A method and system for visually inspecting electronic components for defects

Abstract

The application discloses an electronic component visual defect detection method and system, and relates to the technical field of visual detection; the technical points are as follows: identity records, sidewall uncovered areas and supplementary visual angle patterns of target components are acquired, residual backfill permission results are formed according to cross-station identity closure results; carrier pattern repetition verification and component posture following verification are performed on candidate residuals admitted, and candidate residual sources are identified; when the candidate residual is not attributed to the carrier pattern, follows the posture of the target component and is connected with the sidewall uncovered area, sidewall defect evidence is generated, and a visual defect detection result is output; when the sources conflict, records are missing or the connection relationship is not established, a detection limited result is output, and writing into stable samples and benchmark updating is blocked; and the application can reduce the interference of carrier fixed textures and residual defects of adjacent bearing positions on sidewall defect judgment.

Description

Technical Field

[0001] This invention relates to the field of visual inspection technology, specifically to a method and system for visual defect detection of electronic components. Background Technology

[0002] As electronic components evolve towards miniaturization, high-density assembly, and high-speed shipping, surface mount capacitors, surface mount resistors, miniature packaged chips, and small connector terminals often utilize carrier bags, blister tray positioning cavities, and tape grooves as carrier units during incoming material inspection, tape packaging, and outgoing shipment verification. Visual image acquisition is performed during continuous transport. Existing visual inspection equipment can typically identify surface contamination, end-face defects, package chipping, and obvious scratches based on the component's main outline, surface grayscale changes, and visible solder joint boundaries. Taking the tape inspection of miniature surface mount capacitors as an example, the component falls into... After the carrier tape is pocketed, its upper surface can usually form a relatively complete appearance pattern in the top view. However, the side wall of the solder end near the inner wall of the pocket is easily affected by the cavity wall, the afterimage of the transparent cover tape, and local reflection. When the carrier tape moves at high speed and is accompanied by slight vibration, the component may undergo slight deflection inside the pocket, which further reduces the visible range of the solder end side wall, chamfer, and bottom corner in the top view. For defects such as unplated end face, pocket friction scratches, and slight chamfer chipping, their image features are often concentrated in the above-mentioned boundary areas. Relying solely on top-view inspection can easily lead to uncovered areas that are difficult to judge stably.

[0003] In existing technologies, to improve the coverage of electronic component appearance inspection, fixed region of interest detection, template matching, edge segmentation, and learning model recognition are commonly used. Additionally, oblique, side, and flip-over acquisition stations can be added beyond the top-view station to supplement side information that is difficult to obtain from a single perspective. These methods can improve the identification conditions of surface and end-face defects in conventional components and are applicable to most stable transport and attitude-controlled inspection scenarios. However, for micro-components transported at high speeds on carrier belts, it is necessary to confirm whether the uncovered sidewall area formed by the top-view station corresponds to the same component within the same pocket, compared to the sidewall residual acquired from the supplementary perspectives. If there are fluctuations in the carrier tape pitch, timing shifts in station triggering, cavities in the pocket for material replenishment, or sudden changes in component movement or posture at the same sidewall location, identification discrepancies may occur between images from different stations. At the same time, repeated scratches, dust indentations, and blister textures on the inner wall of the carrier tape pocket will form similar residuals in consecutive pockets, overlapping with component sidewall defects in the image. Therefore, if the solder end sidewall residuals appearing in the supplementary view lack cross-station identification closure criteria and carrier residual separation criteria, it is difficult to stably attribute them to the target electronic component, thus affecting the integrity of defect detection results and quality traceability after shipment. Summary of the Invention

[0004] To achieve the above objectives, the present invention is implemented through the following technical solution: A method for visual defect detection of electronic components includes the following specific steps: The system acquires multi-view pattern reading records of the carrier carrying electronic components during continuous transport. It uses carrier pattern and component appearance pattern separation and identification processing to form a target component identity record. The target component identity record is used to maintain the correspondence between the target carrier position identifier and the target component posture. Obtain the target component identity record, and based on the sidewall visibility projection processing, perform weld end sidewall visibility recognition on the top view pattern in the multi-view pattern reading record, generate the sidewall uncovered area, and keep the sidewall uncovered area as the area to be compensated for detection. Obtain the target component identity record and the uncovered area of ​​the side wall. Based on the cross-station identity closure processing, perform side wall residual identification and verification on the supplementary view pattern in the multi-view pattern reading record. When the side wall residual closes with the target component identity record, generate the residual backfill permission result. Candidate residuals corresponding to the residual backfill permission results are obtained. Source identification is performed by handling the competition between vehicle residuals and component residuals. Vehicle pattern repetition verification and component posture following verification are performed on the candidate residuals. Candidate residuals belonging to the vehicle pattern are excluded from the component defect candidates. Candidate residuals belonging to the target component and connected to the sidewall uncovered area are generated as sidewall defect evidence. The electronic component visual defect detection results are output based on the sidewall defect evidence.

[0005] Furthermore, forming a target component identity record includes: identifying the vehicle pattern from the multi-view pattern reading record and generating a vehicle identification record; identifying the component appearance pattern within the target bearing position defined by the vehicle identification record and generating a component identification record; wherein, the supplementary view pattern is only used for cross-workstation identity closure processing and does not participate in the validity confirmation of the target component identity record.

[0006] Furthermore, forming the target component identity record also includes: performing identity acceptance matching processing on the vehicle identification record and the component identification record to generate acceptance matching results between the candidate component corresponding to the component identification record and the target bearing position; when the acceptance matching result meets the acceptance permission requirements formed by the acceptance relationship of the target bearing position in the stable reading sample, a target component identity record with a valid identity status is generated; when the acceptance matching result does not meet the acceptance permission requirements, a target component identity record with a restricted identity status is generated, and it is prohibited to output qualified test results based on the target component identity record with a restricted identity status.

[0007] Furthermore, the weld end sidewall visibility identification includes: retrieving the top view diagram using the target component identity record as an index; when the target component identity record has an identity valid state, performing sidewall projection on the visible weld end boundary according to the target component posture under the top view diagram coordinates to form the sidewall projection position; when the target component identity record has an identity restricted state, generating a restricted detection area to be compensated, and not allowing it to enter the residual backfill permission processing.

[0008] Furthermore, generating the uncovered sidewall area includes: overlaying the sidewall projection position with the occlusion boundary in the top view; generating the uncovered sidewall area when at least one of the following conditions is met: the sidewall projection position overlaps with the occlusion boundary, or the continuous weld end boundary reading result is not formed; marking the sidewall orientation of the uncovered sidewall area according to the target component orientation; and maintaining it as the area to be compensated and detected.

