inspection system provided by the present invention. The present invention also provides an LED chip packaging inspection method, comprising:

[0048] S1. Obtain the coaxial diffuse positioning image of the LED bead to be tested, identify the cup edge, chip light-emitting area, gold wire channel and outer edge of adhesive surface, and establish the coordinates of the packaging area containing multiple packaging areas;

[0049] Specifically, in this embodiment, the LED bead to be tested is placed at the packaging inspection station, which includes a camera, a coaxial diffuse light source, left and right obliquely incident dark field light sources, a low-current pulse driving unit, and an image processing unit. The image processing unit first controls the coaxial diffuse light source to illuminate the LED bead to be tested, and the camera acquires a coaxial diffuse positioning image. The coaxial diffuse positioning image is used to identify the cup rim boundary, the chip light-emitting area, the gold wire channel, and the outer edge of the adhesive surface in the same coordinate system, rather than being directly used for final defect determination.

[0050] When identifying the cup rim boundary, the image processing unit performs grayscale normalization and edge enhancement on the coaxial diffuse positioning image, extracting approximately annular or arc-shaped boundaries as the cup rim boundary. When identifying the chip light-emitting area, it extracts regions whose brightness, texture, or geometric position meets the chip window characteristics. When identifying the gold wire channel, it determines it based on the linear high-reflectivity contour, the slender structure connecting the chip light-emitting area and the cup wall, and its positional relationship relative to the chip light-emitting area. When identifying the outer edge of the adhesive surface, it extracts the edge gradient changes between the adhesive surface and the cup wall or external region. After completing the identification, the image processing unit writes each boundary and region into the packaging area coordinates, enabling subsequent multi-optical path images to be correlated and mutually verified for defects under the same packaging area coordinates.

[0051] Furthermore, in one embodiment of the present invention, the plurality of packaging regions include a chip window region, a gold wire channel region, a colloid center region, a colloid edge region, a cup wall region, and an overflow inspection region; the step of establishing the packaging region coordinates comprising the plurality of packaging regions includes:

[0052] Determine the center of the cup rim based on the described cup rim boundary;

[0053] The center of the chip's light-emitting area is determined based on the chip's light-emitting area;

[0054] A positioning reference is determined from the center of the cup rim, the center of the chip light-emitting area, and the outer edge of the adhesive surface, and the image coordinates of the LED bead to be tested are normalized according to the positioning reference.

[0055] The multiple encapsulation regions are divided into normalized image coordinates to generate the coordinates of the encapsulation regions.

[0056] Specifically, the image processing unit fits a circle or ellipse based on the cup rim boundary, and uses the center of the fitted circle or the center of the fitted ellipse as the cup rim center; it determines the center of the chip light-emitting area based on the center of the circumscribed rectangle of the connected domain of the chip light-emitting area or the gray-level centroid; and it determines the adhesive coverage area based on the outer edge of the adhesive surface. The positioning reference can be formed solely by the cup rim center, or it can be formed jointly by the cup rim center, the chip light-emitting area center, and the outer edge of the adhesive surface. In this embodiment, the image processing unit uses the cup rim center as the global origin, the cup rim radius or the adhesive surface outer edge radius as the normalization scale, and the direction of the chip light-emitting area center relative to the cup rim center as the angular reference to normalize the image coordinates of the LED bead to be detected.

[0057] In the normalized image coordinates, the chip window area corresponds to the chip's light-emitting area and its adjacent colloid-covered area; the gold wire channel area corresponds to the gold wire extension path and its surrounding preset width range; the colloid center area corresponds to the main encapsulation colloid area located away from the outer edge of the colloid surface and outside the chip window area; the colloid edge area corresponds to the preset radial width range inside and outside the outer edge of the colloid surface; the cup wall area corresponds to the reflective or structural transition area near the cup rim boundary; and the overflow inspection area corresponds to the area outside the cup wall area used to determine whether the colloid has overflowed. If a local image is overexposed or has missing boundaries, the image processing unit uses the equipment calibration parameters of the same batch of LEDs and the currently identified abnormal boundary segments of the LEDs to complete the corresponding encapsulation area, avoiding the inability to assign subsequent defect locations due to local reflections.

[0058] S2. Acquire multi-optical-path images in the coordinates of the encapsulation area. The multi-optical-path images include a coaxial diffuse image, a left oblique incident dark field image, a right oblique incident dark field image, and a low-current self-emissive image.

