An artificial and machine collaborative AOI circuit board quality inspection method and system

Abstract

The application provides an artificial and machine cooperative AOI circuit board quality inspection method and system, and belongs to the technical field of measurement. The method comprises the following steps: step 1: based on an AI quality inspection robot, a circuit board on a production line is scanned to obtain a global scanning result; based on the global scanning result, a target parameter sequence with the minimum camera adjustment complexity is determined, and local high-precision collection is performed according to the target parameter sequence; step 2: based on the scanning result, intelligent analysis and output are performed to determine a circuit board to be re-inspected; step 3: artificial marking information of the circuit board to be re-inspected is obtained; and step 4: the artificial marking information is fed back to an AI quality inspection model of the AI quality inspection robot to update the model. The artificial and machine cooperative AOI circuit board quality inspection method and system are artificial and machine cooperative, the quality inspection efficiency is higher, the camera parameters are gradually adjusted with the minimum adjustment complexity, and the local high-precision collection process control is more suitable.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of measurement, in particular to an AOI circuit board quality inspection method and system with human-machine cooperation. BACKGROUND

[0002] AOI circuit board quality inspection is affected by many factors such as light conditions, solder reflection and component color changes, which can easily misjudge normal solder joints as false solder joints, bridging and other defective solder joints, resulting in low false alarm rate of automatic quality inspection, and requiring a large amount of manual intervention for re-inspection. However, a large amount of manual intervention for re-inspection cannot take advantage of the original advantages of automatic quality inspection, and the overall quality inspection efficiency and cost compared with traditional manual quality inspection have not been significantly improved, and may even be lower in quality inspection efficiency and higher in cost.

[0003] Therefore, there is an urgent need for an AOI circuit board quality inspection method and system with human-machine cooperation to at least solve the above problems. SUMMARY

[0004] One of the purposes of the present application is to provide an AOI circuit board quality inspection method and system with human-machine cooperation, which first uses an AI quality inspection robot to detect all circuit boards. During detection, a global scanning result is obtained, and based on the global scanning result, a target parameter sequence with the minimum camera adjustment complexity is determined, and local high-precision collection is performed according to the target parameter sequence. The AI quality inspection model of the AI quality inspection robot is updated in time after obtaining the manual annotation information of the circuit board to be re-inspected, and the human-machine cooperation is more efficient, the camera parameters are adjusted gradually with the minimum adjustment complexity, and the local high-precision collection process is more suitable.

[0005] The AOI circuit board quality inspection method with human-machine cooperation provided by the embodiment of the present application comprises:

[0006] Step 1: AOI automatic scanning of circuit boards on a production line based on an AI quality inspection robot;

[0007] Step 2: intelligent analysis and output based on the scanning result to determine the circuit board to be re-inspected;

[0008] Step 3: obtaining manual annotation information of the circuit board to be re-inspected;

[0009] Step 4: feeding back the manual annotation information to the AI quality inspection model of the AI quality inspection robot for model updating;

[0010] The step 1: AOI automatic scanning of circuit boards on a production line based on an AI quality inspection robot comprises:

[0011] fixing the circuit board to a detection position;

[0012] After the fixing is completed, the global scanning is performed on the circuit board to be detected;

[0013] Based on the global scanning result, the local high-precision acquisition is performed;

[0014] Based on the global scanning result, the local high-precision acquisition is performed, including:

[0015] Based on the global scanning result, the suspected defect area is output;

[0016] According to the suspected defect type of the suspected defect area, the suspected defect area is grouped to obtain a region group;

[0017] Each region group is traversed in turn, and the camera configuration sub-parameter is obtained according to the suspected defect type and the suspected defect area distribution of the region group being traversed;

[0018] When each region group is traversed, the camera configuration sub-parameter group of different parameter sequences is obtained;

[0019] The camera configuration sub-parameter group with the minimum camera adjustment complexity is determined as the target parameter sequence, and the local high-precision acquisition is performed according to the target parameter sequence.

[0020] Preferably, step 3: obtaining the manual annotation information of the circuit board to be re-inspected, including:

[0021] The mechanical arm is used to transfer the circuit board to be re-inspected to the manual station;

[0022] The scanning result corresponding to the circuit board to be re-inspected is displayed to the target personnel, and the manual annotation information is obtained after the target personnel annotates.

[0023] The parameter sequence is used for local high-precision acquisition.