[0009] Furthermore, the sidewall residual identification and verification includes: defining the retrieval object of the supplementary view pattern based on the target element identity record; forming a sidewall identification range in the supplementary view pattern based on the sidewall orientation of the sidewall-uncovered area; identifying candidate sidewall residuals within the sidewall identification range; and generating a residual backfilling restricted result when at least one of the following conditions is met: the supplementary view pattern does not match the target bearing position identifier in the target element identity record or the sidewall identification range is not formed.

[0010] Furthermore, generating residual backfill permission results includes: performing a closure check on the candidate sidewall residuals corresponding to the target bearing position identifier, target component attitude, and sidewall orientation; generating residual backfill permission results when all closure checks are successful; generating residual backfill restricted results when no candidate sidewall residuals are formed; and generating residual backfill restricted results when not all closure checks are successful, and excluding the corresponding candidate sidewall residuals from residual attribution competition processing.

[0011] Furthermore, the source identification process includes: performing a vehicle pattern repetition check on the candidate residuals based on the vehicle identification record, which verifies whether the candidate residuals repeat the vehicle pattern position defined by the vehicle identification record; performing a component attitude following check on the candidate residuals based on the target component attitude and sidewall orientation, which verifies whether the candidate residuals are located at the corresponding sidewall position defined by the target component attitude and sidewall orientation; and generating a residual attribution restriction result when at least one of the following conditions is met: the vehicle identification record cannot define the vehicle pattern position, or the target component attitude and sidewall orientation cannot define the corresponding sidewall position, without generating the corresponding candidate residual as evidence of a sidewall defect.

[0012] Furthermore, the output of the electronic component visual defect detection results includes: when the vehicle pattern repetition verification is successful but the component posture following verification is unsuccessful, the candidate residual is excluded from the component defect candidates; when the vehicle pattern repetition verification is unsuccessful, the component posture following verification is successful, and the candidate residual is connected to the uncovered area of ​​the sidewall, the candidate residual is generated as evidence of sidewall defects, and the electronic component visual defect detection results are output based on the sidewall defect evidence; when at least one of the following is successful: the vehicle pattern repetition verification and the component posture following verification are both successful; the vehicle pattern repetition verification and the component posture following verification are both unsuccessful; the vehicle pattern repetition verification is unsuccessful and the component posture following verification is successful but the candidate residual is not connected to the uncovered area of ​​the sidewall, a residual attribution restriction result is generated, and the corresponding candidate residual is not generated as evidence of sidewall defects.

[0013] A visual defect detection system for electronic components, comprising: The identity record forming module acquires multi-view pattern reading records of the carrier carrying electronic components during continuous transport. It uses carrier pattern and component appearance pattern separation and recognition processing to form target component identity record. The target component identity record is used to maintain the correspondence between the target carrier position identifier and the target component posture. The module for forming the area to be compensated acquires the identity record of the target component, performs weld end sidewall visibility recognition on the top view pattern in the multi-view pattern reading record based on sidewall visibility projection processing, generates the sidewall uncovered area, and keeps the sidewall uncovered area as the area to be compensated and detected. The residual backfilling permission module obtains the target component identity record and the uncovered area of ​​the side wall. Based on cross-station identity closure processing, it performs side wall residual identification and verification on the supplementary view pattern in the multi-view pattern reading record. When the side wall residual closes with the target component identity record, it generates a residual backfilling permission result. The defect result output module obtains the candidate residuals corresponding to the residual backfill permission result, performs source identification by handling the competition between vehicle residuals and component residuals, performs vehicle pattern repetition verification and component posture following verification on the candidate residuals, excludes the candidate residuals belonging to the vehicle pattern from the component defect candidates, generates the candidate residuals belonging to the target component and connected to the side wall uncovered area as side wall defect evidence, and outputs the electronic component visual defect detection result based on the side wall defect evidence.

[0014] This invention provides a method and system for visual defect detection of electronic components, which has the following beneficial effects: 1. This invention uses the residual backfill permission result as the only admission condition for source identification. Before generating sidewall defect evidence, it performs vehicle pattern repetition verification and component attitude following verification. This allows candidate residuals in the supplementary viewpoint pattern to first pass through source competition before entering the defect confirmation link. As a result, repetition residuals caused by vehicle edges, inner wall patterns of bearing positions, and adjacent bearing positions will not be directly interpreted as sidewall defects of the target component. This can correct the misjudgment problem caused by sidewall occlusion and unclear source of supplementary viewpoint residuals in traditional solutions, so that the output electronic component visual defect detection results have clear target component attribution and sidewall location attribution.

[0015] 2. This invention uses vehicle pattern repetition verification, component attitude tracking verification, and the relationship between the uncovered sidewall area and the target component as conditions for generating sidewall defect evidence. This ensures that candidate residuals must simultaneously satisfy the conditions of being a non-vehicle repetition source, following the target component attitude, and accepting the compensated sidewall area before they are converted into defect evidence. This process ensures that sidewall defect evidence originates from closed candidate residuals and defined target component identity records, avoiding the direct output of defect conclusions based solely on grayscale anomalies or edge residuals in the supplementary viewpoint. As a result, the compensation detection results of the occluded sidewall area can be consistent with the preceding identity closure, sidewall orientation, and uncovered area.

[0016] 3. This invention also excludes candidate residuals that are missing valid reading records of vehicle patterns, cannot form attitude boundaries, have conflicts between repeated verification and follow-up verification, or whose inheritance relationship is not established from the sidewall defect evidence generation link by setting residual attribution restrictions. It also prohibits them from entering stable reading sample records and subsequent benchmark updates. As a result, the system will not output highly reliable defect conclusions when there are missing measurements, delays, insufficient samples, or conflicting sources, nor will it write back uncertain residuals as stable samples. This can maintain the cleanliness of the visual inspection model and cross-station reading benchmarks, so that subsequent batch inspections are not continuously contaminated by erroneous residuals. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system framework diagram of the present invention. Detailed Implementation

[0018] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: Please see Figure 1This embodiment provides a method for visual defect detection of electronic components, including the following specific steps: S1 is used to form the target component identity record during continuous transport of the carrier, so that subsequent identification of uncovered sidewall areas, cross-station identity closure processing, and candidate residual assignment competition all revolve around the same target carrier position and the same target component posture. The system acquires multi-view pattern reading records of the carrier carrying the electronic component during continuous transport. In this step, the multi-view pattern reading records are used to read the top view, carrier pattern, and transport phase synchronized with the top view. The system generates supplementary view to be closed identifiers based on the target carrier position. When the supplementary view pattern has been acquired, the system only registers its station sequence identifier and reading time identifier for subsequent cross-station identity closure processing. It does not participate in the validity confirmation of the target component identity record. In the tape and reel shipment inspection scenario of micro chip capacitors, the top view can form the pattern on the upper surface of the component and the edge pattern of the carrier position. When the side wall of the solder end near the inner wall of the carrier position is blocked by the cavity wall and affected by local reflection, its side wall pattern appears as a partially missing state in the top view. The system first forms the target component identity record based on the top view and carrier pattern, so that the supplementary view pattern only enters S3 as a subsequent object to be closed.