[0059] Specifically, after establishing the coordinates of the encapsulation area, the detection controller acquires multi-path images according to a preset acquisition sequence. When acquiring coaxial diffuse images, the coaxial diffuse light source is turned on to obtain the contour of the adhesive surface, the reflection of the cup wall, the contour of the gold wire channel, and local high-reflection areas. When acquiring left oblique incident dark field images, the left oblique incident light source is turned on to highlight minor surface contamination, scratches, and edge scattering. When acquiring right oblique incident dark field images, the right oblique incident light source is turned on to form directional verification with the left oblique incident dark field images. When acquiring low-current self-emissive images, the external light source is turned off, and the LED bead under test is driven to emit light by the low-current pulse drive unit to observe the occlusion, scattering, and attenuation caused by the colloid on the self-emissive path.

[0060] After multi-path image acquisition is completed, the image processing unit registers each image according to the coordinates of the packaging area. During registration, a positioning reference formed by the center of the cup, the center of the chip's light-emitting area, the outer edge of the adhesive surface, or a combination thereof, can be used to map the coaxial diffuse image, the left and right oblique incident dark field image, and the low-current self-emissive image to a unified coordinate system. For local edge position deviations caused by differences in the light source angle, the image processing unit does not directly treat them as defects, but retains them in the subsequent packaging optical path response verification matrix, and uses them together with the self-emissive attenuation value and the packaging area number for judgment.

[0061] Furthermore, in one embodiment of the present invention, the low-current self-emissive image is acquired after the external light source is turned off; when acquiring the low-current self-emissive image, the LED bead to be detected is driven in a short pulse mode, and the short pulse driving timing is synchronized with the camera exposure timing; the driving current of the short pulse mode is lower than the preset temperature rise influence current, which is determined by the device calibration parameters and is used to make the low-current self-emissive image characterize the blocking, scattering and attenuation of the light emission path by the encapsulating colloid.

[0062] Specifically, before acquiring low-current self-emissive images, the detection controller shuts down the coaxial diffuse light source, the left-oblique-incident dark-field light source, and the right-oblique-incident dark-field light source, ensuring that the main light signal received by the camera comes from the self-emission of the LED bead under test. The low-current pulse drive unit outputs a short pulse drive current, with the pulse width and the camera exposure window synchronized by the same trigger signal, ensuring that the exposure window covers the stable self-emissive segment while avoiding the rising and falling edges of the drive. The preset temperature rise-affecting current is determined by the equipment calibration parameters. For example, by calibrating the colloidal temperature rise and brightness fluctuations of qualified LED beads under different drive currents, an upper limit of current is selected that will not cause changes in colloidal temperature rise and affect image judgment.

[0063] Under these acquisition conditions, the local attenuation in the low-current self-emissive image mainly reflects the obstruction, scattering, and attenuation of the light emission path caused by bubbles, cracks, abnormal adhesive content, or gold wire blockage inside the encapsulation colloid, rather than being caused by external illumination reflection or thermal changes due to operating current. If the short pulse trigger fails, the exposure is out of sync, or the overall brightness of the self-emissive image is below the device's allowable range, the image processing unit marks the frame as an invalid acquisition frame and re-executes the low-current self-emissive image acquisition to prevent invalid frames from entering the encapsulation optical path response verification matrix.

[0064] S3. Based on the regional response reference corresponding to the coordinates of the encapsulation region, extract abnormal connected components from the multi-optical path image as suspected defects. Generate an encapsulation optical path response verification matrix according to the optical path number, encapsulation region number, and the abnormal connected component position, response intensity, centroid coordinates, and self-emission attenuation value of the suspected defects.

[0065] Specifically, the image processing unit establishes a regional response benchmark for each packaged area. The regional response benchmark can include the normal brightness range in the coaxial diffuse image, the normal scattering intensity range in the dark field image, the radial attenuation benchmark in the low-current self-emissive image, the edge gradient direction benchmark, and the color channel ratio benchmark. The regional response benchmark is determined based on samples of qualified LED chips from the same batch, the non-abnormal reference area of ​​the currently tested LED chip, or equipment calibration parameters, and is associated with and stored in relation to the packaged area number.

[0066] When extracting anomalous connected components, the image processing unit reads the corresponding region response benchmark according to the encapsulation region number within each optical path image and compares the response value of the current pixel or local window with this benchmark. For coaxial diffuse images, the focus is on extracting high-reflectivity contours, occlusion contours, and areas where the adhesive surface deviates from its edge; for left and right oblique incidence dark field images, the focus is on extracting high-scattering or directional scattering anomalous regions; for low-current self-emission images, the focus is on extracting local attenuation regions and regions with varying radial attenuation ranges. Adjacent anomalous pixels form anomalous connected components after connected component analysis, and each anomalous connected component serves as a potential defect candidate.

[0067] In one embodiment, the response intensity can be determined as follows: .in, This represents the average response value within the currently abnormal connected component. This represents the mean regional response baseline of the corresponding packaged area. This represents the standard deviation of the baseline response for the corresponding package area. To prevent constants with a denominator of zero, the response intensity is written into the abnormal response record for subsequent determination of whether a suspected defect has reached the surface scattering threshold, the threshold for reassessment, or other rule-based thresholds.

