Automatic optical inspection decision memory method, device, equipment, storage medium and program product

Abstract

The application discloses an automatic optical inspection decision memory method, device, equipment, storage medium and program product. The method comprises the following steps: acquiring an image to be analyzed and corresponding metadata of industrial automatic optical inspection, generating a task state vector based on the metadata and a detection task of the image to be analyzed; in the process of generating a decision action sequence by an intelligent agent, a multi-source evidence set supporting generation of the decision action sequence is extracted based on the task state vector, and the multi-source evidence set is structured and packaged to generate an evidence package; the evidence package and the decision action sequence are stored by binding with a unified task identifier, and the evidence package and the decision action sequence are used for restoring a decision process.

Description

Technical Field

[0001] This application relates to the field of industrial automated optical inspection, and more specifically, to an automated optical inspection decision memory method, apparatus, equipment, storage medium, and program product. Background Technology

[0002] In industrial automated optical inspection (AOI) scenarios such as semiconductor wafer defect detection and panel appearance defect detection, intelligent agents have gradually replaced traditional fixed templates and static rules, undertaking key decision-making tasks such as defect detection, result review, strategy switching, and anomaly handling, becoming a core component in ensuring inspection accuracy and stable production line operation. As inspection demands become increasingly complex, agents typically need to integrate multi-source information, including specification clauses, historical cases, model versions, equipment status, and manual review results, to make dynamic judgments.

[0003] Existing industrial AOI agent systems mostly adopt a distributed logging scheme, which only retains simple records of decision-making actions such as threshold adjustment, model switching, and equipment alerts. Some systems retain operation logs, but only record the results of the actions.

[0004] However, the existing recording methods cannot restore the complete decision-making chain. Not only is it difficult to trace the cause when an anomaly occurs, which can easily lead to repeated trial and error, but it also cannot distinguish the applicable boundaries of evidence under different processes, versions, and equipment, which can easily lead to the misuse of experience and make it difficult to meet the requirements of long-term stable operation and regulatory compliance in industrial scenarios. Summary of the Invention

[0005] The main objective of this application is to provide an automatic optical inspection decision memory method, device, equipment, storage medium, and program product to solve the problems of scattered, untraceable, and easily misused decision-making basis of industrial AOI agents, and to achieve the effects of decision playback, verifiable basis, and stable system operation.

[0006] To achieve the above objectives, a first aspect of this application proposes an automatic optical inspection decision memory method, comprising: acquiring an image to be analyzed and corresponding metadata for industrial automatic optical inspection; generating a task state vector based on the metadata and the inspection task of the image to be analyzed; extracting a multi-source evidence set supporting the generation of the decision action sequence based on the task state vector during the process of an agent generating a decision action sequence; and structurally encapsulating the multi-source evidence set to generate an evidence package; binding and storing the evidence package and the decision action sequence using a unified task identifier, wherein the evidence package and the decision action sequence are used to reconstruct the decision process.

[0007] According to the automatic optical detection decision memory method provided in this application, the multi-source evidence set includes at least one of the following: normative clauses, historical cases, strategy versions, equipment operation information, and manual review records.

[0008] According to the automatic optical inspection decision memory method provided in this application, the standard clauses are derived from customer specifications, internal signed documents, and standard operating procedures; the historical cases are derived from the automatic optical inspection anomaly closed-loop record; the strategy version is derived from the skill script version, model version, and process parameter version library; the equipment operation information is derived from the equipment monitoring system, maintenance system, and operation log; and the manual review record is derived from the review platform and expert review results.

[0009] According to the automatic optical detection decision memory method provided in this application, each piece of evidence in the evidence package includes at least one of the following: evidence identifier, source path, version number, effective domain, confidence level, summary explanation, and timestamp.

[0010] According to the automatic optical detection decision memory method provided in this application, the decision action sequence includes at least one of the following: main execution action, verification action, and rollback trigger condition.