[0020] The system identifies the carrier pattern from the multi-view pattern reading record and generates a carrier identification record. The carrier pattern is a readable pattern on the carrier used to define the position of the electronic component. The carrier identification record is used to characterize the position boundary and reading phase of the target carrier position during continuous transport. Specifically, the system reads the carrier position boundary and the positioning pattern that can characterize the continuous transport phase in the carrier pattern to form the boundary record and phase record corresponding to the target carrier position. The system also writes the reading time of the top view pattern corresponding to the target carrier position into the carrier identification record. In the tape carrier scenario, when the continuous carrier positions move with the transport mechanism, the edge of the carrier position forms a repeating structure in the top view pattern. When the electronic component is slightly deflected within the carrier position due to transport vibration, the system first identifies the target carrier position and then identifies the appearance pattern of the component within the target carrier position to suppress the inclusion of the inner wall texture of the carrier position, the cover tape afterimage, and the edge pattern of adjacent carrier positions into the target component range.

[0021] After generating the carrier identification record, the system identifies the component appearance pattern within the target carrier position defined by the carrier identification record and generates a component identification record. The component appearance pattern is the main outline of the electronic component within the target carrier position, the visible solder end boundary, and a readable appearance pattern that can characterize the target component's posture. For electronic components with clear polarity markings or asymmetrical solder end structures, the target component posture is formed by the polarity marking position, the main body's long axis direction, and the solder end connection direction. For symmetrical electronic components, the target component posture is formed by the center of the main body's outer contour, the visible solder end boundary, and the offset state relative to the target carrier position. In the scenario where a micro-surface capacitor is slightly tilted within the carrier position, its main body's outer contour is still within the target carrier position, and the distance between the two solder ends and the edge of the carrier position forms a difference. The system writes this difference into the target component posture in the component identification record, so that the subsequent sidewall visibility projection generates the sidewall uncovered area according to the target component posture. If the component appearance pattern is insufficient to support the target component posture recognition, the system writes the posture-restricted state into the component identification record and causes the corresponding candidate component to enter the identity-restricted processing.

[0022] The system performs identity matching processing on vehicle identification records and component identification records, generating matching results between candidate components and target bearing positions. The identity matching processing does not rely on a single contour coincidence result as the basis for identity confirmation. Instead, it checks whether the candidate component falls within the effective bearing range defined by the boundary of the target bearing position, whether the attitude state of the candidate component corresponds permissibly with the bearing direction of the target bearing position, and whether the reading time of the top view pattern of the candidate component corresponds with the reading phase of the target bearing position. If the appearance contour of the candidate component is close to the center of the target bearing position, but the reading phase points to an adjacent bearing position, the system treats the candidate component as a trigger misalignment object and generates an identity record of the target component with an identity-restricted state. If the reading phase of the candidate component is consistent with the target bearing position, but the component appearance pattern crosses the boundary of the target bearing position, the system treats the candidate component as a boundary contamination object and generates an identity record of the target component with an identity-restricted state.

[0023] The acceptance permission requirement is formed by the acceptance relationship of the target bearing position in the stable reading sample, and is used to limit whether the candidate component can be written into the target component identity record with a valid identity status. The stable reading sample is the sample record in which the carrier transport phase is continuous, the carrier pattern is read effectively, the component appearance pattern is not obstructed or mismatched, and the first piece inspection review record confirms that it is in a normal bearing state. The system obtains the normal acceptance relationship of the target component in the target bearing position from the stable reading sample, and forms the acceptance permission requirement for identity acceptance matching processing based on the same type of carrier, the same specification of electronic component, and the same inspection cycle. The equipment calibration record forms the limitation range of the acceptance permission requirement. The equipment calibration record includes the camera calibration record, the carrier bearing position design boundary, and the transport trigger calibration record, so that abnormal offsets, carrier contamination patterns, and trigger drift in the stable reading sample are not written into the acceptance permission requirement. When the inspection cycle, camera calibration record, or carrier specifications change, the system re-obtains the stable reading sample of the current batch and performs the current batch calibration on the acceptance permission requirement.

[0024] If there are insufficient stable read samples in the current batch, the system uses equipment calibration records, initial component pattern records that have passed the first article inspection and verification, and carrier pattern records that have been continuously and effectively read to form temporary acceptance requirements. Continuous and effective reading means that the carrier pattern can form the target bearing position boundary and reading phase during continuous transportation, and there is no reading state with missing bearing position, carrier contamination obstruction, or trigger misalignment. The temporary acceptance requirements and their screening results are only used for identity screening of the current batch and are not written into the stable read sample records. The target component identity record formed by the temporary acceptance requirements still needs to complete the cross-station identity closure processing in S3 before it can participate in the subsequent defect evidence generation. When the equipment calibration records, initial component pattern records, and carrier pattern records that have been continuously and effectively read cannot form temporary acceptance requirements, the system generates a target component identity record with an identity restricted state.

[0025] When the acceptance matching result meets the acceptance permission requirements, the system generates a target element identity record with a valid identity status. The target element identity record is used to maintain the correspondence between the target bearing position and the target element posture. It records at least the target bearing position identifier, the target element posture, the top view pattern reading time, the acceptance matching result, and the valid identity status. The target element identity record with a valid identity status is used as the object boundary for the side wall visibility projection processing in S2, enabling S2 to generate the side wall uncovered area around the same target element. In S3, it is used as the verification benchmark for cross-station identity closure processing, so that the side wall residual in the supplementary view pattern must be closed with the target element identity record before it can enter the residual backfill permission judgment. In the scenario where a suspected weld end side wall scratch is collected in the supplementary view station, the system first identifies whether the supplementary view pattern corresponds to the same target bearing position and the same target element posture in the previous top view station, and then controls whether the residual participates in the subsequent defect evidence generation based on the identification result.

[0026] When the acceptance matching result does not meet the acceptance permission requirements, the system generates a target element identity record with an identity restriction status. The target element identity record with an identity restriction status retains the reading object boundary and reading time of the target bearing position, which is used in subsequent steps to identify the existence of identity instability at the bearing position. The target element identity record with an identity restriction status cannot be used as the identity basis for directly outputting qualified test results. In high-speed transportation, if the target element jumps within the bearing position due to the vibration of the carrier, making the relationship between the main body outline in the top view and the target bearing position boundary unstable, the system records the reading object boundary of the target bearing position and sets the corresponding target element identity record to an identity restriction status, so that the subsequent side wall uncovered area and supplementary viewing angle residual cannot be directly used as the basis for outputting qualified test results.