[0024] Preferably, the scanning result corresponding to the circuit board to be re-inspected is displayed to the target personnel, and the manual annotation information is obtained after the target personnel annotates, including:

[0025] The re-inspection semantic sequence of the target personnel when viewing the circuit board to be re-inspected is obtained; the re-inspection semantic sequence is obtained by sorting the re-inspection semantics generated by the target personnel when viewing the circuit board to be re-inspected according to the time sequence;

[0026] According to the re-inspection semantics and the fault type trigger semantics, the trigger semantics is determined;

[0027] According to the fault type corresponding to the preset verification basis source, the verification basis is obtained following the trigger time of the trigger semantics;

[0028] According to the verification strategy corresponding to the verification basis, the verification result is determined;

[0029] If the verification result is a verification pass, the verification basis is parsed to obtain the visual positioning basis;

[0030] If the visual positioning basis is successfully obtained, the display visual field of the scanning result is determined according to the visual positioning basis;

[0031] Based on the display visual field, the scanning result corresponding to the circuit board to be rechecked is displayed to the target personnel.

[0032] The AOI circuit board quality inspection method provided by the embodiment of the application further comprises:

[0033] If the visual positioning basis is not successfully obtained, a rechecking semantic set in a time range preset for a distance triggering semantic is obtained;

[0034] According to the rechecking semantic set, a first rechecking entity description vector is constructed;

[0035] A second rechecking entity description vector of a plurality of local results of the scanning result is obtained;

[0036] According to a matching result of the first rechecking entity description vector and the second rechecking entity description vector, a target local result meeting a preset result threshold is determined;

[0037] If the target local result is unique, the display visual field is determined according to the target local result.

[0038] The AOI circuit board quality inspection method provided by the embodiment of the application further comprises:

[0039] If the target local result is not unique, the target local result is globally highlighted in the scanning result, and a display visual field selected by the target personnel is determined.

[0040] The AOI circuit board quality inspection system provided by the embodiment of the application comprises:

[0041] The machine scanning module is configured to perform AOI automatic scanning on the circuit board on the production line based on the AI quality inspection robot;

[0042] The intelligent analysis module is configured to perform intelligent analysis and output based on the scanning result, and determine the circuit board to be rechecked;

[0043] The manual annotation module is configured to obtain manual annotation information of the circuit board to be rechecked;

[0044] The model updating module is configured to feed back the manual annotation information to an AI quality inspection model of the AI quality inspection robot for model updating;

[0045] The machine scanning module performs AOI automatic scanning on the circuit board on the production line based on the AI quality inspection robot, and the machine scanning module comprises:

[0046] Fix the circuit board to the detection position;

[0047] After fixing, globally scan the circuit board to be detected;

[0048] Based on the global scanning result, perform local high-precision acquisition;

[0049] Among them, based on the global scanning result, performing local high-precision acquisition comprises:

[0050] Based on the global scanning result, output the suspected defect area;

[0051] According to the suspected defect type of the suspected defect area, the suspected defect area is grouped to obtain the area group;

[0052] Iterate each area group in turn, and according to the suspected defect type and suspected defect area distribution of the area group being iterated, obtain the camera configuration sub-parameter;

[0053] When each area group is iterated, obtain the camera configuration sub-parameter group of different parameter sequences;

[0054] Determine the camera configuration sub-parameter group with the smallest camera adjustment complexity as the target parameter sequence, and perform local high-precision acquisition according to the target parameter sequence.

[0055] Preferably, the artificial marking module obtains artificial marking information of the circuit board to be rechecked, comprising:

[0056] Use the mechanical arm to transfer the circuit board to be rechecked to the artificial work station;

[0057] Show the scanning result corresponding to the circuit board to be rechecked to the target personnel, and obtain the artificial marking information after the target personnel marks.

[0058] Preferably, the artificial marking module shows the scanning result corresponding to the circuit board to be rechecked to the target personnel, and obtains the artificial marking information after the target personnel marks, comprising:

[0059] Obtain the rechecking semantic sequence when the target personnel views the circuit board to be rechecked; the rechecking semantic sequence is obtained by sorting the rechecking semantics generated when the target personnel views the circuit board to be rechecked according to the generation time;

[0060] According to the rechecking semantics and the fault type trigger semantics, determine the trigger semantics;

[0061] According to the fault type corresponding to the preset verification basis source, obtain the verification basis following the trigger time of the trigger semantics;

[0062] According to the verification strategy corresponding to the verification basis, determine the verification result;

[0063] If the verification result is a verification pass, the verification basis is parsed to obtain the visual positioning basis;

[0064] If the visual positioning basis is successfully obtained, the display visual field of the scanning result is determined according to the visual positioning basis;

[0065] Based on the display visual field, the scanning result corresponding to the circuit board to be rechecked is displayed to the target personnel.