[0011] According to the automatic optical detection decision memory method provided in this application, after binding and storing the evidence package and the decision action sequence using a unified task identifier, the method further includes: obtaining a target task identifier; reading the corresponding target evidence package and target decision action sequence according to the target task identifier; performing decision playback based on the target evidence package and the target decision action sequence and outputting an audit report; wherein the content of the decision playback includes at least one of the following: decision triggering conditions, execution path and result indicators, and the audit report includes at least one of the following: referenced normative clauses, triggered historical cases, model version or skill script version used, current device status, action execution order and final result description.

[0012] This application also provides an automatic optical inspection decision memory device, comprising the following modules: an acquisition module and a processing module; the acquisition module is used to acquire the image to be analyzed and the corresponding metadata for industrial automatic optical inspection; the processing module is used to generate a task state vector based on the metadata and the inspection task of the image to be analyzed; during the process of the agent generating a decision action sequence, a multi-source evidence set supporting the generation of the decision action sequence is extracted based on the task state vector, and the multi-source evidence set is structurally encapsulated to generate an evidence package; a unified task identifier is used to bind and store the evidence package and the decision action sequence, and the evidence package and the decision action sequence are used to reconstruct the decision process.

[0013] According to the automatic optical detection decision memory device provided in this application, the multi-source evidence set includes at least one of the following: normative clauses, historical cases, strategy versions, equipment operation information, and manual review records.

[0014] According to the automatic optical inspection decision memory device provided in this application, the specified terms are derived from customer specifications, internal signed documents, and standard operating procedures; the historical cases are derived from the automatic optical inspection anomaly closed-loop record; the strategy version is derived from the skill script version, model version, and process parameter version library; the equipment operation information is derived from the equipment monitoring system, maintenance system, and operation log; and the manual review record is derived from the review platform and expert review results.

[0015] According to the automatic optical detection decision memory device provided in this application, each piece of evidence in the evidence package includes at least one of the following: evidence identifier, source path, version number, effective domain, confidence level, summary explanation, and timestamp.

[0016] According to the automatic optical detection decision memory device provided in this application, the decision action sequence includes at least one of the following: main execution action, verification action, and rollback trigger condition.

[0017] According to the automatic optical detection decision memory device provided in this application, after binding and storing the evidence package and the decision action sequence using a unified task identifier, the acquisition module is used to acquire the target task identifier; the processing module is used to read the corresponding target evidence package and target decision action sequence according to the target task identifier; perform decision playback based on the target evidence package and the target decision action sequence and output an audit report; wherein, the content of the decision playback includes at least one of the following: decision triggering conditions, execution path and result indicators, and the audit report includes at least one of the following: referenced normative clauses, triggered historical cases, model version or skill script version used, current device status, action execution order and final result description.

[0018] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the automatic optical detection decision memory method as described above.

[0019] This application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the automatic optical detection decision memory method as described above.

[0020] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the automatic optical detection decision memory method as described above.

[0021] The technical solutions provided by the embodiments of this application can include the following beneficial effects: Since a multi-source evidence set supporting the decision-making action sequence can be extracted based on the task state vector, and the multi-source evidence set can be structurally encapsulated to generate an evidence package, it is possible to completely collect the normative clauses, historical cases, model versions, and other multi-source evidence relied upon by the intelligent agent's decision-making, forming a standardized evidence package, thereby achieving complete and standardized retention of the decision-making basis; Since a unified task identifier is used to bind and store the evidence package with the decision-making action sequence, the decision-making chain can be completely restored through the evidence package and the decision-making action sequence, quickly tracing the cause of the decision when production line anomalies occur, avoiding repeated trial and error; At the same time, the structured and encapsulated evidence package can clearly distinguish the applicable boundaries of evidence under different processes, different versions, and different equipment, effectively preventing the misuse of experience, thereby improving the accuracy and stability of cross-scenario experience reuse, and further enhancing the operational stability, decision credibility, and regulatory compliance of the industrial automatic optical inspection system. Attached Figure Description

[0022] The accompanying drawings, which form part of this application, are used to provide a further understanding of the application and to make other features, objects, and advantages of the application more apparent. The illustrative embodiments and descriptions of this application are used to explain the application and do not constitute an undue limitation of the application. In the drawings: Figure 1 A flowchart illustrating the automated optical detection decision memory method provided in this application; Figure 2 This is a schematic diagram of the structure of the automatic optical detection decision memory device provided by the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present application.