[0068] In a low-current self-emissive optical path, the self-emissive attenuation value can be determined as follows: .in, This indicates the self-illuminating reference brightness of the corresponding packaged area or an adjacent reference area without abnormalities. The emissivity of the region representing the location of the abnormal connected component. To prevent constants with a denominator of zero, the self-luminescence attenuation value is used to distinguish between surface scattering anomalies and internal attenuation anomalies, and also to determine whether gold wire shading anomalies form localized linear attenuation.

[0069] Furthermore, in one embodiment of the present invention, the encapsulated optical path response verification matrix uses optical path numbers as matrix rows and encapsulation region numbers as matrix columns; the optical path numbers include at least a coaxial diffuse optical path, a left oblique incident dark field optical path, a right oblique incident dark field optical path, and a low-current self-emissive optical path; the matrix units of the encapsulated optical path response verification matrix are used to store at least one abnormal response record, each abnormal response record including the optical path number, encapsulation region number, abnormal connected component position, response intensity, centroid coordinates, edge gradient direction, and color channel ratio under the corresponding optical path; wherein, the abnormal response record corresponding to the low-current self-emissive optical path also includes a self-emissive attenuation value.

[0070] Specifically, the encapsulated optical path response verification matrix is ​​not merely a simple table indicating the presence or absence of anomalies, but a structured technical object used to express the correspondence between the same suspected defect across optical paths. The matrix rows are determined by the optical path number, and the matrix columns are determined by the encapsulation region number. A single matrix cell can store zero or more anomaly response records to accommodate situations where multiple anomalous connected regions exist within the same encapsulation region. Each anomaly response record includes at least the record number, optical path number, encapsulation region number, location of the anomalous connected region, response intensity, centroid coordinates, edge gradient direction, and color channel ratio.

[0071] The location of anomaly connected components can be represented by a set of pixel coordinates, a contour polygon, or a bounding rectangle; the centroid coordinates are obtained by weighting the pixel coordinates within the anomaly connected component according to the response value; the edge gradient direction is statistically obtained from the gradient direction at the boundary of the anomaly connected component; the color channel ratio is determined by the color channel response in the coaxial diffuse image or the low-current self-emission image. Anomaly response records corresponding to low-current self-emission optical paths also include self-emission attenuation values. For anomaly response records of non-low-current self-emission optical paths, self-emission attenuation values ​​are not mandatory; when subsequently aggregating the same suspected defect, the dark field or coaxial diffuse response is associated with the low-current self-emission response through the location of the anomaly connected component and the centroid coordinates.

[0072] S4. Based on the response consistency of the same suspected defect in the encapsulated optical path response mutual verification matrix among the dark field image, the self-illuminating image and the encapsulated area, generate a defect mutual verification object.

[0073] Specifically, the image processing unit traverses the abnormal response records in the packaged optical path response verification matrix and determines whether the abnormal responses in different optical paths belong to the same suspected defect based on the location of the abnormal connected components, centroid coordinates, packaged region number, and response intensity change trend. The same suspected defect does not need to produce abnormal responses in all optical paths; for example, surface contamination or scratches usually have obvious responses in left and right oblique incidence dark field images, while the self-emission attenuation is weak in low current self-emission images; bubbles or cracks usually cause direction-related centroid relative shifts in dark field images and cause local attenuation in low current self-emission images.

[0074] When generating a defect verification object, the image processing unit uses the suspected defect as the core and writes the set of abnormal response records participating in the verification, the corresponding encapsulation region number, the response consistency type, the candidate response type, and the severity criteria into the same object. The defect verification object is used to carry out the rule matching process of S5, so that subsequent rules do not directly deal with isolated single-optical-path abnormal connected domains, but rather with technical objects that have already completed cross-optical-path verification.

[0075] Furthermore, in one embodiment of the present invention, generating a defect verification object based on the response consistency of the same suspected defect in the encapsulated optical path response verification matrix among the dark field image, the self-illuminating image, and the encapsulated region includes:

[0076] Based on the overlapping relationships of abnormal connected domain positions, the proximity relationships of centroid coordinates, the consistency relationships of encapsulation region numbers, and the changing trends of response intensity in the abnormal response records, abnormal response records belonging to the same suspected defect are grouped into a set of abnormal response records for mutual verification.

[0077] The response consistency type is determined based on the set of abnormal response records participating in the mutual authentication;

[0078] Based on the suspected defect, the encapsulation region number, the set of abnormal response records participating in the mutual verification, the response consistency type, the candidate response type, and the severity, the defect mutual verification object is generated.