[0066] Preferably, the artificial marking module further performs the following operations:

[0067] If the visual positioning basis fails to be obtained, a rechecking semantic set within a time range preset for a distance triggering semantic is obtained;

[0068] According to the rechecking semantic set, a first rechecking entity description vector is constructed;

[0069] A second rechecking entity description vector of a plurality of local results of the scanning result is obtained;

[0070] According to the matching result of the first rechecking entity description vector and the second rechecking entity description vector, a target local result meeting a preset result threshold is determined;

[0071] If the target local result is unique, the display visual field is determined according to the target local result.

[0072] Preferably, the artificial marking module further performs the following operations:

[0073] If the target local result is not unique, the target local result is globally highlighted in the scanning result, and a display visual field selected by the target personnel is determined.

[0074] The beneficial effects of the present application are:

[0075] The present application first utilizes an AI quality inspection robot to detect all circuit boards, obtains a global scanning result during detection, determines a target parameter sequence with the minimum camera adjustment complexity based on the global scanning result, and performs local high-precision collection according to the target parameter sequence; a circuit board to be rechecked is obtained, and after obtaining artificial marking information of the circuit board to be rechecked, an AI quality inspection model of the AI quality inspection robot is timely updated, and the human and the machine cooperate, so that the quality inspection efficiency is higher, the camera parameters are gradually adjusted with the minimum adjustment complexity, and the local high-precision collection process control is more suitable.

[0076] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be realized and obtained by the structure particularly pointed out in the present application document.

[0077] The technical solutions of the present application are described in further detail below with reference to the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0078] The accompanying drawings are included to provide a further understanding of the present application, and constitute a part of the specification, illustrate the present application, and are used to explain the present application, and do not constitute a limitation on the present application. In the drawings:

[0079] Figure 1 A schematic diagram of an AOI circuit board quality inspection method of human-machine cooperation in an embodiment of the present application;

[0080] Figure 2 A schematic diagram of an AOI circuit board quality inspection system of human-machine cooperation in an embodiment of the present application. DETAILED DESCRIPTION

[0081] The preferred embodiments of the present application are described below with reference to the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and do not limit the present application.

[0082] The embodiment of the present application provides an AOI circuit board quality inspection method of human-machine cooperation, as shown in Figure 1 , comprising:

[0083] Step 1: based on the AI quality inspection robot, the circuit board on the production line is automatically scanned by AOI;

[0084] Step 2: based on the scanning result, intelligent analysis output is performed to determine the circuit board to be rechecked;

[0085] Step 3: obtaining the manual annotation information of the circuit board to be rechecked;

[0086] In step 3, the manual annotation information of the circuit board to be rechecked is obtained, comprising:

[0087] The circuit board to be rechecked is transferred to the manual station by the mechanical arm;

[0088] The scanning result corresponding to the circuit board to be rechecked is displayed to the target personnel, and the manual annotation information is obtained after the target personnel annotates;

[0089] Step 4: the manual annotation information is fed back to the AI quality inspection model of the AI quality inspection robot for model updating;

[0090] The step 1: based on the AI quality inspection robot, the circuit board on the production line is automatically scanned by AOI, comprising:

[0091] Fix the circuit board to the detection position;

[0092] After fixing, the circuit board to be detected is globally scanned;

[0093] based on the global scanning result, performing local high-precision acquisition;

[0094] The local high-precision acquisition based on the global scanning result comprises:

[0095] based on the global scanning result, outputting a suspected defect region;

[0096] grouping the suspected defect regions according to suspected defect types of the suspected defect regions to obtain a region group;

[0097] sequentially traversing each region group, and obtaining camera configuration sub-parameters according to a suspected defect type of the region group being traversed and a suspected defect region distribution;

[0098] After each region group is traversed, a camera configuration sub-parameter group with different parameter sequences is obtained.