[0024] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0025] In this application, the terms "upper," "lower," "left," "right," "front," "rear," "top," "bottom," "inner," "outer," "middle," "vertical," "horizontal," "lateral," and "longitudinal" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are primarily for the purpose of better describing this application and its embodiments, and are not intended to limit the indicated device, element, or component to having a specific orientation, or to be constructed and operated in a specific orientation.

[0026] Furthermore, in addition to indicating location or positional relationship, some of the aforementioned terms may also have other meanings. For example, the term "above" may also be used in some cases to indicate a certain dependency or connection relationship. Those skilled in the art can understand the specific meaning of these terms in this application based on the specific circumstances.

[0027] Furthermore, the terms "installation," "setup," "equipped with," "connection," "linked," and "socketing" should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral structure; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or an internal connection between two devices, components, or parts. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0028] This application describes some exemplary embodiments for illustrative purposes. It should be understood that this application may be implemented in other ways not specifically shown in the accompanying drawings.

[0029] like Figure 1 As shown, this application provides an automatic optical detection decision memory method, which can be applied to an automatic optical detection decision memory device. The automatic optical detection decision memory method may include steps S101-S103: S101. The automatic optical inspection decision memory device acquires the image to be analyzed and the corresponding metadata for industrial automatic optical inspection, and generates a task state vector based on the metadata and the inspection task of the image to be analyzed.

[0030] Specifically, the automated optical inspection decision memory device can acquire images to be analyzed from semiconductor automated optical inspection production lines or panel automated optical inspection production lines, and simultaneously extract metadata corresponding to the images to be analyzed. This metadata may include at least one or more of the following: process parameters, production line number, product model, batch number, time window, optical configuration information, equipment health score, model version, process parameter version, or review result. The automated optical inspection decision memory device can associate and integrate the above metadata with task information corresponding to the images to be analyzed, such as defect type, inspection area, equipment number, product model, batch status, process layer number, camera number, light source mode, and anomaly level, to establish a mapping relationship between evidence and task status, forming a task status vector used to define the applicable boundaries of the evidence.

[0031] It is understandable that by generating task state vectors, precise applicable boundary constraints can be provided for the evidence set, ensuring the relevance and accuracy of evidence extraction and decision-making, and improving the reliability and consistency of decision-making basis.

[0032] Optionally, in scenarios such as front-end micro-defect inspection of semiconductor wafers, package appearance inspection, and chip surface contamination inspection, the automatic optical inspection decision memory device can extract metadata related to wafer batch, package type, machine number, camera number, light source mode, process parameter version, model version, and equipment health score, and combine it with task information such as defect type, inspection area, process layer number, and anomaly level to generate a task state vector adapted to the semiconductor inspection scenario.

[0033] Optionally, in scenarios such as panel Mura inspection, bright spot inspection, dark spot inspection, line defect inspection, stain inspection, and edge anomaly inspection, the automatic optical inspection decision memory device can extract metadata related to product model, backlight mode, exposure time, conveyor speed, batch number, optical configuration, maintenance cycle, and environmental conditions. Combined with task information such as defect type, inspection area, and anomaly level, it can generate a task state vector adapted to the panel inspection scenario. S102. During the process of generating a decision action sequence by an intelligent agent, the automatic optical detection decision memory device extracts a set of multi-source evidence supporting the generation of the decision action sequence based on the task state vector, and performs structured encapsulation of the multi-source evidence set to generate an evidence package.

