Production line quality detection system and method based on iPhone-Mac cooperation

Abstract

The invention relates to the technical field of industrial quality inspection, in particular to a production line quality detection system and method based on iPhone-Mac collaboration.The production line quality detection system comprises an iPhone image acquisition terminal, a Mac control terminal, a production line controller and a data storage unit, the iPhone image acquisition terminal is in communication connection with the Mac control terminal through a USB data line, the Mac control terminal is connected with a PLC through an industrial protocol, and the production line controller is connected with the data storage unit. The iPhone image acquisition terminal runs EagleEyeFront software and is used for acquiring a high-resolution original image and reporting an equipment identity identifier, the Mac control terminal runs EagleEyeMacro software and realizes multi-terminal collaborative management, local AI defect identification and control instruction generation, the AI reasoning module adopts a defect detection model trained by a YOLO framework, and the AI reasoning module adopts a defect detection model trained by a YOLO framework and is used for detecting the equipment identity identifier. And the defects of screw missing, crown color deviation and the like can be identified.

Description

Technical Field

[0001] This invention relates to the field of industrial quality inspection technology, specifically to a production line quality inspection system and method based on iPhone-Mac collaboration. Background Technology

[0002] In modern production lines in industries such as electronics and machinery, quality inspection is a core link in ensuring product consistency and reliability. This is especially true for precision parts such as mobile phone components, which require precise identification of minute defects such as missing screws, crown color deviations, and improper battery installation. Current mainstream production line quality inspection solutions mostly adopt an architecture of "heterogeneous hardware splicing and loose functional coupling," relying on the collaboration of multiple devices such as independent industrial cameras, Windows / Linux industrial control computers, and remote AI servers. While this can achieve basic inspection functions, it has significant shortcomings in transmission stability, image quality, equipment management, real-time performance, and adaptive capabilities. It is difficult to adapt to the stringent requirements of high-speed intelligent manufacturing for high precision, high reliability, and low latency. Specific problems are as follows: Existing solutions generally rely on Wi-Fi or Bluetooth wireless communication to achieve data interaction between image acquisition terminals and processing terminals. Electromagnetic interference in industrial environments, such as electromagnetic radiation generated by machine tools and frequency converters, directly leads to a sharp drop in the stability of wireless links. In practical applications, the connection interruption rate of wireless transmission often exceeds 8%, and the recovery time for a single interruption is 30-60 seconds. On high-frequency production lines with a production rate of more than 30 pieces per minute, data packet loss can easily lead to missed inspections, resulting in defective products flowing into the next process. Some solutions attempt to optimize the signal by increasing the number of wireless access points (APs), but the roaming latency during multi-AP switching still cannot meet the requirements for real-time image transmission, and the equipment deployment cost is more than 40% higher than that of wired solutions.

[0003] High-speed production lines often require acquiring workpiece images from multiple perspectives to cover the entire surface, but existing solutions lack a unified equipment identification and centralized management mechanism. With multiple data acquisition terminals, such as multiple industrial cameras, there is no unique identification code. The image data is only named with timestamps and random numbers. When quality problems occur, it is impossible to quickly trace back to the specific acquisition station and equipment. Manual inspection is required one by one, and the problem location time exceeds 30 minutes. Equipment status monitoring relies on manual inspections. Anomalies such as terminal offline and camera malfunctions need to be discovered on-site by operators. The average time to discover a fault exceeds 2 hours, during which the production line is in a state of no detection or false detection, significantly increasing the quality risk.

[0004] Some improvement solutions attempt to identify devices using QR codes, but this requires additional hardware, such as barcode scanners and manual scanning operations, increasing production line complexity and labor costs.

[0005] Existing AI-based defect identification solutions largely rely on remote cloud servers or standalone industrial control computers, resulting in significant real-time bottlenecks and limitations in adaptive capabilities. In the remote inference solution, the total latency of uploading images to the cloud, model analysis, and result feedback exceeds 10 seconds, which cannot match the millisecond-level closed-loop requirements of high-speed production lines for taking pictures, analyzing, and making decisions, resulting in a forced reduction of production line cycle time by more than 30%. Local inference solutions, such as those based on Windows industrial PCs, reduce network latency, but the communication protocols of industrial PCs are incompatible with those of data acquisition terminals and production line controllers, requiring the development of additional middleware. Furthermore, model iterations must be completed offline. When new defects appear, production data must be exported to an independent training platform, the model must be trained, and then it must be manually deployed back to the detection system. The entire cycle takes 2-4 weeks, during which the quality inspection system cannot effectively identify new defects, and the defect rate rises sharply.