[0099] determining a camera configuration sub-parameter group with the minimum camera adjustment complexity as a target parameter sequence, and performing local high-precision acquisition according to the target parameter sequence.

[0100] The working principle and beneficial effects of the above technical solution are:

[0101] When performing AOI automatic scanning on a circuit board, the circuit board needs to be mechanically positioned, and a conveyor belt or a mechanical arm is used to position the circuit board to a detection position. After confirming that the circuit board is in place, the circuit board is fixed. After fixing, a low-pixel (such as 8 million pixels) high-speed camera is used to shoot a low-resolution image of the entire circuit board to obtain a global scanning result. Then, a high-pixel (such as 20 million pixels) camera is used for local high-precision shooting to determine the key area (such as the solder joint) of the circuit board.

[0102] Next, intelligent analysis is performed on the scanning result. The intelligent analysis output means that the scanning result is characterized and compared with the CAD design features of the corresponding circuit board. According to the comparison result and a preset comparison result-determination result comparison library, the determination result of the circuit board is determined. For example, if the comparison result is missing parts, it is determined as a major defect; if the comparison result is a silk screen offset, it is determined as a minor defect; if the comparison result has no corresponding matching item in the comparison result-determination result comparison library, the corresponding circuit board is a circuit board to be rechecked. After determining the circuit board to be rechecked, the mechanical arm is used to transfer the circuit board to be rechecked to a manual work station. The scanning result corresponding to the circuit board to be rechecked is displayed to the target personnel. After the target personnel mark, the manual marking information is obtained. After obtaining the manual marking information, the AI quality inspection robot AI quality inspection model learns the manual marking information and applies it to subsequent quality inspection. The model is updated in a targeted manner.

[0103] When performing local high-precision acquisition based on the global scanning result, a defect detection algorithm (such as edge detection, texture analysis, etc.) is used to mark the area where a defect may exist and the suspected defect type (such as “missing part”, “element offset”, and “print offset”, etc.) in the suspected defect area. When determining the suspected defect area and the suspected defect type, the global scanning result is input into a suspected defect determination model (the suspected defect determination model is a defect determination model obtained by using a deep learning model to learn the AOI detection image of a scanning device whose scanning device parameter similarity is greater than a scanning parameter similarity threshold (such as 80%) with the global scanning result (the image is marked with a defect area and a defect type corresponding to the defect area)). The global scanning result is input into the defect determination model, and the defect area and defect type output by the model are used as the suspected defect area and the suspected defect type.

[0104] Since the local high-precision acquisition parameters (camera configuration parameters) for identifying different suspected defect types are different, the suspected defect areas are grouped according to the suspected defect types to obtain multiple area groups, such as the suspected defect types of areas A and E are missing parts, which are divided into the same area group; the suspected defect types of areas B, C, and D are print offsets, which are divided into the same area group.

[0105] Each area group is traversed in turn, and camera configuration sub-parameters are obtained according to the suspected defect type and suspected defect area distribution of the area group being traversed. The suspected defect area distribution is the local area distribution of the area in the area group being traversed in the entire AOI circuit board. When obtaining the camera configuration sub-parameters, a machine learning technique is used to learn the process record of manually configuring the camera setting parameters of a high-precision scanning camera according to the suspected defect type and suspected defect area distribution, and then the camera configuration sub-parameters (such as resolution, light source angle, exposure time, etc.) are obtained according to the suspected defect type and suspected defect area distribution of the area group being traversed.

[0106] After each area group is traversed, a plurality of area groups corresponding camera configuration sub-parameters are obtained. For local fine scanning of the same circuit board, a plurality of camera configuration sub-parameter changes are involved, a camera configuration sub-parameter group is a parameter setting sequence of the change strategy of the plurality of camera configuration sub-parameters, a camera configuration sub-parameter group with the minimum camera adjustment complexity is a target parameter sequence (a parameter setting sequence of the change of the camera configuration sub-parameter with the minimum camera adjustment complexity), when determining the camera adjustment complexity, first, according to the first camera parameter in the camera configuration sub-parameter group and a preset first camera adjustment complexity determination table, the adjustment complexity value of the camera from the initial state to the first camera parameter is determined, the first camera adjustment complexity determination table stores the camera from the initial state to different camera parameters and the corresponding adjustment complexity value in one-to-one correspondence; then, according to the parameter difference between the two adjacent camera configuration parameters in the camera configuration sub-parameter group and a preset second camera adjustment complexity determination table, the adjustment complexity value corresponding to the parameter difference is determined, the second camera adjustment complexity determination table includes the parameter difference and the adjustment complexity value in one-to-one correspondence; the adjustment complexity value corresponding to the camera configuration sub-parameter group is calculated by summation, the total camera adjustment complexity is obtained, and the target parameter sequence with the minimum camera adjustment complexity is screened, so that the camera parameter can be adjusted gradually with the minimum adjustment complexity, and the local high-precision acquisition process control is more suitable.