[0027] Through the processing of S1, the vehicle pattern and the component appearance pattern are separated and identified. The vehicle identification record first defines the target bearing position, and the component identification record then defines the target component posture. The identity matching process further converts the two into a target component identity record. This target component identity record provides both the object boundary of the side wall visibility projection for S2 and the cross-station identity closure benchmark for the supplementary view pattern for S3. This allows subsequent side wall uncovered areas, residual backfilling permission results, and electronic component visual defect detection results to be traced back to the same target bearing position and the same target component posture. In this embodiment, in detection scenarios where vehicle edge reflection, slight component deflection, and interference from adjacent bearing position patterns coexist, a traceable target component identity basis is first established before entering the side wall defect compensation detection process, so that subsequent detection results have a clear object source and access boundary.

[0028] S2, obtain the target component identity record, and based on the side wall visibility projection processing, perform weld end side wall visibility recognition on the top view pattern in the multi-view pattern reading record, generate the side wall uncovered area, and keep the side wall uncovered area as the area to be compensated for detection.

[0029] S2 is used to obtain the target component identity record formed by S1, and based on the side wall visibility projection processing, to perform weld end side wall visibility recognition on the top view pattern in the multi-view pattern reading record, generate the side wall uncovered area, and keep the side wall uncovered area as the area to be compensated for detection. The target component identity record serves as the object boundary, the top view pattern serves as the evidence of weld end side wall visibility, the target component posture serves as the side wall projection direction, the occlusion boundary serves as the visibility restriction, and the side wall uncovered area serves as the spatial object for subsequent cross-station identity closure processing.

[0030] The system uses the target component identity record as an index to retrieve the top view pattern from the multi-view pattern reading record. The target component identity record at least records the target bearing position identifier, the target component posture, the top view pattern reading time, and the identity status. The retrieved top view pattern and the target component identity record belong to the same top view pattern reading time and the same target bearing position. Through this index relationship, the top view pattern, the target component posture, and the uncovered area of ​​the sidewall in S2 are all limited to the same top view pattern coordinates, restricting patterns of different bearing positions and patterns of different reading times from participating in the same weld end sidewall identification process. If the top view pattern cannot be retrieved according to the target component identity record, the system writes the target bearing position into a restricted compensation state, stops the ordinary sidewall visibility projection processing, and does not allow the residual backfill permission processing. If the retrieved top view pattern reading time is inconsistent with the target component identity record, the system generates a restricted compensation detection area, stops the ordinary sidewall visibility projection processing, and does not allow the residual backfill permission processing.

[0031] The sidewall visibility determination criteria used in this step are formed by stable reading samples under the same model of carrier, the same specification of electronic components, and the same top-view imaging cycle, combined with camera calibration records, carrier bearing position design boundaries, and first-piece inspection verification records. Camera calibration records are used to limit the available range of top view coordinates and sidewall projection positions; carrier bearing position design boundaries are used to limit the extraction range of occlusion boundaries; stable reading samples are used to form a continuous reading state of normal weld end boundaries; first-piece inspection verification records are used for the initial visibility calibration of the current batch. When there are insufficient stable reading samples in the current batch, the system uses camera calibration records, carrier bearing position design boundaries, and first-piece inspection verification records to form initial sidewall visibility determination criteria. The initial sidewall visibility determination criteria are only used for S2 identification of the current batch, are not written into the stable reading sample records, and do not directly form defect results. When the initial sidewall visibility determination criteria cannot be formed, the system generates a restricted detection area to be compensated and stops generating ordinary sidewall uncovered areas.

[0032] When the target component's identity record is valid, the top view retrieval result meets the reading time consistency condition, the sidewall visibility determination condition is valid, and the visible solder end boundary can be read, the system performs sidewall projection on the visible solder end boundary in the top view coordinate according to the target component's posture, forming the sidewall projection position. The sidewall projection position is used to characterize the identifiable projection range of the target component's solder end sidewall in the top view. Its source is the target component's posture and the visible solder end boundary in the top view, and it is constrained by the available range of the top view coordinate limited by the camera calibration record. When the micro-surface capacitor is in a slightly tilted state within the carrier position, the system does not directly divide the solder end sidewall with the center direction of the carrier position, but limits the sidewall projection direction of the visible solder end boundary with the target component's posture, and includes the solder end sidewall near the inner wall of the carrier position into the sidewall visibility identification object. When at least one of the following is true: the target component's identity record is in a restricted state, or the visible solder end boundary cannot be read, the system generates a restricted detection area to be compensated, stops the ordinary sidewall projection process, and must not enter the residual backfill permission process.

[0033] After forming the sidewall projection position, the system extracts the pattern boundary that affects the continuous reading of the weld end boundary in the top view within the range related to the target bearing position and the sidewall projection position, forming an occlusion boundary. The sidewall projection position is then superimposed with the occlusion boundary. The occlusion boundary includes the inner wall of the target bearing position, the edge of the carrier, the afterimage of the cover strip, and the boundary formed by the reflective covering pattern that causes the weld end boundary to be unreadable in the top view. The occlusion boundary is only used as a limiting condition for the visibility identification of the weld end sidewall and is not used as a defect result. When the occlusion boundary conflicts with the main outline of the target component, the system generates a restricted detection area to be compensated and stops the generation of ordinary sidewall uncovered areas.

[0034] The system determines the superposition result of the side wall projection position and the occlusion boundary. When the side wall projection position does not overlap with the occlusion boundary and forms a continuous weld end boundary reading result, the system confirms the side wall projection position as the weld end side wall range covered by the top view. The continuous weld end boundary reading result means that in the side wall direction defined by the target component's posture, the top view forms interconnected weld end boundary reading segments, and the reading segments meet the current batch's side wall visibility determination conditions, so that the range can be observed by the top view.

[0035] When at least one of the following conditions is met: the sidewall projection position overlaps with the occlusion boundary, or the continuous weld end boundary reading result is not formed, the system generates a sidewall uncovered area. The sidewall uncovered area indicates that the corresponding weld end sidewall range of the target component cannot be effectively observed from the top view. The reasons for its formation include occlusion of the inner wall of the target bearing position, occlusion of the carrier edge, occlusion of the cover strip shadow, and occlusion of the reflective surface, which makes the weld end boundary unreadable. The system does not interpret the sidewall uncovered area as a qualified state of the electronic component, nor does it directly interpret the pattern abnormality within the occlusion boundary as a defective state of the electronic component.

[0036] The system marks the sidewall orientation of the uncovered area according to the target component's posture. The sidewall orientation is the positional relationship of the uncovered area relative to the target component's posture, indicating that the uncovered area belongs to the corresponding weld end sidewall of the target component. The sidewall orientation does not use the global orientation of the carrier to replace the orientation of the target component. When the component has a slight deflection, S3 performs sidewall residual same-object closure according to the target component's posture. If the sidewall orientation can be formed, the system keeps the uncovered area as the area to be compensated for detection. If the sidewall orientation cannot be formed according to the target component's posture, the system generates a restricted area to be compensated for detection and does not treat the corresponding area as a normal area to be compensated for detection.