[0034] Specifically, during the entire process of generating a decision action sequence by an intelligent agent, the automatic optical detection decision memory device can use the task state vector as the retrieval and filtering basis to selectively extract a set of multi-source evidence that matches the current detection task, equipment status, and process conditions. It establishes a one-to-one association mapping between the evidence and the task state vector, ensuring that all extracted evidence is a valid basis for supporting the current decision generation. Then, the selected set of multi-source evidence is uniformly organized and structured. The fields of each piece of evidence are filled and the information is solidified according to a preset format to form an evidence package with a unified structure that can be parsed by the program and audited and played back. This completes the transformation of evidence from a scattered state to a standardized structured object.

[0035] Understandably, by extracting evidence based on task state vectors, irrelevant evidence can be accurately filtered out, improving the validity of evidence and the rationality of decision-making; through structured encapsulation, scattered evidence can be unified and standardized, improving the convenience of subsequent storage, retrieval, playback and auditing.

[0036] It should be noted that the multi-source evidence set refers to the complete set of evidence that supports the intelligent agent in making decisions on detection, re-judgment, strategy switching and anomaly handling, and is used to completely restore the decision basis; the evidence package refers to the decision basis carrier that is not arbitrarily tampered with after the multi-source evidence is standardized and structured and packaged.

[0037] Optionally, the multi-source evidence set includes at least one of the following: regulatory clauses, historical cases, strategy versions, equipment operation information, and manual review records. The regulatory clauses originate from customer specifications, internal signed documents, and standard operating procedures; the historical cases originate from closed-loop records of automatic optical inspection anomalies; the strategy versions originate from skill script versions, model versions, and process parameter version libraries; the equipment operation information originates from equipment monitoring systems, maintenance systems, and operation logs; and the manual review records originate from the review platform and expert review results.

[0038] Specifically, regulatory clauses can be extracted from customer specifications, internal signed documents, and standard operating procedures to provide compliance and standardization basis for decision-making; historical cases can be extracted from the closed-loop records of automatic optical inspection anomalies to provide historical experience reference for decision-making; strategy versions can be extracted from skill script versions, model versions, and process parameter version libraries to clarify the execution logic and parameter versions adopted for decision-making; equipment operation information can be extracted from equipment monitoring systems, maintenance systems, and operation logs to reflect the current working status and operating conditions of the equipment; and manual review records can be extracted from review platforms and expert review results to provide manual verification and confirmation basis for decision-making. Through the fusion and extraction of multi-dimensional information, a complete, comprehensive, and decision-supporting multi-source evidence set can be formed. The automatic optical inspection decision memory device can uniformly number and locate the source of the above multi-source evidence set, thereby laying the foundation for subsequent encapsulation and tracking.

[0039] Optionally, each piece of evidence in the evidence package includes at least one of the following: evidence identifier, source path, version number, effective domain, confidence level, summary explanation, and timestamp.

[0040] Specifically, the automated optical inspection decision memory device can standardize field assignments and record information for each piece of evidence packaged into the evidence package. It assigns a globally unique evidence identifier to each piece of evidence for independent identification and traceability, records the source path of the evidence to clarify the data acquisition channel, marks version numbers to distinguish iterative versions of strategies, models, and process parameters, sets effective domains to limit the applicable product models, process layers, equipment families, and time windows, assigns confidence levels to characterize the reliability of the evidence, adds a summary explanation to describe the role and support direction of the evidence in this decision-making process, and records a timestamp to mark the exact moment of evidence extraction and packaging. Through the recording of these complete fields, each piece of evidence possesses the attributes of traceability, verifiability, delimitability, and interpretability, thereby forming a structured evidence package that can be parsed and audited. After the evidence items are packaged, the automated optical inspection decision memory device can perform a hash signature on the entire evidence package to prevent subsequent tampering.

[0041] Optionally, the decision-making action sequence includes at least one of the following: a main execution action, a verification action, and a rollback trigger condition. The main execution action may include at least one or more of the following: threshold adjustment, skill script switching, model switching, process parameter switching, re-judgment escalation, or equipment warning. The verification action may include at least one or more of the following: statistical sampling verification, full-scale re-inspection verification, or sampled manual review. The rollback trigger condition may be one or more of the following: false positive rate exceeding a threshold, false negative rate exceeding a threshold, abnormally amplified defect density, or cycle time degradation.