[0107] The AI quality inspection robot is first used to detect all circuit boards, the global scanning result is obtained during detection, the target parameter sequence with the minimum camera adjustment complexity is determined based on the global scanning result, and local high-precision acquisition is performed according to the target parameter sequence; the to-be-rechecked circuit board is obtained, the AI quality inspection model of the AI quality inspection robot is updated in time after the artificial annotation information of the to-be-rechecked circuit board is obtained, the human and the machine cooperate, the quality inspection efficiency is higher, the camera parameter is adjusted gradually with the minimum adjustment complexity, and the local high-precision acquisition process control is more suitable.

[0108] In one embodiment, the scanning result corresponding to the to-be-rechecked circuit board is displayed to the target personnel, the artificial annotation information is obtained after the target personnel labels, and the artificial annotation information includes:

[0109] The rechecking semantic sequence when the target personnel views the to-be-rechecked circuit board is obtained; the rechecking semantic sequence is obtained by sorting the rechecking semantics generated when the target personnel views the to-be-rechecked circuit board according to the time sequence;

[0110] The trigger semantics are determined according to the rechecking semantics and the fault type trigger semantics;

[0111] According to the preset verification basis source corresponding to the fault type, the verification basis is obtained at the trigger time following the trigger semantics;

[0112] According to the corresponding verification strategy of the verification basis, the verification result is determined;

[0113] If the verification result is a verification pass, the verification basis is parsed to obtain the visual positioning basis;

[0114] If the visual positioning basis is successfully obtained, the display visual field of the scan result is determined according to the visual positioning basis, and the scan result corresponding to the circuit board to be rechecked is displayed to the target person based on the display visual field;

[0115] If the visual positioning basis fails to be obtained, a rechecking semantic set within a time range preset for a distance trigger semantic is obtained;

[0116] According to the rechecking semantic set, a first rechecking entity description vector is constructed;

[0117] A second rechecking entity description vector of a plurality of local results of the scan result is obtained;

[0118] According to the matching result of the first rechecking entity description vector and the second rechecking entity description vector, a target local result that meets a preset result threshold is determined;

[0119] If the target local result is unique, the display visual field is determined according to the target local result;

[0120] If the target local result is not unique, the target local result is globally highlighted in the scan result, and a display visual field selected by the target person is determined.

[0121] The working principle and beneficial effects of the above technical solution are:

[0122] The rechecking semantic sequence is obtained by sorting the language semantics generated when the target person views the circuit board to be rechecked according to the chronological order, such as: [this block, board, somewhere, seems, virtual welding].

[0123] The fault type trigger semantic is a description semantic of the type of circuit board fault, such as: "missing parts", "component offset", and "silkscreen offset", etc., which is pre-set by the staff; the rechecking semantics and the fault type trigger semantics are semantically matched, if the semantic matching is successful (for example, the semantic matching similarity is greater than or equal to 90%), the corresponding rechecking semantic matched with the fault type trigger semantic is taken as the trigger semantic.

[0124] After the trigger semantics is generated, actual verification is performed manually, such as measuring a certain place with a multimeter; the verification basis source preset for the fault type is derived from the verification basis source of the fault corresponding to the fault type trigger semantics, and the verification basis is obtained according to the trigger time of the trigger semantics following the preset verification basis source corresponding to the fault type, such as: the fault type is: virtual welding, and the verification basis source is: the sensing source of the welding point connection strength, such as: a connected multimeter; the verification basis is: the sensing result of the welding point connection strength, such as: the conduction data of the welding point measured by the multimeter.

[0125] The verification result is that the verification is passed, which indicates that the corresponding fault type fault actually exists, such as: the welding point is not conductive, which indicates that the virtual welding actually exists at a certain place.