[0037] This step ultimately outputs the state of the uncovered sidewall area and the area to be compensated for detection. The uncovered sidewall area is associated with the target component identity record, the top view reading time, the sidewall orientation, and the state to be compensated. As the spatial object for S3 to perform cross-station identity closure processing, S3 only performs residual backfill permission judgment on the supplementary view sidewall residual that can be closed to the uncovered sidewall area.

[0038] S3, obtain the target component identity record and the uncovered area of ​​the side wall, and perform side wall residual identification and verification on the supplementary view pattern in the multi-view pattern reading record based on cross-station identity closure processing. When the side wall residual closes with the target component identity record, generate the residual backfill permission result.

[0039] S3 takes over the target component identity record formed by S1 and the sidewall uncovered area formed by S2. It is used to determine the backfill permission of the candidate sidewall residual in the supplementary view pattern and the sidewall uncovered area of ​​the same target component. The target component identity record serves as the identity benchmark for cross-station objects, the sidewall uncovered area serves as the object to be compensated for in the detection space, the supplementary view pattern serves as supplementary identification evidence for the sidewall residual, and the residual backfill permission result serves as the admission function for the residual attribution competition processing. S3 does not directly generate electronic component defect results, nor does it directly generate electronic component qualified results.

[0040] The system acquires the target component identity record and the uncovered area of ​​the sidewall. The target component identity record records at least the target bearing position identifier, the target component attitude, the top view reading time, and the identity status. The uncovered area of ​​the sidewall records at least the sidewall orientation, the state to be compensated, and the corresponding target component identity record. When the target component identity record has a valid identity status and the uncovered area of ​​the sidewall remains a detection area to be compensated, the system enters the sidewall residual identification and verification process. When at least one of the following is true: the target component identity record has a restricted identity status, the uncovered area of ​​the sidewall is a restricted detection area to be compensated, or the uncovered area of ​​the sidewall has not formed a sidewall orientation, the system generates a residual backfilling restricted result and does not perform residual backfilling permission processing.

[0041] The workstation connection relationship is used to supplement the admission criteria for viewpoint pattern retrieval and to confirm cross-workstation connection chains; the sidewall recognition range is used to limit the recognition location of candidate sidewall residuals; the residual readable range is used to limit the area in the supplementary viewpoint pattern where sidewall residuals can be read; the closure check condition is used to control whether candidate sidewall residuals generate residual backfill permission results. Time connection, spatial connection, and sample stability state are used for pattern retrieval admission, sidewall recognition range limitation, and closure check condition formation, respectively, and are not directly added together.

[0042] Before forming the sidewall identification range, the system performs an access check on the supplementary viewpoint pattern. The system limits the retrieval objects of the supplementary viewpoint pattern by the target element identity record, and retrieves the supplementary viewpoint pattern that matches the target bearing position identifier from the multi-view pattern reading record in combination with the supplementary viewpoint closing identifier. The supplementary viewpoint closing identifier comes from the cross-station call relationship established by S1 around the target bearing position, which is used to limit the station connection chain between the top view pattern and the supplementary viewpoint pattern. The retrieval object of the supplementary viewpoint pattern is jointly limited by the target bearing position identifier and the supplementary viewpoint closing identifier in the target element identity record, so that the supplementary viewpoint pattern corresponds to the same target bearing position and the same cross-station connection chain. The supplementary viewpoint pattern is directly selected without following the pattern acquisition order. If the supplementary viewpoint pattern does not match the target bearing position identifier in the target element identity record or the supplementary viewpoint closing identifier is missing, the system generates a residual backfilling limited result and does not form a sidewall identification range.

[0043] After the supplementary view pattern passes the access check, the system forms a sidewall recognition range in the supplementary view pattern based on the sidewall orientation of the uncovered area. The top view pattern and the supplementary view pattern are not directly superimposed using pixel coordinates. The system uses the target bearing position object in the supplementary view pattern as the local object reference, locates the sidewall area corresponding to the uncovered area based on the workstation connection relationship and the sidewall orientation, and takes the overlapping part of the sidewall area and the residual readable range as the sidewall recognition range. The sidewall recognition range is limited to the sidewall area corresponding to the target bearing position object in the supplementary view pattern and does not extend to the entire image of the supplementary view pattern. If at least one of the following is true: the workstation connection relationship cannot be formed, the residual readable range cannot be formed, or the sidewall recognition range is not formed, the system generates a residual backfilling limited result and does not perform candidate sidewall residual recognition.

[0044] After the sidewall recognition range is formed, the system identifies candidate sidewall residuals within the sidewall recognition range. Candidate sidewall residuals refer to local pattern anomalies in the supplementary view pattern that are located within the sidewall recognition range and participate in the closure check. Candidate sidewall residuals are only used as objects to be closed and are not used as evidence of sidewall defects. When the vehicle pattern residuals and adjacent bearing position residuals in the supplementary view pattern do not fall within the sidewall recognition range, they are not used as candidate sidewall residuals. When candidate sidewall residuals are not formed, the system generates a residual backfilling limitation result and does not perform a closure check.

[0045] Candidate sidewall residuals can be formed by grayscale abrupt changes, edge breaks, local texture discontinuities, or weld end boundary defects relative to the normal sidewall pattern of the current batch within the sidewall recognition range. The system only uses local anomalies located within the sidewall recognition range that can record the pattern position, reading time, and sidewall orientation as candidate sidewall residuals.

[0046] After the candidate sidewall residuals are formed, the system performs closure checks on the candidate sidewall residuals corresponding to the target bearing position identifier, the target element attitude, and the sidewall orientation. The closure check corresponding to the target bearing position identifier is used to determine whether the candidate sidewall residual belongs to the target bearing position defined by the target element identity record; the closure check corresponding to the target element attitude is used to determine whether the sidewall position where the candidate sidewall residual is located accepts the target element attitude; and the closure check corresponding to the sidewall orientation is used to determine whether the candidate sidewall residual corresponds to the same sidewall orientation marked by S2. When all three closure checks are valid, the system determines that the candidate sidewall residual is closed with the target element identity record and generates a residual backfill permission result. When not all closure checks are valid, the system generates a residual backfill restriction result and excludes the corresponding candidate sidewall residual from the residual attribution competition processing.

[0047] The closure check conditions are uniformly formed by the station calibration record, camera calibration record, stable cross-station reading sample, and first piece inspection verification record. The station calibration record is used to limit the station connection relationship between the top view pattern and the supplementary view pattern; the camera calibration record is used to limit the readable range of the residual in the supplementary view pattern; the stable cross-station reading sample is used to form the normal connection state of the same target bearing position between the top view pattern and the supplementary view pattern; the first piece inspection verification record is used for the initial closure calibration of the current batch. The stable cross-station reading sample is not directly generated from the residual backfill permission result and the residual backfill limitation result.