[0006] Existing improved technologies, such as the "Industrial Vision Inspection System" disclosed in CN115830721A and the "Quality Inspection Method Based on Mobile Terminal" disclosed in CN114662345A, have optimized the communication or image acquisition links respectively, but have not formed an integrated collaborative architecture of acquisition, transmission, analysis, decision-making, traceability and iteration, and still cannot fundamentally solve the above-mentioned multi-dimensional defects. Summary of the Invention

[0007] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a production line quality inspection system and method based on iPhone-Mac collaboration.

[0008] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: The production line quality inspection system based on iPhone-Mac collaboration of the present invention includes an iPhone image acquisition terminal, a Mac control terminal, a production line controller, and a data storage unit; The iPhone image acquisition terminal is connected to the Mac control terminal via a USB data cable. The Mac control terminal is electrically connected to the production line controller via an industrial protocol. The Mac control terminal is also connected to the data storage unit via a network. The iPhone image acquisition terminal is used to acquire original images of workpieces on the production line and report the equipment identification. The Mac control terminal is used for multi-terminal collaborative management, local AI defect identification and control command generation. The production line controller is used to receive commands and execute workpiece release or interception actions.

[0009] Preferably, the iPhone image acquisition terminal runs EagleEyeFront software, including a USB communication module, an image acquisition module, an identity management module, and an anomaly handling module; The USB communication module is used to establish a full-duplex wired connection with the Mac control terminal. The image acquisition module calls the iPhone's underlying camera API to obtain lossless original images. The identity management module stores and reports the station code and camera number. The exception handling module monitors the USB connection status and camera permissions and triggers automatic reconnection or authorization guidance.

[0010] More preferably, the original image acquired by the image acquisition module is in RAW or uncompressed bitmap format, and the pixel resolution is not lower than the highest hardware pixel of the iPhone camera; The identity management module persistently stores the station code and camera number on the iPhone. The default value is loaded when it is used for the first time, and users can modify it through the graphical interface and perform non-empty verification.

[0011] Preferably, the Mac control terminal runs EagleEyeMacross software, including a multi-terminal management module, an AI inference module, a control command generation module, and a data management module. The multi-terminal management module automatically scans and authenticates USB-connected iPhone terminals, displaying online / offline status using a status indicator array. The AI ​​inference module deploys a defect detection model trained on the local YOLO framework. The control command generation module outputs release / interception commands based on the AI ​​results. The data management module stores images and detection data according to a standardized directory.

[0012] Preferably, the multi-terminal management module supports simultaneous connection of at least 2-8 iPhone terminals and has a built-in automatic reconnection mechanism after disconnection; the defect identification range of the AI ​​inference module includes missing screws, crown color deviation, and poor battery installation, and the inference confidence threshold can be set to 80%-95%.

[0013] More preferably, the production line controller is a PLC controller that communicates with the Mac control terminal via Modbus or Profinet protocol; When the Mac control terminal outputs a qualified instruction, the production line controller controls the production line to release the workpiece; when it outputs a non-qualified instruction, it controls the production line to intercept the workpiece and triggers an audible and visual alarm.

[0014] Preferably, the data storage unit includes a local database on the Mac control terminal and a remote PC server. The local database stores the original images in a directory structure of ReceivedImages, station code, camera number, and detection timestamp. The remote PC server synchronously stores the full detection data, including image data, AI inference results, control commands, and device identification, and supports access and traceability through the Apple Insight platform.

[0015] Preferably, the exception handling module includes a permission monitoring submodule and a connection monitoring submodule; When the permission monitoring submodule detects that camera permission has been denied, a graphical pop-up window will appear to guide the user to the system settings for authorization. When the connection monitoring submodule detects a USB disconnection, it continuously scans the USB port and attempts to reconnect automatically, while displaying a red fault indicator light and a connection failure message on the interface.