[0126] In order to facilitate the target personnel to mark the virtual welding position in the scanning result for subsequent model training, the field of view is positioned at the relevant marking position to assist manual rapid marking, therefore, the verification basis is analyzed to obtain the field of view positioning basis (such as: the actual measurement point of the connected multimeter on the circuit board).

[0127] The display field of view of the scanning result is determined according to the field of view positioning basis (such as: the actual measurement point corresponds to the display field of view of the scanning result after magnification.

[0128] In addition, there are cases where the field of view basis cannot be physically obtained from the verification basis, such as: the worker directly checks the fault by subjective viewing, at this time, the display field of view is determined according to the identified recheck semantics. When obtaining, first determine the recheck semantics set within the preset time range from the trigger semantics, the preset time range is, for example: within 3 seconds before and after the trigger semantics corresponding to the generation time. The first recheck entity description vector is: the relationship (such as: connection relationship) description vector of the circuit components mentioned in the recheck semantics, such as: the vector describing the component connection relationship of "w component and r are connected in series, and r and H are connected in parallel"; the second recheck entity description vector is the relationship description vector of the circuit components of the local scanning result (local result), and the multiple local results in the scanning result are randomly determined and correspond to the multiple local circuit component descriptions in the scanning result. The second recheck entity description vector is the vector of the multiple local circuit component descriptions, and its construction method is the same as that of the first recheck entity description vector.

[0129] The first review entity description vector and the second review entity description vector are matched, and a target local result meeting a preset result threshold is determined as a matching result; the matching result is a vector similarity, and the preset result threshold is, for example, 80%. If the target local result is only one, the display field can be directly located, for example, the target local result is enlarged for the annotator to view. Otherwise, the target local result needs to be globally highlighted in the scanning result first, and then the display field selected by the target personnel is obtained. The global highlighting refers to that the whole display scanning result is displayed, and then all the target local results are highlighted (for example, a color difference is formed with the non-highlighted part to display), and then the target personnel selects a certain target local result, and the display field is further enlarged to obtain the display field, which greatly improves the annotation efficiency of artificial review of the scanning result.

[0130] The application introduces a fault type trigger semantic, extracts a verification basis according to the matched trigger semantic, and performs verification according to a verification strategy corresponding to the verification basis. If the verification is passed, the verification basis is explained to attempt to obtain a field of view positioning basis. If the obtaining is successful, the display field is directly determined according to the field of view positioning basis. Otherwise, a local review semantic set within a preset time range from the trigger semantic is obtained, a first review entity description vector describing a local area circuit component distribution relationship that needs to be manually annotated is constructed according to the review semantic set, and a second review entity description vector describing a scanning result multiple local result area circuit component distribution relationship is obtained. The first review entity description vector and the second review entity description vector are matched, and a result threshold is set. According to the different number of target results, the display field is adaptively obtained, and the adaptability of the display field acquisition is improved.

[0131] The embodiment of the application provides an AOI circuit board quality inspection system for manual and machine cooperation, as shown in Figure 2 The embodiment of the application provides an AOI circuit board quality inspection system for manual and machine cooperation, as shown in

[0132] The machine scanning module 1 is used for performing AOI automatic scanning on the circuit board on the production line based on the AI quality inspection robot.

[0133] The intelligent analysis module 2 is used for performing intelligent analysis and output based on the scanning result, and determining the circuit board to be reviewed.

[0134] The artificial annotation module 3 is used for obtaining the artificial annotation information of the circuit board to be reviewed.

[0135] The model updating module 4 is used for feeding back the artificial annotation information to the AI quality inspection model of the AI quality inspection robot to update the model.

[0136] The machine scanning module is used for performing AOI automatic scanning on the circuit board on the production line based on the AI quality inspection robot, and includes the following steps.

[0137] The circuit board is fixed to the detection position.

[0138] After the fixing is completed, the global scanning is performed on the circuit board to be detected;

[0139] Based on the global scanning result, local high-precision acquisition is performed;

[0140] Based on the global scanning result, local high-precision acquisition is performed, including:

[0141] Based on the global scanning result, a suspected defect area is output;

[0142] According to the suspected defect type of the suspected defect area, the suspected defect area is grouped to obtain a region group;

[0143] Each region group is traversed in turn, and according to the suspected defect type and suspected defect area distribution of the region group being traversed, a camera configuration sub-parameter is obtained;

[0144] After each region group is traversed, a camera configuration sub-parameter group with different parameter sequences is obtained;

[0145] The camera configuration sub-parameter group with the minimum camera adjustment complexity is determined as a target parameter sequence, and local high-precision acquisition is performed according to the target parameter sequence.