[0048] When there are insufficient stable cross-station read samples in the current batch, the system uses station calibration records, camera calibration records, and first-piece inspection verification records to form initial closure conditions. The initial closure conditions are only used for S3 closure processing of the current batch, are not written to the stable cross-station read sample records, and do not directly generate defect results. After the current batch forms stable cross-station read samples, the system updates the closure check conditions based on the stable cross-station read samples and stops using the initial closure conditions on unprocessed objects. The already generated residual backfill permission results and residual backfill restricted results are not rewritten in reverse according to the updated closure check conditions. When the initial closure conditions cannot be formed, the system generates residual backfill restricted results and does not execute residual backfill permission processing.

[0049] The residual backfilling permission result shall at least record the target component identity record, the uncovered area of ​​the sidewall, the sidewall orientation, the candidate sidewall residual, and the permission status. The permission status is used to indicate that the candidate sidewall residual is permitted to enter the residual attribution competition process, but does not indicate that the candidate sidewall residual has been identified as an electronic component defect, nor is it used as the basis for the electronic component's qualified test result.

[0050] The results of the restricted residual backfill should at least record the restricted object and the reason for the restriction. The reasons for the restriction include at least one of the following: the supplementary view pattern does not match the target bearing position mark, the supplementary view mark to be closed is missing, the work station connection relationship cannot be formed, the residual readable range cannot be formed, the side wall identification range is not formed, the candidate side wall residual is not formed, and the closure check is not fully established. The results of the restricted residual backfill should not be used as evidence of side wall defects, nor as the basis for the qualified test results of electronic components.

[0051] After processing in this step, the uncovered area of ​​the sidewall formed by S2 is converted into the sidewall identification range in the supplementary view pattern. The local pattern anomaly in the supplementary view pattern is only generated as a residual backfill permission result after the candidate sidewall residual is formed and the target bearing position identifier, target component attitude and sidewall orientation closure check is completed. The residual backfill permission result provides an access boundary for the residual attribution competition processing. Candidate sidewall residuals that have not completed cross-station identity closure cannot be backfilled to the uncovered area of ​​the target component's sidewall.

[0052] S4. Obtain the candidate residuals corresponding to the residual backfill permission result. Perform source identification by using the competition processing of vehicle residual and component residual attribution. Perform vehicle pattern repetition verification and component posture following verification on the candidate residuals. Exclude the candidate residuals belonging to the vehicle pattern from the component defect candidates. Generate the candidate residuals belonging to the target component and connected to the side wall uncovered area as side wall defect evidence. Output the electronic component visual defect detection result based on the side wall defect evidence.

[0053] S4 follows the residual backfill permission result generated by S3. The residual backfill permission result only indicates that the candidate residual has completed cross-station identity closure and has the admission conditions to enter the competition processing of vehicle residual and component residual attribution. This result does not indicate that the candidate residual has been assigned to the target component, nor does it indicate that the candidate residual has constituted evidence of sidewall defect. S4 is used to perform source identification on the candidate residuals that have obtained the residual backfill permission result, so that the vehicle pattern residual, adjacent bearing position residual and target component sidewall residual in the supplementary view pattern are respectively subjected to exclusion processing, evidence generation processing and restricted output processing.

[0054] The system obtains the candidate residuals corresponding to the residual backfill permission result. The candidate residuals are the candidate sidewall residuals recorded in the residual backfill permission result of S3, which are collectively referred to as candidate residuals in S4. The system simultaneously retrieves the target component identity record, sidewall uncovered area, sidewall orientation, supplementary viewpoint pattern, and vehicle identification record associated with the candidate residuals. The target component identity record is used to limit the target bearing position identifier and target component attitude corresponding to the candidate residuals; the sidewall uncovered area is used to limit whether the candidate residuals are connected to the compensation detection area formed in S2; the sidewall orientation is used to limit the direction of the weld end sidewall corresponding to the candidate residuals; the supplementary viewpoint pattern is used to provide pattern evidence of the candidate residuals; and the vehicle identification record is used to limit the vehicle pattern position required for repeated vehicle pattern verification.

[0055] This step is used to determine the criteria for vehicle pattern repetition verification, component attitude tracking verification, and acceptance relationship confirmation. These criteria are uniformly formed from vehicle identification records, camera calibration records, stable reading samples, and first-piece inspection verification records. The reading time condition is only used to limit the order of available records. The pattern positional relationship is compared under the local positional relationship of the target bearing position defined by the vehicle identification record, and is not numerically fused with the reading time condition. When there are insufficient stable reading samples in the current batch, the system uses the vehicle identification record, camera calibration record, and first-piece inspection verification record to form the initial criteria. The initial criteria are only used for S4 source identification in the current batch and are not written to the stable reading sample record. When the initial criteria cannot be formed, the system generates a residual attribution restriction result and stops executing vehicle pattern repetition verification, component attitude tracking verification, and acceptance relationship confirmation.

[0056] The system performs source identification access checks on the S4 processing object. When the residual backfill permission result does not have a permission status or when the residual backfill permission result does not record at least one of the candidate residuals, the system generates a residual attribution restricted result and stops performing source identification on the residual backfill permission result. When a candidate residual has been formed, but at least one of the following conditions is met: the candidate residual is not associated with the target component identity record, the vehicle identification record cannot limit the vehicle pattern position, or the target component attitude and sidewall orientation cannot limit the corresponding sidewall position, the system generates a residual attribution restricted result and does not generate the candidate residual as evidence of sidewall defect.

[0057] After the source identification access check is passed, the system performs vehicle pattern duplication verification on the candidate residuals based on the vehicle identification record. This verification checks whether the candidate residual repeats the vehicle pattern position defined by the vehicle identification record. The system uses the target bearing position boundary, vehicle edge, and bearing position inner wall pattern defined by the vehicle identification record as the vehicle pattern position boundary. It retrieves valid vehicle pattern reading records from the multi-view pattern reading records whose reading time is no later than the reading time of the supplementary view pattern of the candidate residual and which can form the vehicle pattern position. Reading records that cannot form the vehicle pattern position do not participate in the vehicle pattern duplication verification. If a valid vehicle pattern reading record cannot be retrieved, the system... The system generates a limited result for residual assignment, but does not generate the candidate residual as evidence of sidewall defects. Based on the valid reading record of the vehicle pattern, the system verifies whether the candidate residual appears repeatedly relative to the same vehicle pattern position under the local positional relationship of the target bearing position defined by the vehicle identification record. When the candidate residual appears repeatedly relative to the vehicle pattern position defined by the vehicle identification record, the system determines that the vehicle pattern repetition verification is valid. When no repetition relationship is formed after the valid reading record of the vehicle pattern, the system determines that the vehicle pattern repetition verification is invalid. Whether the vehicle pattern repetition verification is valid or invalid is only used as input for the competition processing of vehicle residual and component residual assignment, and does not directly determine the final assignment of the candidate residual.