[0079] Specifically, when collecting abnormal response records, the image processing unit first compares the overlap between the locations of abnormal connected regions in different optical paths. If the intersection-union ratio (IUU) of two abnormal connected regions reaches the region overlap threshold, or if their centroid coordinate distance is less than the proximity threshold of the corresponding encapsulation region, then both are considered as candidate responses for the same suspected defect. If the abnormal connected regions are located in adjacent encapsulation regions, such as the colloid edge region and the overflow inspection region, then it is further determined whether their contours are continuous, whether the edge gradient directions are compatible, and whether the response intensity changes continuously along the outer edge of the colloid surface.

[0080] Response consistency types can include dark field unidirectional enhancement, local attenuation of self-emission, relative shift of centroids at left and right oblique incidence, deviation of adhesive surface profile, continuous overflow response, and linear attenuation of gold wire channels. Candidate response types are initially determined based on the set of anomalous response records participating in mutual verification. For example, a strong dark field with weak self-emission results in a surface scattering candidate response type, while a relative shift of the dark field centroid and significant self-emission attenuation result in an internal attenuation candidate response type. The severity can be determined by the area of ​​the anomalous connected region, the peak response intensity, the self-emission attenuation value, the profile deviation, and the weight of the encapsulation region. The final severity is further determined in S5 by combining preset optical path response rules.

[0081] S5. Match the defect verification object with a preset optical path response rule that includes multiple response types to determine the defect type, the encapsulation area where the defect is located, the contour mask, and the severity.

[0082] Specifically, the defect determination process is executed by the image processing unit or the defect determination module. The image processing unit reads the set of abnormal response records, response consistency type, candidate response type, encapsulation region number, and severity criteria from the defect verification object, and calls the preset optical path response rules for matching. The preset optical path response rules are not black-box rules, but a set of rules that at least include surface scattering type, internal attenuation type, contour deviation type, overflow continuity type, and gold line occlusion type.

[0083] When a defect verification object matches a preset optical path response rule, the image processing unit outputs the defect type according to the rule and converts the location of the abnormal connected region with the highest response intensity or the best match to the rule in the defect verification object into a contour mask. The encapsulation region where the defect is located is determined by the corresponding encapsulation region number; if the defect verification object spans two encapsulation regions, the main encapsulation region is determined according to the area ratio of the abnormal connected region, the main response intensity region, and the rule type, while adjacent encapsulation regions can be retained as auxiliary positions. The severity can be determined based on the area ratio, response intensity, self-emission attenuation value, and contour deviation, for example, by writing the detection results as low, medium, and high levels.

[0084] Furthermore, in one embodiment provided by the present invention, the preset optical path response rules include surface scattering rules, internal attenuation rules, contour deviation rules, overflow continuity rules, and gold line occlusion rules.

[0085] The surface scattering rule is used to determine the corresponding defect verification object as surface contamination or scratch when the response intensity of the suspected defect in the left oblique incident dark field image and the right oblique incident dark field image reaches the surface scattering threshold and the self-luminescence attenuation value is lower than the internal defect judgment threshold.

[0086] The internal attenuation rule is used to determine the corresponding defect verification object as a bubble or crack when the centroid coordinates of the suspected defect in the left oblique incident dark field image and the right oblique incident dark field image form a relative offset related to the corresponding incident direction, and the self-luminous attenuation value reaches the internal defect judgment threshold.

[0087] The contour deviation rule is used to determine the corresponding defect verification object as excess glue, insufficient glue, or leakage glue when the abnormal connected domain position in the edge region of the colloid coincides with the outer edge of the glue surface, and the contour deviation range or the self-luminous radial attenuation range reaches the corresponding benchmark.

[0088] The overflow continuity rule is used to identify the corresponding defect verification object as glue overflow when a continuous high-response area is formed between the overflow inspection area and the cup wall area.

[0089] The gold wire occlusion rule is used to identify the corresponding defect verification object as a gold wire occlusion anomaly when the gold wire channel region forms a linear high reflectivity or occlusion contour in the coaxial diffuse image, forms a local linear attenuation in the low current self-luminous image, and the dark field scattering response does not meet the surface scattering rule.

[0090] Specifically, for surface scattering type rules, the image processing unit reads the response intensity of the corresponding abnormal response record in the left oblique incidence dark field image and the right oblique incidence dark field image, respectively. When the response intensity in both dark field directions reaches the surface scattering threshold, or one direction reaches the surface scattering threshold and the other direction reaches the auxiliary scattering threshold, and the self-emission attenuation value at the associated position in the low-current self-emission image is lower than the internal defect judgment threshold, it indicates that the anomaly is mainly manifested as surface scattering rather than internal occlusion. The image processing unit determines the corresponding defect verification object as surface contamination or scratch.