[0146] In one embodiment, the artificial marking module obtains artificial marking information of the circuit board to be re-inspected, including:

[0147] The mechanical arm is used to transfer the circuit board to be re-inspected to the artificial station;

[0148] The scanning result corresponding to the circuit board to be re-inspected is displayed to the target personnel, and after the target personnel mark, the artificial marking information is obtained.

[0149] In one embodiment, the artificial marking module displays the scanning result corresponding to the circuit board to be re-inspected to the target personnel, and after the target personnel mark, the artificial marking information is obtained, including:

[0150] The re-inspection semantic sequence of the target personnel when viewing the circuit board to be re-inspected is obtained; the re-inspection semantic sequence is obtained by sorting the re-inspection semantics generated when the target personnel views the circuit board to be re-inspected according to the time sequence;

[0151] According to the re-inspection semantics and the fault type trigger semantics, the trigger semantics is determined;

[0152] According to the fault type corresponding to the preset verification basis source, the verification basis is obtained following the trigger time of the trigger semantics;

[0153] According to the verification strategy corresponding to the verification basis, the verification result is determined;

[0154] If the verification result is a verification pass, the verification basis is parsed to obtain the visual positioning basis;

[0155] If the visual positioning basis is successfully obtained, the display visual field of the scan result is determined according to the visual positioning basis;

[0156] Based on the display visual field, the scan result corresponding to the circuit board to be re-inspected is displayed to the target person.

[0157] In one embodiment, the artificial annotation module further performs the following operations:

[0158] If the visual positioning basis fails to be obtained, a re-inspection semantic set within a time range preset for a distance trigger semantic is obtained;

[0159] According to the re-inspection semantic set, a first re-inspection entity description vector is constructed;

[0160] A second re-inspection entity description vector of a plurality of local results of the scan result is obtained;

[0161] According to the matching result of the first re-inspection entity description vector and the second re-inspection entity description vector, a target local result that matches a preset result threshold is determined.

[0162] If the target local result is unique, the display visual field is determined according to the target local result.

[0163] In one embodiment, the artificial annotation module further performs the following operations:

[0164] If the target local result is not unique, the target local result is globally highlighted in the scan result, and a display visual field selected by the target person is determined.

[0165] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.

Claims

1. An artificial and machine collaborative AOI circuit board quality inspection method, characterized in that, The method comprises the following steps: Step 1: based on the AI quality inspection robot, the circuit board on the production line is scanned automatically by AOI; Step 2: based on the scanning result, intelligent analysis and output are carried out to determine the circuit board to be re-inspected; Step 3: obtaining the manual annotation information of the circuit board to be re-inspected; Step 4: feeding back the manual annotation information to the AI quality inspection model of the AI quality inspection robot to update the model; The step 1: based on the AI quality inspection robot, the circuit board on the production line is scanned automatically by AOI, comprising: Fix the circuit board to the detection position; After fixing, globally scan the circuit board to be detected; Based on the global scanning result, local high-precision collection is carried out; Among them, based on the global scanning result, local high-precision collection is carried out, comprising: Based on the global scanning result, output the suspected defect area; According to the suspected defect type of the suspected defect area, the suspected defect area is grouped to obtain the area group; Iterate each area group in turn, and according to the suspected defect type and suspected defect area distribution of the area group being iterated, obtain the camera configuration sub-parameter; After each area group is iterated, the camera configuration sub-parameter group with different parameter sequences is obtained; Determine the camera configuration sub-parameter group with the smallest camera adjustment complexity as the target parameter sequence, and carry out local high-precision collection according to the target parameter sequence; Among them, step 3: obtaining the manual annotation information of the circuit board to be re-inspected, comprising: Use the mechanical arm to transfer the circuit board to be re-inspected to the manual work station; Show the scanning result corresponding to the circuit board to be re-inspected to the target personnel, and obtain the manual annotation information after the target personnel mark; Obtain the re-inspection semantic sequence when the target personnel views the circuit board to be re-inspected; the re-inspection semantic sequence is obtained by sorting the re-inspection semantics generated when the target personnel views the circuit board to be re-inspected according to the time sequence; Match the re-inspection semantics and the fault type trigger semantics, if the semantics match successfully, the re-inspection semantics corresponding to the fault type trigger semantics are matched as the trigger semantics; According to the corresponding preset check basis of the fault type, the trigger time of the trigger semantics is obtained to obtain the check basis; According to the corresponding check strategy of the check basis, the check result is determined; If the check result is passed, analyze the check basis to obtain the field of view positioning basis; If the field of view positioning basis is obtained successfully, the display field of view of the scanning result is determined according to the field of view positioning basis; Based on the display field of view, the scanning result corresponding to the circuit board to be re-inspected is displayed to the target personnel; If the field of view positioning basis fails to be obtained, a set of re-inspection semantics within a preset time range from the trigger semantics is obtained; According to the re-inspection semantic set, a first re-inspection entity description vector is constructed; Obtain the second re-inspection entity description vector of the multiple local results of the scanning result; According to the matching result of the first re-inspection entity description vector and the second re-inspection entity description vector, the target local result meeting the preset result threshold is determined; If the target local result is unique, the display field of view is determined according to the target local result; If the target local result is not unique, the target local result is globally highlighted in the scanning result, and the display field of view selected by the target personnel is determined.