[0058] The system performs component attitude following verification on candidate residuals based on the target component's attitude and sidewall orientation. Component attitude following verification is used to verify whether the candidate residual is located at the corresponding sidewall position jointly defined by the target component's attitude and sidewall orientation. The system defines the target component's attitude direction in the supplementary viewpoint pattern by the target component's attitude and defines the weld end sidewall direction corresponding to the sidewall not covered in S2 by the sidewall orientation. It verifies whether the candidate residual is located at the corresponding sidewall position jointly defined by the attitude direction and sidewall orientation. When the candidate residual is located at the corresponding sidewall position, the system determines that the component attitude following verification is successful; when the candidate residual is not located at the corresponding sidewall position, the system determines that the component attitude following verification is unsuccessful. Whether the component attitude following verification is successful or unsuccessful is only used as input for the competition processing of the attribution of the carrier residual and the component residual, and is not directly equivalent to evidence of sidewall defects.

[0059] After the component attitude tracking verification is successful, the system performs a connection relationship confirmation on the candidate residual and the sidewall uncovered area. The connection relationship is jointly defined by the same target component identity record, the same sidewall orientation, and the corresponding sidewall position. When the target component identity record corresponding to the candidate residual is consistent with the target component identity record corresponding to the sidewall uncovered area, the sidewall position corresponding to the candidate residual is consistent with the sidewall orientation of the sidewall uncovered area, and the supplementary view pattern evidence corresponding to the candidate residual comes from the residual backfill permission result, the system determines that the candidate residual and the sidewall uncovered area are connected. When the connection relationship is not established, the candidate residual does not enter the sidewall defect evidence generation link.

[0060] The system uses a competitive process of assigning vehicle residuals and component residuals to divide the results of candidate residuals. When the vehicle pattern duplication verification is successful and the component posture following verification is unsuccessful, the system assigns the candidate residual to the vehicle pattern and excludes the candidate residual from the component defect candidates. The excluded candidate residuals do not generate sidewall defect evidence and are not used as the basis for the target component defect in the electronic component visual defect detection results.

[0061] When the vehicle pattern repeat verification fails, the component attitude follow verification succeeds, and the candidate residual is connected to the uncovered area of ​​the sidewall, the system will assign the candidate residual to the target component, generate the candidate residual as evidence of sidewall defect, and output the visual defect detection result of the electronic component based on the evidence of sidewall defect. The evidence of sidewall defect comes from the candidate residual corresponding to the residual backfill permission result. Its corresponding object is the target component defined by the target component identity record. Its corresponding position is the sidewall position defined by the target component attitude and the sidewall orientation. Its connected object is the uncovered area of ​​the sidewall formed by S2. The evidence of sidewall defect does not come from the sidewall area that is not effectively observed in the top view, nor is it written into the stable reading sample record, and it does not participate in updating the closure check condition in S3.

[0062] When at least one of the following conditions is met: vehicle pattern repetition verification and component attitude following verification are both true; vehicle pattern repetition verification and component attitude following verification are both false; vehicle pattern repetition verification is false and component attitude following verification is true but the candidate residual does not connect with the sidewall uncovered area, the system generates a residual attribution restriction result. The corresponding candidate residual is not generated as evidence of sidewall defect. The residual attribution restriction result is used to characterize that the source of the candidate residual cannot form a unique attribution between the vehicle pattern and the target component. It is not used as a basis for target component defect or as a basis for target component qualification. The corresponding candidate residual does not enter the component defect candidate confirmation link, does not enter the sidewall defect evidence generation link, does not enter the stable reading sample record, and does not participate in subsequent benchmark updates.

[0063] The system outputs the visual defect detection results of electronic components based on the sidewall defect evidence. When the candidate residual generates sidewall defect evidence, the visual defect detection results of electronic components include the target bearing position identifier, the target component posture, the sidewall orientation, and the sidewall defect evidence. When the candidate residual is assigned to the vehicle pattern, the system does not output the target component sidewall defect result based on the candidate residual. When the candidate residual generates a residual assignment restriction result, the system outputs the detection restriction result and retains the target component identity record, the sidewall uncovered area, and the restriction reason for subsequent review processing.

[0064] In this step, the candidate residuals corresponding to the residual backfill permission result are not directly generated as sidewall defect evidence. Instead, vehicle pattern repetition verification and component attitude following verification are performed before generating sidewall defect evidence. Only candidate residuals that fail vehicle pattern repetition verification, succeed component attitude following verification, and connect with the uncovered area of ​​the sidewall are generated as sidewall defect evidence. Candidate residuals that succeed vehicle pattern repetition verification are excluded from component defect candidates. Candidate residuals with conflicting attribution states, no attribution states, or no connection relationship are included in the residual attribution restricted result. Thus, the visual defect detection results of electronic components can maintain the consistency between the target component object, sidewall position, and residual source.

[0065] Example 2: Please see Figure 2 Based on Embodiment 1, this embodiment also provides an electronic component visual defect detection system, including: an identity record forming module, which acquires multi-view pattern reading records of the carrier carrying the electronic component during continuous transport, and uses carrier pattern and component appearance pattern separation and recognition processing to form a target component identity record. The target component identity record is used to maintain the correspondence between the target carrier position identifier and the target component posture. The module for forming the area to be compensated acquires the identity record of the target component, performs weld end sidewall visibility recognition on the top view pattern in the multi-view pattern reading record based on sidewall visibility projection processing, generates the sidewall uncovered area, and keeps the sidewall uncovered area as the area to be compensated and detected. The residual backfilling permission module obtains the target component identity record and the uncovered area of ​​the side wall. Based on cross-station identity closure processing, it performs side wall residual identification and verification on the supplementary view pattern in the multi-view pattern reading record. When the side wall residual closes with the target component identity record, it generates a residual backfilling permission result. The defect result output module obtains the candidate residuals corresponding to the residual backfill permission result, performs source identification by handling the competition between vehicle residuals and component residuals, performs vehicle pattern repetition verification and component posture following verification on the candidate residuals, excludes the candidate residuals belonging to the vehicle pattern from the component defect candidates, generates the candidate residuals belonging to the target component and connected to the side wall uncovered area as side wall defect evidence, and outputs the electronic component visual defect detection result based on the side wall defect evidence.