[0091] For internal attenuation type rules, the image processing unit compares the centroid coordinates of associated abnormal connected regions in the left-oblique-incident dark-field image and the right-oblique-incident dark-field image. If the centroid coordinates of the two images show a relative offset related to the corresponding incident direction, and the self-emission attenuation value in the low-current self-emission image reaches the internal defect judgment threshold, it indicates that the suspected defect is located inside the encapsulant or attenuates the emission path. At this time, the image processing unit determines the corresponding defect verification object as a bubble or crack, and further outputs a contour mask based on whether the abnormal connected region contour is circular, cracked, or irregular striped.

[0092] For contour deviation rules, the image processing unit reads the outer edge reference of the adhesive surface and the position of the abnormal connected region in the adhesive edge region, and calculates the deviation of the current outer edge of the adhesive surface relative to the region response reference. If the adhesive edge expands outward and the self-emission radial attenuation range expands outward, it is identified as a candidate for excess adhesive; if the adhesive edge contracts inward and the self-emission radial attenuation range is compressed, it is identified as a candidate for insufficient adhesive; if there is a local missing adhesive surface that causes abnormal light transmission or attenuation breakage in the chip window area or the adhesive edge region, it is identified as a candidate for missing adhesive. After the above candidates meet the corresponding references, the system outputs whether the adhesive is excess, insufficient, or missing.

[0093] For overflow continuity rules, the image processing unit determines whether a spatially continuous high-response region is formed between the overflow inspection area and the cup wall area. If the high-response region extends from the colloid edge area across the cup wall area to the overflow inspection area, and this continuous region has a continuous outline in the coaxial diffuse image or dark field image, then the corresponding defect verification object is identified as overflow. If only an isolated reflective point appears in the cup wall area and is not continuous with the overflow inspection area, it is not treated as overflow.

[0094] For the gold wire occlusion type rule, the image processing unit reads the linear high-reflectance or occlusion contour in the coaxial diffuse image within the gold wire channel region, and reads the local linear attenuation at the same channel location in the low-current self-emission image. If the linear attenuation is consistent with the direction of the gold wire channel, and the scattering response in the left and right obliquely incident dark field image does not satisfy the surface scattering type rule, it indicates that the anomaly is not a surface foreign object or scratch, but rather that the gold wire channel occludes or abnormally reflects the light emission path. The image processing unit then identifies the corresponding defect verification object as a gold wire occlusion anomaly.

[0095] Furthermore, in one embodiment of the present invention, the corresponding benchmarks for the surface scattering threshold, the internal defect judgment threshold, the contour deviation range, and the self-luminous radial attenuation range are determined based on at least one of the following: qualified LED bead samples from the same batch, the non-abnormal reference area of ​​the current LED bead to be tested, or equipment calibration parameters.

[0096] Specifically, qualified LED samples from the same batch are used to establish the normal response range at the batch level. This includes metrics such as the mean, standard deviation, radial decay curve, and edge gradient distribution of each packaged area in coaxial diffuse images, dark field images, and low-current self-emissive images. An anomaly-free reference area for the currently tested LED is used to correct for local brightness differences within the same LED. For example, a boundary segment in the cup wall region without abnormal connected regions or a stable response area in the colloidal center region can be selected as a reference. Equipment calibration parameters are used to limit hardware conditions such as camera exposure, light source brightness, oblique incidence angle, low-current pulse drive current, and optical magnification. These sources collectively define thresholds and benchmarks, ensuring that the surface scattering threshold, internal defect judgment threshold, contour deviation range, and self-emissive radial decay range are not arbitrary values, but rather technical parameters corresponding to the batch, the current sample, and the equipment status.

[0097] Furthermore, in one embodiment of the present invention, if the same defect verification object simultaneously satisfies multiple preset optical path response rules, or does not satisfy any preset optical path response rule but its response intensity reaches the threshold for re-judgment, then the defect verification object is marked as a defect verification object to be re-judged; supplementary acquisition is performed on the encapsulation area corresponding to the defect verification object to be re-judged, and the defect type is re-determined after updating the encapsulation optical path response verification matrix based on the image obtained from the supplementary acquisition; the supplementary acquisition includes at least one of adjusting the oblique incidence angle, adjusting the self-illumination exposure time, or reducing the local acquisition window.

[0098] Specifically, when a defect verification object simultaneously satisfies both the internal attenuation rule and the contour deviation rule, or simultaneously satisfies both the surface scattering rule and the gold line occlusion rule, the image processing unit does not immediately output the final defect type, but instead marks it as a defect verification object to be re-evaluated. If a defect verification object does not satisfy any preset optical path response rule, but its response intensity reaches the re-evaluation threshold, it indicates that the suspected defect should not be directly ignored and also enters the re-evaluation process.