[0042] Specifically, the main execution action is the core operation directly implemented by the intelligent agent for the current detection task, used to complete detection adjustment, strategy switching, re-judgment upgrade and equipment early warning; the verification action refers to the auxiliary operation to verify and confirm the execution result of the main execution action, used to ensure the accuracy and effectiveness of decision execution; the rollback trigger condition is the preset anomaly judgment rule, which triggers the system to restore to the previous stable state when the detection index deteriorates. The three work together to form a complete, controllable decision execution logic with security protection capabilities.

[0043] S103. The automatic optical detection decision memory device uses a unified task identifier to bind and store the evidence package and the decision action sequence.

[0044] The evidence package and the decision action sequence are used to reconstruct the decision-making process.

[0045] Specifically, the automatic optical inspection decision memory device can assign a globally unique unified task identifier to each independent inspection decision. Using this unified task identifier as the association key, a strong association is established between the evidence package and the decision action sequence corresponding to the same decision. The associated evidence package and decision action sequence are synchronously written into the long-term memory system, and a searchable index structure is constructed with dimensions such as task identifier, defect type, product model, batch number, version information, and time window. This enables long-term stable storage and rapid location and retrieval of the evidence package and decision action sequence. When the evidence package and the decision action sequence are invoked, the decision basis, execution logic, and operation process of the intelligent agent can be completely restored, realizing a full-link review of the decision-making process.

[0046] Understandably, by binding and storing evidence packages and decision-making action sequences with unified task identifiers, it is possible to ensure that each evidence package corresponds one-to-one with the decision-making action sequence, providing a unique retrieval basis for subsequent decision playback, accountability tracing, and compliance auditing; it is possible to achieve integrated and persistent retention of decision-making basis and execution actions, forming a decision memory that can be reviewed, reused, and verified, thereby improving the operational stability and decision credibility of industrial automated optical inspection systems; it is possible to quickly locate historical decision data and restore the complete decision-making chain, reducing the cost of anomaly analysis and experience reuse.

[0047] It should be noted that the unified task identifier is a unique number used to identify a single detection decision, enabling precise binding of evidence packages and decision action sequences; the long-term memory system is a storage system used to persistently store decision-related data, supporting decision playback, auditing, and experience reuse.

[0048] Optionally, while binding the evidence package and decision action sequence into the long-term memory system, the automated optical inspection decision memory device can also simultaneously write the execution results and post-execution quality indicators corresponding to this decision. The execution results are used to characterize the actual completion of the decision action sequence, and the post-execution quality indicators include at least one or more of the following: change in false detection rate, change in false negative rate, change in defect density, consistency of re-judgment, and change in cycle time. The automated optical inspection decision memory device associates the above execution results and post-execution quality indicators together with a unified task identifier, which, together with the evidence package and decision action sequence, constitutes a complete decision memory record. This can further enrich the data dimensions of decision review and audit verification, and provide quantitative basis for experience reuse, strategy optimization, and effect evaluation for subsequent similar tasks.

[0049] For example, in a semiconductor inspection scenario, when the micro-defect density of a certain process layer in the wafer front-end increases significantly within a short period of time, the agent can combine information such as the performance of the current model version, historical anomaly cases of the same process layer, and equipment health scores to generate a decision action sequence that rolls back the model version and increases the re-judgment ratio. The automated optical inspection decision memory device can encapsulate the historical cases, version records, equipment status, and decision action sequence supporting this decision into an evidence package and bind it to a long-term memory system, facilitating rapid retrieval and experience reuse for similar anomalies in the future.

[0050] In panel inspection scenarios, when a batch of organic light-emitting diode (OLED) products exhibits increased alarms due to uneven local brightness, and the equipment maintenance system reports abnormal light source attenuation scores, the intelligent agent can generate a decision-making sequence to activate the uniformity correction sub-link and output an equipment maintenance warning. The automated optical inspection decision memory device can encapsulate the equipment maintenance records, light source status, historical cases, and action execution sequence upon which this decision is based into an evidence package and bind it for storage. This allows engineers to clearly distinguish the cause of the anomaly and accurately determine whether it is a process anomaly or an equipment status anomaly.