2. An artificial and machine collaborative AOI circuit board quality inspection system, characterized in that, The method comprises the following steps: The machine scanning module is configured to perform AOI automatic scanning on the circuit board on the production line based on the AI quality inspection robot. The intelligent analysis module is configured to perform intelligent analysis and output based on the scanning result, and determine the circuit board to be re-inspected. The manual annotation module is configured to obtain manual annotation information of the circuit board to be re-inspected. The model updating module is configured to feed back the manual annotation information to the AI quality inspection model of the AI quality inspection robot for model updating. The machine scanning module performs AOI automatic scanning on the circuit board on the production line based on the AI quality inspection robot, including: fixing the circuit board to the detection position; after fixing, globally scanning the circuit board to be detected; based on the global scanning result, performing local high-precision collection; wherein, based on the global scanning result, performing local high-precision collection, including: outputting suspected defect areas based on the global scanning result; grouping the suspected defect areas according to the suspected defect types of the suspected defect areas to obtain area groups; sequentially traversing each area group, and obtaining camera configuration sub-parameters according to the suspected defect types and suspected defect area distribution of the area group being traversed; after each area group is traversed, obtaining camera configuration sub-parameter groups in different parameter sequences; determining the camera configuration sub-parameter group with the smallest camera adjustment complexity as the target parameter sequence, and performing local high-precision collection according to the target parameter sequence; wherein, the manual annotation module obtains manual annotation information of the circuit board to be re-inspected, including: transferring the circuit board to be re-inspected to the manual work station by using the mechanical arm; displaying the scanning result corresponding to the circuit board to be re-inspected to the target personnel, and obtaining the manual annotation information after the target personnel annotates; obtaining a re-inspection semantic sequence when the target personnel views the circuit board to be re-inspected; the re-inspection semantic sequence is obtained by sorting the re-inspection semantics generated when the target personnel views the circuit board to be re-inspected according to the time sequence; performing semantic matching on the re-inspection semantics and the fault type trigger semantics, and if the semantic matching is successful, the re-inspection semantics corresponding to the fault type trigger semantics are taken as the trigger semantics; obtaining a verification basis according to the trigger time of the trigger semantics based on the verification basis source corresponding to the fault type; determining a verification result according to the verification strategy corresponding to the verification basis; if the verification result is a verification pass, analyzing the verification basis to obtain a field of view positioning basis; if the field of view positioning basis is successfully obtained, determining a display field of view of the scanning result according to the field of view positioning basis; displaying the scanning result corresponding to the circuit board to be re-inspected based on the display field of view; if the field of view positioning basis fails to be obtained, obtaining a re-inspection semantic set within a preset time range from the trigger semantics; constructing a first re-inspection entity description vector according to the re-inspection semantic set; obtaining a second re-inspection entity description vector of multiple local results of the scanning result; determining a target local result that matches the preset result threshold according to the matching result of the first re-inspection entity description vector and the second re-inspection entity description vector; if the target local result is unique, determining a display field of view according to the target local result; if the target local result is not unique, globally highlighting the target local result in the scanning result, and determining a display field of view selected by the target personnel.

Images