[0099] During supplementary data acquisition, the detection controller determines the acquisition range based on the encapsulation area corresponding to the defect verification object to be reassessed. For objects with insufficient dark field directionality, the oblique incidence angle is adjusted before reacquiring a left-oblique-incident dark field image or a right-oblique-incident dark field image; for objects with unclear self-illumination attenuation boundaries, the self-illumination exposure time is adjusted before reacquiring a low-current self-illumination image; for local small-sized defects, the local acquisition window is reduced to increase the local sampling density. The supplementary acquired images are entered into the original encapsulation area coordinates, and new abnormal response records are generated. The image processing unit updates the encapsulation optical path response verification matrix with the new abnormal response records, and then re-executes the defect verification object generation and preset optical path response rule matching to obtain the reassessed defect type.

[0100] In one example of a through-processing operation, taking a suspected defect in the colloidal central region of an LED bead under test as an example, the image processing unit first determines the center of the cup rim and the center of the chip's light-emitting area in the coaxial diffuse positioning image, and numbers the colloidal central region as A3. Subsequently, coaxial diffuse images, left oblique incidence dark field images, right oblique incidence dark field images, and low-current self-emissive images are acquired. Based on the response benchmark of region A3, an abnormal connected component P1 is extracted from the left oblique incidence dark field image, with a response intensity of 3.6 and centroid coordinates of [missing information]. An anomalous connected component P2 was extracted from the right-oblique-incidence dark-field image, with a response intensity of 3.4 and centroid coordinates of [missing information]. The self-emission attenuation value at the corresponding position in the low-current self-emission image is 0.24. If the surface scattering threshold is 2.8 and the internal defect judgment threshold is 0.18 in this embodiment, then although P1 and P2 both have dark field responses, their centroid coordinates form a relative offset related to the left and right incident directions, and the self-emission attenuation value reaches the internal defect judgment threshold. The image processing unit determines the corresponding defect verification object as a bubble or a crack.

[0101] In another re-evaluation example, a continuous high-response region appears between the colloid edge area and the overflow inspection area, but the radial attenuation range of this region in the low-current self-illuminating image also shows local expansion. During the initial matching, the image processing unit triggers both the contour deviation rule and the overflow continuity rule, thus marking this defect verification object as a defect verification object to be re-evaluated. After the detection controller narrows the local acquisition window and re-acquires the coaxial diffuse image and dark field image, if the supplementary image shows that the high-response region extends continuously from the cup wall area to the overflow inspection area, the updated encapsulation optical path response verification matrix supports the overflow continuity rule, and the final output defect type is overflow; if the supplementary image shows that the high-response region is only located in the colloid edge area and does not cross the cup wall area, the output defect type is excess glue.

[0102] Please see Figure 6 The present invention also provides an LED chip packaging inspection system for implementing the above-mentioned LED chip packaging inspection method, comprising:

[0103] The packaging area coordinate establishment module is used to acquire the coaxial diffuse positioning image of the LED bead to be tested, identify the cup edge, chip light-emitting area, gold wire channel and outer edge of adhesive surface, and establish the packaging area coordinates containing multiple packaging areas;

[0104] A multi-optical-path image acquisition module is used to acquire multi-optical-path images under the coordinates of the encapsulation area. The multi-optical-path images include a coaxial diffuse image, a left oblique incident dark field image, a right oblique incident dark field image, and a low-current self-emissive image.

[0105] The mutual verification matrix generation module is used to extract abnormal connected components as suspected defects from the multi-optical path image based on the regional response reference corresponding to the coordinates of the encapsulation area, and generate an encapsulation optical path response mutual verification matrix according to the optical path number, encapsulation area number, and the abnormal connected component position, response intensity, centroid coordinates and self-emission attenuation value of the suspected defect.

[0106] The mutual verification object generation module is used to generate a defect mutual verification object based on the response consistency of the same suspected defect in the encapsulated optical path response mutual verification matrix among the dark field image, the self-illuminating image and the encapsulated region.

[0107] The defect result determination module is used to match the defect verification object with a preset optical path response rule containing multiple response types to determine the defect type, the encapsulation area where the defect is located, the contour mask, and the severity.

[0108] Specifically, the encapsulation region coordinate establishment module is connected to the camera and coaxial diffuse light source to output the encapsulation region coordinates; the multi-optical path image acquisition module is connected to the coaxial diffuse light source, left and right obliquely incident dark field light sources, low-current pulse drive unit, and camera to acquire multi-optical path images under the same encapsulation region coordinates; the mutual verification matrix generation module reads the encapsulation region coordinates and multi-optical path images to generate an encapsulation optical path response mutual verification matrix containing abnormal response records; the mutual verification object generation module reads the encapsulation optical path response mutual verification matrix and generates defect mutual verification objects; the defect result determination module reads the defect mutual verification objects and preset optical path response rules, and outputs the defect type, the encapsulation region where the defect is located, the contour mask, and the severity. These modules can be deployed in the same image detection host or implemented collaboratively by the camera controller, light source controller, and image processing server.