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

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

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

Claims

1. A method for visual defect detection in electronic components, characterized in that, The specific steps include the following: The system acquires multi-view pattern reading records of the carrier carrying electronic components during continuous transport. It uses carrier pattern and component appearance pattern separation and identification processing to form a target component identity record. The target component identity record is used to maintain the correspondence between the target carrier position identifier and the target component posture. Obtain the target component identity record, and based on the sidewall visibility projection processing, perform weld end sidewall visibility recognition on the top view pattern in the multi-view pattern reading record, generate the sidewall uncovered area, and keep the sidewall uncovered area as the area to be compensated for detection. Obtain the target component identity record and the uncovered area of ​​the side wall. Based on the cross-station identity closure processing, perform side wall residual identification and verification on the supplementary view pattern in the multi-view pattern reading record. When the side wall residual closes with the target component identity record, generate the residual backfill permission result. Candidate residuals corresponding to the residual backfill permission results are obtained. Source identification is performed by handling the competition between vehicle residuals and component residuals. Vehicle pattern repetition verification and component posture following verification are performed on the candidate residuals. Candidate residuals belonging to the vehicle pattern are excluded from the component defect candidates. Candidate residuals belonging to the target component and connected to the sidewall uncovered area are generated as sidewall defect evidence. The electronic component visual defect detection results are output based on the sidewall defect evidence.

2. The method for visual defect detection of electronic components according to claim 1, characterized in that, The process of creating a target component identity record includes: identifying a vehicle pattern from a multi-view pattern reading record and generating a vehicle identification record; identifying a component appearance pattern within the target bearing position defined by the vehicle identification record and generating a component identification record; wherein, the supplementary view pattern is only used for cross-workstation identity closure processing and does not participate in the validity confirmation of the target component identity record.

3. The method for visual defect detection of electronic components according to claim 2, characterized in that, The formation of a target component identity record also includes: performing identity acceptance matching processing on the vehicle identification record and the component identification record to generate acceptance matching results between the candidate component corresponding to the component identification record and the target bearing position; when the acceptance matching result meets the acceptance permission requirements formed by the acceptance relationship of the target bearing position in the stable reading sample, a target component identity record with a valid identity status is generated; when the acceptance matching result does not meet the acceptance permission requirements, a target component identity record with a restricted identity status is generated, and it is prohibited to output qualified test results based on the target component identity record with a restricted identity status.

4. The method for visual defect detection of electronic components according to claim 1, characterized in that, The weld end sidewall visibility identification includes: retrieving the top view chart using the target component identity record as an index; when the target component identity record has an identity valid state, performing sidewall projection on the visible weld end boundary according to the target component posture under the top view chart coordinates to form the sidewall projection position; when the target component identity record has an identity restricted state, generating a restricted detection area to be compensated, and not allowing it to enter the residual backfill permission processing.

5. The method for visual defect detection of electronic components according to claim 4, characterized in that, Generating an uncovered area on the sidewall includes: overlaying the sidewall projection position with the occlusion boundary in the top view; generating an uncovered area on the sidewall when at least one of the following conditions is met: the sidewall projection position overlaps with the occlusion boundary, or the continuous weld end boundary reading result is not formed; marking the sidewall orientation of the uncovered area on the sidewall according to the target component orientation; and keeping it as a detection area to be compensated.

6. The method for visual defect detection of electronic components according to claim 1, characterized in that, The sidewall residual identification and verification includes: defining the retrieval object of the supplementary view pattern based on the target element identity record; forming a sidewall identification range in the supplementary view pattern based on the sidewall orientation of the sidewall-uncovered area; identifying candidate sidewall residuals within the sidewall identification range; and generating a residual backfilling restricted result when at least one of the following conditions is met: the supplementary view pattern does not match the target bearing position identifier in the target element identity record or the sidewall identification range is not formed.

7. The method for visual defect detection of electronic components according to claim 6, characterized in that, The process of generating residual backfill permission results includes: performing a closure check on the candidate sidewall residuals corresponding to the target bearing position identifier, target component attitude, and sidewall orientation; generating residual backfill permission results when all closure checks are successful; generating residual backfill restricted results when no candidate sidewall residuals are formed; and generating residual backfill restricted results when not all closure checks are successful, and excluding the corresponding candidate sidewall residuals from the residual attribution competition process.

8. The method for visual defect detection of electronic components according to claim 1, characterized in that, The source identification process includes: performing vehicle pattern duplication verification on candidate residuals based on vehicle identification records, which verifies whether the candidate residuals repeat the vehicle pattern position defined by the vehicle identification records; performing component attitude following verification on candidate residuals based on the target component attitude and sidewall orientation, which verifies whether the candidate residuals are located at the corresponding sidewall position defined by the target component attitude and sidewall orientation; and generating a residual attribution restriction result when at least one of the following conditions is met: the vehicle identification records cannot define the vehicle pattern position, or the target component attitude and sidewall orientation cannot define the corresponding sidewall position. The corresponding candidate residuals are not generated as evidence of sidewall defects.

9. The method for visual defect detection of electronic components according to claim 8, characterized in that, The output of the electronic component visual defect detection results includes: when the vehicle pattern repetition verification is successful but the component posture following verification is unsuccessful, the candidate residual is excluded from the component defect candidates; when the vehicle pattern repetition verification is unsuccessful, the component posture following verification is successful, and the candidate residual is connected to the side wall uncovered area, the candidate residual is generated as side wall defect evidence, and the electronic component visual defect detection results are output based on the side wall defect evidence; when at least one of the following is true: the vehicle pattern repetition verification and the component posture following verification are both successful, the vehicle pattern repetition verification and the component posture following verification are both unsuccessful, or the vehicle pattern repetition verification is unsuccessful and the component posture following verification is successful but the candidate residual is not connected to the side wall uncovered area, a residual attribution restriction result is generated, and the corresponding candidate residual is not generated as side wall defect evidence.

10. A visual defect detection system for electronic components, employing the visual defect detection method for electronic components as described in any one of claims 1-9, characterized in that, include: The identity record forming module acquires multi-view pattern reading records of the carrier carrying electronic components during continuous transport. It uses carrier pattern and component appearance pattern separation and recognition processing to form target component identity record. The target component identity record is used to maintain the correspondence between the target carrier position identifier and the target component posture. The module for forming the area to be compensated acquires the identity record of the target component, performs weld end sidewall visibility recognition on the top view pattern in the multi-view pattern reading record based on sidewall visibility projection processing, generates the sidewall uncovered area, and keeps the sidewall uncovered area as the area to be compensated and detected. The residual backfilling permission module obtains the target component identity record and the uncovered area of ​​the side wall. Based on cross-station identity closure processing, it performs side wall residual identification and verification on the supplementary view pattern in the multi-view pattern reading record. When the side wall residual closes with the target component identity record, it generates a residual backfilling permission result. The defect result output module obtains the candidate residuals corresponding to the residual backfill permission result, performs source identification by handling the competition between vehicle residuals and component residuals, performs vehicle pattern repetition verification and component posture following verification on the candidate residuals, excludes the candidate residuals belonging to the vehicle pattern from the component defect candidates, generates the candidate residuals belonging to the target component and connected to the side wall uncovered area as side wall defect evidence, and outputs the electronic component visual defect detection result based on the side wall defect evidence.

Images