[0051] Optionally, after binding and storing the evidence package and the decision action sequence using a unified task identifier, the automatic optical detection decision memory device can obtain the target task identifier; read the corresponding target evidence package and target decision action sequence according to the target task identifier; perform decision playback based on the target evidence package and the target decision action sequence and output an audit report; wherein, the content of the decision playback includes at least one of the following: decision triggering conditions, execution path and result indicators, and the audit report includes at least one of the following: referenced normative clauses, triggered historical cases, model version or skill script version used, current device status, action execution order and final result description.

[0052] Specifically, after binding and storing the evidence package and the decision action sequence, the automatic optical detection decision memory device can obtain the target task identifier according to external query commands or anomaly analysis needs. Using the target task identifier as the retrieval key, it can accurately retrieve the uniquely corresponding target evidence package and target decision action sequence from the long-term memory system. Based on the various evidence information in the target evidence package and the execution logic in the target decision action sequence, it can completely reproduce the decision generation and execution process of the intelligent agent in chronological order and dependency relationship, realize decision playback and generate a standardized audit report. Decision playback is used to intuitively reproduce the decision triggering conditions, the overall execution path and related result indicators. The audit report is used to fully record the normative clauses cited in the decision, the triggered historical cases, the model version or skill script version used, the equipment operating status, the action execution order and the final result description.

[0053] It should be noted that by quickly retrieving the corresponding target evidence package and target decision action sequence based on the target task identifier, and by performing decision playback and outputting standardized audit reports based on the two, the decision triggering conditions, overall execution path and final result indicators of the intelligent agent can be completely and clearly reproduced, realizing full-link traceability, verifiability and auditability of the decision-making process; it can quickly locate the root cause of production line anomalies or decision errors, efficiently complete problem analysis and responsibility tracing, and significantly reduce the cost of anomaly investigation and experience reuse; it can also provide real, complete and traceable data support for decision optimization, model iteration and process improvement, promote the long-term stable evolution of industrial automatic optical inspection systems, and meet the core needs of industry regulatory compliance and quality system traceability.

[0054] In this embodiment, since a multi-source evidence set supporting the decision-making action sequence can be extracted based on the task state vector, and the multi-source evidence set can be structurally encapsulated to generate an evidence package, it is possible to completely collect the normative clauses, historical cases, model versions, and other multi-source evidence on which the intelligent agent's decision-making depends, forming a standardized evidence package, thereby achieving complete and standardized retention of the decision-making basis. Since a unified task identifier is used to bind and store the evidence package with the decision-making action sequence, the decision-making chain can be completely restored through the evidence package and the decision-making action sequence, and the cause of the decision can be quickly traced when production line anomalies occur, avoiding repeated trial and error. At the same time, the structured encapsulated evidence package can clearly distinguish the applicable boundaries of evidence under different processes, different versions, and different equipment, effectively preventing the misuse of experience, thereby improving the accuracy and stability of cross-scenario experience reuse, and thus enhancing the operational stability, decision credibility, and regulatory compliance of the industrial automatic optical inspection system.

[0055] The foregoing mainly describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, it includes corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0056] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0057] The automatic optical detection decision memory method provided in this application can be executed by an automatic optical detection decision memory device, or a control module for automatic optical detection decision memory within that device. This application uses the execution of the automatic optical detection decision memory method by an automatic optical detection decision memory device as an example to illustrate the automatic optical detection decision memory device provided in this application.

[0058] It should be noted that the embodiments of this application can divide the automatic optical detection decision memory device into functional modules according to the above method examples. For example, each function can be divided into its own functional modules, or two or more functions can be integrated into one processing module. The integrated modules can be implemented in hardware or as software functional modules. Optionally, the module division in the embodiments of this application is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0059] like Figure 2 As shown in the figure, this application embodiment provides an automatic optical detection decision memory device 200. The automatic optical detection decision memory device 200 includes: an acquisition module 201 and a processing module 202.