[0109] In the experimental verification, samples of the same batch of LED beads were selected for defect detection. The control scheme used a single optical path detection method, determining the defect type solely based on the response intensity of abnormal connected components in a single image. The present invention uses encapsulation region coordinates, multi-optical path images, encapsulation optical path response cross-verification matrix, defect cross-verification objects, and preset optical path response rules for judgment. In the experiment, the false positive rate was defined as the proportion of normal structures or non-target responses judged as defects; the false negative rate was defined as the proportion of labeled defects not detected; and the type confusion rate was defined as the proportion of detected defects judged as other defect types. Figure 4 As can be seen, the misjudgment rate, missed judgment rate, and type confusion rate of the control scheme are 7.8%, 9.4%, and 11.6%, respectively, while those of the present invention are 2.1%, 2.8%, and 3.2%, respectively, representing a reduction of 5.7 percentage points, 6.6 percentage points, and 8.4 percentage points.

[0110] Figure 5A confusion matrix for defect type identification is shown. During experimental statistics, for ease of matrix display, excess glue, insufficient glue, and leakage glue were grouped into the abnormal glue quantity group; however, actual test results can still be output as excess glue, insufficient glue, or leakage glue. Figure 5 As can be seen, the diagonal recognition rates for surface contamination or scratches, bubbles or cracks, abnormal adhesive content, adhesive overflow, and abnormal gold wire obstruction are 92%, 91%, 90%, 92%, and 92%, respectively, while the maximum confusion rate for non-diagonal defects is 5%. These results demonstrate that by cross-validating the left and right obliquely incident dark-field images, the self-illumination attenuation value, and the encapsulation area, this invention can reduce the confusion caused by cup wall reflection, transparent colloid curvature, and localized obstruction of gold wire channels.

[0111] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0112] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for testing LED chip packaging, characterized in that, include: S1. Obtain the coaxial diffuse positioning image of the LED bead to be tested, identify the cup edge, chip light-emitting area, gold wire channel and outer edge of adhesive surface, and establish the coordinates of the packaging area containing multiple packaging areas; S2. Acquire multi-optical-path images in the coordinates of the encapsulation area. The multi-optical-path images include a coaxial diffuse image, a left oblique incident dark field image, a right oblique incident dark field image, and a low-current self-emissive image. S3. Based on the regional response reference corresponding to the coordinates of the encapsulation region, extract abnormal connected components from the multi-optical path image as suspected defects. Generate an encapsulation optical path response verification matrix according to the optical path number, encapsulation region number, and the abnormal connected component position, response intensity, centroid coordinates, and self-emission attenuation value of the suspected defects. S4. Based on the response consistency of the same suspected defect in the encapsulated optical path response mutual verification matrix among the dark field image, the self-illuminating image and the encapsulated area, generate a defect mutual verification object. S5. Match the defect verification object with a preset optical path response rule that includes multiple response types to determine the defect type, the encapsulation area where the defect is located, the contour mask, and the severity.

2. The LED bead packaging testing method according to claim 1, characterized in that, The multiple packaging areas include a chip window area, a gold wire channel area, a colloid center area, a colloid edge area, a cup wall area, and an overflow inspection area; The establishment of the packaging region coordinates, which includes multiple packaging regions, includes: Determine the center of the cup rim based on the described cup rim boundary; The center of the chip's light-emitting area is determined based on the chip's light-emitting area; A positioning reference is determined from the center of the cup rim, the center of the chip light-emitting area, and the outer edge of the adhesive surface, and the image coordinates of the LED bead to be tested are normalized according to the positioning reference. The multiple encapsulation regions are divided into normalized image coordinates to generate the coordinates of the encapsulation regions.

3. The LED bead packaging testing method according to claim 1, characterized in that, The low-current self-emissive image is acquired after the external light source is turned off. When acquiring the low-current self-emissive image, the LED bead to be tested is driven in a short pulse mode, and the short pulse driving timing is synchronized with the camera exposure timing. The driving current of the short pulse mode is lower than the preset temperature rise influence current, which is determined by the device calibration parameters and is used to make the low-current self-emissive image characterize the blocking, scattering and attenuation of the light emission path by the encapsulating colloid.

4. The LED bead packaging testing method according to claim 1, characterized in that, The encapsulated optical path response verification matrix uses optical path numbers as matrix rows and encapsulation region numbers as matrix columns. The optical path numbers include at least a coaxial diffuse optical path, a left oblique incident dark field optical path, a right oblique incident dark field optical path, and a low-current self-emissive optical path. The matrix units of the encapsulated optical path response verification matrix are used to store at least one abnormal response record. Each abnormal response record includes the optical path number, encapsulation region number, abnormal connected component location, response intensity, centroid coordinates, edge gradient direction, and color channel ratio under the corresponding optical path. Among them, the abnormal response record corresponding to the low-current self-emissive optical path also includes the self-emissive attenuation value.