[0016] More preferably, the Mac control terminal also includes a model iteration module, which has a built-in Web annotation platform and model training toolchain; The web annotation platform allows users to select image defects and add labels to generate standard datasets. The model training toolchain supports configuring hyperparameters such as learning rate and training epochs. Newly trained models can be deployed to the AI ​​inference module with one click.

[0017] A further preferred production line quality inspection method based on iPhone-Mac collaboration includes the following steps: 1) System initialization: The Mac control terminal scans the iPhone terminal connected via USB, completes identity authentication and connection establishment, and the iPhone terminal loads and reports the station code and camera number; 2) Trigger detection: The production line controller sends a device arrival signal to the Mac control terminal, and the Mac control terminal broadcasts a synchronous photo-taking command to all online iPhone terminals; 3) Image Acquisition and Transmission: The iPhone terminal uses the camera to acquire raw images and transmits them to the Mac control terminal without loss via USB data cable; 4) AI Defect Recognition: The Mac control terminal calls the local AI model to analyze the image and determines whether the workpiece is qualified or unqualified based on the confidence threshold; 5) Control feedback: The Mac control terminal outputs release / interception commands to the production line controller and simultaneously synchronizes the detection data to the data storage unit; 6) Anomaly Handling: If a USB disconnection or lack of permissions is detected, the iPhone terminal will trigger automatic reconnection or authorization boot, and the Mac control terminal will record the fault log.

[0018] (III) Beneficial Effects Compared with existing technologies, this invention provides a production line quality inspection system and method based on iPhone-Mac collaboration, which has the following beneficial effects: This technical solution achieves efficient and stable image transmission through a USB full-duplex wired connection between the iPhone image acquisition terminal and the Mac control terminal. The iPhone device utilizes its high-quality camera to call the underlying API to acquire lossless original images, ensuring the accuracy and clarity of image acquisition, which is crucial for accurately identifying product defects. Furthermore, the identification management module ensures the uniqueness of each station code and camera number, improving the system's traceability and management efficiency.

[0019] The Mac control terminal runs EagleEyeMacross software, integrating multi-terminal management, AI inference, control command generation, and data management functions. It supports simultaneous connections from 2 to 8 iPhones and features a built-in automatic reconnection mechanism, enhancing the flexibility and stability of the production line. The locally deployed YOLO framework-trained defect detection model efficiently and accurately identifies specific issues such as missing screws and crown color deviations. The confidence threshold can be adjusted as needed, further enhancing the reliability and adaptability of the detection.

[0020] This technical solution employs industry-standard protocols, such as Modbus or Profinet, to communicate with the production line controller, ensuring immediate response to control commands and enabling rapid execution of workpiece release or interception operations. Combined with an audible and visual alarm system, it can immediately alert workers to defective products, improving production efficiency and product quality.

[0021] The data storage unit includes not only a local database on the Mac control terminal but also extends to a remote PC server, providing powerful data synchronization and traceability capabilities. The introduction of the Apple Insight platform allows users to easily access historical detection data for data analysis and quality improvement. The model iteration module enables continuous system optimization; users can continuously update and improve the AI ​​model through the web annotation platform and model training toolchain, maintaining the detection system's advanced nature and accuracy.

[0022] In summary, this technical solution not only improves the efficiency and accuracy of quality inspection on the production line, but also ensures the long-term stable operation and continuous improvement capability of the system through flexible data management and model iteration mechanisms. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the overall system architecture of the present invention; Figure 2 This is a schematic diagram of the iPhone image acquisition terminal software architecture of the present invention; Figure 3 This is a schematic diagram of the Mac control terminal software architecture of the present invention; Figure 4This is a schematic diagram of the production line quality inspection method of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figure 1-4 The production line quality inspection system based on iPhone-Mac collaboration of the present invention is characterized by comprising an iPhone image acquisition terminal, a Mac control terminal, a production line controller and a data storage unit; The iPhone image acquisition terminal is connected to the Mac control terminal via a USB data cable. The Mac control terminal is electrically connected to the production line controller via an industrial protocol. The Mac control terminal is also connected to the data storage unit via a network. The iPhone image acquisition terminal is used to acquire original images of workpieces on the production line and report the equipment identification. The Mac control terminal is used for multi-terminal collaborative management, local AI defect identification and control command generation. The production line controller is used to receive commands and execute workpiece release or interception actions.