[0060] The acquisition module 201 is used to acquire the image to be analyzed and the corresponding metadata of industrial automatic optical inspection; The processing module 202 is used to generate a task state vector based on the metadata and the detection task of the image to be analyzed; during the process of the agent generating a decision action sequence, it extracts a multi-source evidence set that supports the generation of the decision action sequence based on the task state vector, and performs structured encapsulation of the multi-source evidence set to generate an evidence package; it uses a unified task identifier to bind and store the evidence package and the decision action sequence, and the evidence package and the decision action sequence are used to reconstruct the decision process.

[0061] Optionally, the multi-source evidence set includes at least one of the following: normative clauses, historical cases, policy versions, equipment operation information, and manual review records.

[0062] Optionally, the specified terms are derived from customer specifications, internal signed documents, and standard operating procedures; the historical cases are derived from the closed-loop records of abnormalities in automatic optical inspection; the strategy version is derived from the skill script version, model version, and process parameter version library; the equipment operation information is derived from the equipment monitoring system, maintenance system, and operation logs; and the manual review records are derived from the review platform and expert review results.

[0063] Optionally, each piece of evidence in the evidence package includes at least one of the following: evidence identifier, source path, version number, effective domain, confidence level, summary explanation, and timestamp.

[0064] Optionally, the decision action sequence includes at least one of the following: main execution action, verification action, and rollback trigger condition.

[0065] Optionally, after binding and storing the evidence package and the decision action sequence using a unified task identifier, the acquisition module 201 is used to acquire the target task identifier; the processing module 202 is used to read the corresponding target evidence package and target decision action sequence according to the target task identifier; perform decision playback based on the target evidence package and the target decision action sequence and output an audit report; wherein, the content of the decision playback includes at least one of the following: decision triggering conditions, execution path and result indicators, and the audit report includes at least one of the following: referenced normative clauses, triggered historical cases, model version or skill script version used, current device status, action execution order and final result description.

[0066] In this embodiment, since a multi-source evidence set supporting the decision-making action sequence can be extracted based on the task state vector, and the multi-source evidence set can be structurally encapsulated to generate an evidence package, it is possible to completely collect the normative clauses, historical cases, model versions, and other multi-source evidence on which the intelligent agent's decision-making depends, forming a standardized evidence package, thereby achieving complete and standardized retention of the decision-making basis. Since a unified task identifier is used to bind and store the evidence package with the decision-making action sequence, the decision-making chain can be completely restored through the evidence package and the decision-making action sequence, and the cause of the decision can be quickly traced when production line anomalies occur, avoiding repeated trial and error. At the same time, the structured encapsulated evidence package can clearly distinguish the applicable boundaries of evidence under different processes, different versions, and different equipment, effectively preventing the misuse of experience, thereby improving the accuracy and stability of cross-scenario experience reuse, and thus enhancing the operational stability, decision credibility, and regulatory compliance of the industrial automatic optical inspection system.

[0067] Figure 3 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include a processor 310, a communications interface 320, a memory 330, and a communication bus 340, wherein the processor 310, communications interface 320, and memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions in the memory 330 to execute an automatic optical inspection decision memory method. This method includes: acquiring the image to be analyzed and its corresponding metadata for industrial automatic optical inspection; generating a task state vector based on the metadata and the inspection task of the image to be analyzed; extracting a multi-source evidence set supporting the generation of the decision action sequence based on the task state vector during the generation of the decision action sequence during the process of the agent generating the decision action sequence; and structurally encapsulating the multi-source evidence set to generate an evidence package; binding and storing the evidence package and the decision action sequence using a unified task identifier; the evidence package and the decision action sequence are used to reconstruct the decision process.

[0068] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0069] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the automatic optical detection decision memory method provided by the above methods. The method includes: acquiring an image to be analyzed and corresponding metadata for industrial automatic optical detection; generating a task state vector based on the metadata and the detection task of the image to be analyzed; extracting a multi-source evidence set supporting the generation of the decision action sequence based on the task state vector during the process of the agent generating a decision action sequence, and structurally encapsulating the multi-source evidence set to generate an evidence package; binding and storing the evidence package and the decision action sequence using a unified task identifier, wherein the evidence package and the decision action sequence are used to reconstruct the decision process.