5. The LED bead packaging testing method according to claim 4, characterized in that, The step of generating a defect verification object based on the response consistency of the same suspected defect in the encapsulated optical path response verification matrix across the dark field image, the self-illuminating image, and the encapsulated region includes: Based on the overlapping relationships of abnormal connected domain positions, the proximity relationships of centroid coordinates, the consistency relationships of encapsulation region numbers, and the changing trends of response intensity in the abnormal response records, abnormal response records belonging to the same suspected defect are grouped into a set of abnormal response records for mutual verification. The response consistency type is determined based on the set of abnormal response records participating in the mutual authentication; Based on the suspected defect, the encapsulation region number, the set of abnormal response records participating in the mutual verification, the response consistency type, the candidate response type, and the severity, the defect mutual verification object is generated.

6. The LED bead packaging testing method according to claim 1, characterized in that, The preset optical path response rules include surface scattering rules, internal attenuation rules, contour deviation rules, overflow continuity rules, and gold line occlusion rules. The surface scattering rule is used to determine the corresponding defect verification object as surface contamination or scratch when the response intensity of the suspected defect in the left oblique incident dark field image and the right oblique incident dark field image reaches the surface scattering threshold and the self-luminescence attenuation value is lower than the internal defect judgment threshold. The internal attenuation rule is used to determine the corresponding defect verification object as a bubble or crack when the centroid coordinates of the suspected defect in the left oblique incident dark field image and the right oblique incident dark field image form a relative offset related to the corresponding incident direction, and the self-luminous attenuation value reaches the internal defect judgment threshold. The contour deviation rule is used to determine the corresponding defect verification object as excess glue, insufficient glue, or leakage glue when the abnormal connected domain position in the edge region of the colloid coincides with the outer edge of the glue surface, and the contour deviation range or the self-luminous radial attenuation range reaches the corresponding benchmark. The overflow continuity rule is used to identify the corresponding defect verification object as glue overflow when a continuous high-response area is formed between the overflow inspection area and the cup wall area. The gold wire occlusion rule is used to identify the corresponding defect verification object as a gold wire occlusion anomaly when the gold wire channel region forms a linear high reflectivity or occlusion contour in the coaxial diffuse image, forms a local linear attenuation in the low current self-luminous image, and the dark field scattering response does not meet the surface scattering rule.

7. The LED bead packaging testing method according to claim 6, characterized in that, The corresponding benchmarks for the surface scattering threshold, the internal defect judgment threshold, the contour deviation range, and the self-luminous radial attenuation range are determined based on at least one of the following: qualified LED bead samples from the same batch, the non-abnormal reference area of ​​the current LED bead to be tested, or the equipment calibration parameters.

8. The LED bead packaging testing method according to claim 1, characterized in that, If the same defect verification object simultaneously meets multiple preset optical path response rules, or does not meet any preset optical path response rule but its response intensity reaches the threshold for re-judgment, then the defect verification object is marked as a defect verification object to be re-judged; supplementary acquisition is performed on the encapsulation area corresponding to the defect verification object to be re-judged, and the defect type is re-determined after updating the encapsulation optical path response verification matrix based on the image obtained from the supplementary acquisition; the supplementary acquisition includes at least one of adjusting the oblique incidence angle, adjusting the self-illumination exposure time, or reducing the local acquisition window.

9. An LED bead packaging and testing system, characterized in that, A method for testing LED bead packaging according to any one of claims 1 to 8 includes: The packaging area coordinate establishment module is used to acquire the coaxial diffuse positioning image of the LED bead to be tested, identify the cup edge, chip light-emitting area, gold wire channel and outer edge of adhesive surface, and establish the packaging area coordinates containing multiple packaging areas; A multi-optical-path image acquisition module is used to acquire multi-optical-path images under the coordinates of the encapsulation area. The multi-optical-path images include a coaxial diffuse image, a left oblique incident dark field image, a right oblique incident dark field image, and a low-current self-emissive image. The mutual verification matrix generation module is used to extract abnormal connected components as suspected defects from the multi-optical path image based on the regional response reference corresponding to the coordinates of the encapsulation area, and generate an encapsulation optical path response mutual verification matrix according to the optical path number, encapsulation area number, and the abnormal connected component position, response intensity, centroid coordinates and self-emission attenuation value of the suspected defect. The mutual verification object generation module is used to generate a defect mutual verification object based on the response consistency of the same suspected defect in the encapsulated optical path response mutual verification matrix among the dark field image, the self-illuminating image and the encapsulated region. The defect result determination module is used to match the defect verification object with a preset optical path response rule containing multiple response types to determine the defect type, the encapsulation area where the defect is located, the contour mask, and the severity.

Images