[0070] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the automatic optical inspection decision memory method provided by the above methods. The method includes: acquiring an image to be analyzed and corresponding metadata for industrial automatic optical inspection; generating a task state vector based on the metadata and the inspection task of the image to be analyzed; extracting a multi-source evidence set supporting the generation of the decision action sequence based on the task state vector during the process of the agent generating a decision action sequence, and structurally encapsulating the multi-source evidence set to generate an evidence package; binding and storing the evidence package and the decision action sequence using a unified task identifier, wherein the evidence package and the decision action sequence are used to reconstruct the decision process.

[0071] The device embodiments described above are merely illustrative. 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; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0072] Obviously, those skilled in the art should understand that the various units or steps of this application described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device, or fabricating them separately as individual integrated circuit modules, or fabricating multiple modules or steps into a single integrated circuit module. Thus, this application is not limited to any particular combination of hardware and software.

[0073] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. An automatic optical detection decision memory method, characterized in that, include: Acquire the image to be analyzed and the corresponding metadata for industrial automated optical inspection, and generate a task state vector based on the metadata and the inspection task of the image to be analyzed; During the process of generating a decision action sequence by an intelligent agent, a multi-source evidence set supporting the generation of the decision action sequence is extracted based on the task state vector, and the multi-source evidence set is structurally encapsulated to generate an evidence package. A unified task identifier is used to bind and store the evidence package and the decision action sequence, which are used to reconstruct the decision-making process.

2. The automatic optical detection decision memory method according to claim 1, characterized in that, The multi-source evidence set includes at least one of the following: normative clauses, historical cases, strategy versions, equipment operation information, and manual review records.

3. The automatic optical detection decision memory method according to claim 2, characterized in that, The specified terms are derived from customer specifications, internal signed documents, and standard operating procedures; the historical cases are derived from the closed-loop records of abnormalities in automatic optical inspection; the strategy versions are derived from skill script versions, model versions, and process parameter version libraries; the equipment operation information is derived from the equipment monitoring system, maintenance system, and operation logs; and the manual review records are derived from the review platform and expert review results.

4. The automatic optical detection decision memory method according to claim 1, characterized in that, Each piece of evidence in the evidence package includes at least one of the following: evidence identifier, source path, version number, effective domain, confidence level, summary explanation, and timestamp.

5. The automatic optical detection decision memory method according to claim 1, characterized in that, The decision action sequence includes at least one of the following: main execution action, verification action, and rollback trigger condition.

6. The automatic optical detection decision memory method according to any one of claims 1-5, characterized in that, After binding and storing the evidence package and the decision action sequence using a unified task identifier, the method further includes: Obtain the target task identifier; Read the corresponding target evidence package and target decision action sequence based on the target task identifier; Based on the target evidence package and the target decision action sequence, the decision playback is performed and an audit report is output; The decision replay includes at least one of the following: decision triggering conditions, execution path and result indicators. The audit report includes at least one of the following: referenced normative clauses, historical triggering cases, model version or skill script version used, current equipment status, action execution order and final result description.

7. An automatic optical detection decision memory device, characterized in that, include: Acquisition module and processing module; The acquisition module is used to acquire the image to be analyzed and the corresponding metadata of industrial automated optical inspection; The processing module is used to generate a task state vector based on the metadata and the detection task of the image to be analyzed; during the process of the agent generating a decision action sequence, it extracts a multi-source evidence set that supports the generation of the decision action sequence based on the task state vector, and performs structured encapsulation of the multi-source evidence set to generate an evidence package. A unified task identifier is used to bind and store the evidence package and the decision action sequence, which are used to reconstruct the decision-making process.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the automatic optical detection decision memory method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the automatic optical detection decision memory method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the automatic optical detection decision memory method as described in any one of claims 1 to 6.

Images