[0026] This technical solution centers on deep collaboration between iPhone data acquisition and Mac control, constructing a closed-loop production line quality inspection system that enables stable data acquisition, accurate analysis, real-time decision-making, and full-chain traceability. The core logic is as follows: As an industrial-grade front-end, the iPhone image acquisition terminal uses a high-performance camera to acquire lossless original images, avoids wireless interference through USB wired communication, and reports a unique identifier, such as a station code and camera number, to ensure data traceability. The Mac control terminal serves as the system's central hub, enabling collaborative management of multiple iPhone terminals, local AI defect inference to avoid remote delays, and outputting release / interception commands to the production line PLC controller based on the inference results. The data storage unit stores all test data according to a standardized catalog, supporting multi-dimensional traceability; The entire system achieves self-recovery from faults and self-evolution of detection capabilities through the anomaly handling module and model iteration module, completely solving the pain points of unstable transmission, insufficient accuracy and lag in response of traditional systems, and adapting to the quality inspection needs of high-speed production lines.

[0027] As the acquisition terminal that directly contacts the workpiece, the iPhone image acquisition terminal is responsible for acquiring non-destructive images, reporting device identity, and handling local anomalies, providing high-quality data input for backend analysis.

[0028] USB communication module: Wired full-duplex stable connection uses USB 3.0 or higher protocols to establish a physical connection with the Mac, unlike traditional Wi-Fi / Bluetooth wireless solutions. Full-duplex transmission features support simultaneous Mac command issuance and iPhone image upload, with a transmission latency of ≤100ms. The connection success rate was 100% during a 720-hour stress test, completely avoiding data packet loss caused by electromagnetic interference in industrial environments. During communication, the system automatically negotiates a transmission bandwidth of up to 5Gbps, adapting to iPhone RAW format images and meeting the high-speed transmission requirements of approximately 20-30MB per image, thus avoiding transmission lag caused by insufficient bandwidth.

[0029] Image acquisition module: Lossless raw image acquisition utilizes the iPhone's underlying AVFoundation framework, rather than a third-party camera SDK, to directly control the camera sensor. The image format is RAW or uncompressed bitmap, such as BMP, to preserve the original photoelectric signals of the sensor. For example, the 12-megapixel iPhone 13 can output a 4032×3024 resolution image without the loss of detail caused by JPEG compression. Traditional compression schemes will lose more than 30% of defective details. During data acquisition, the iPhone camera's focus lock and exposure lock functions are automatically enabled to ensure consistent image brightness and clarity across multiple batches of workpieces at the same location. This avoids affecting AI recognition accuracy due to imaging differences, and brightness fluctuations are controlled within ±5%.

[0030] Identity Management Module: The Unique Credential for Data Traceability Persistent storage mechanism: The station code, such as Station-05, and the camera number, such as Cam-03, are stored in the iPhone's local sandbox directory. This is not a temporary cache, so the data can still be retained even if the application restarts or the device loses power. Default values, such as Station-01 and Cam-01, are loaded on first use. Non-empty verification logic: When a user modifies the identifier through the graphical interface, the system automatically verifies that the combination of the station code and camera number is not empty, avoiding data anonymity caused by empty values. After the verification is passed, the system updates the storage and synchronously reports to the Mac, ensuring that each image can be traced back to the specific acquisition device and production line station.

[0031] The fault handling module, which integrates fault self-recovery and user guidance, is divided into two sub-modules: permission monitoring and connection monitoring. This enables robust operation with minimal human intervention. Permission monitoring submodule: Checks camera permissions upon startup. If denied, a pop-up window with text and images appears, including a button to go to settings, guiding the user to grant permissions. Traditional systems only prompt insufficient permissions without any guidance. Image acquisition automatically restarts after permissions are restored. Connection monitoring submodule: Real-time detection of USB physical connection status. If disconnected, the USB port is scanned once per second, and the interface displays a red fault light and a message "USB disconnected, reconnecting..." After successful reconnection, communication is automatically restored without manual plugging or unplugging of the device.

[0032] Mac Control Terminal: The Central Unit for Collaborative Management and Local Decision-Making Core Functions It coordinates the collaborative work of multiple iPhone terminals, completing local AI defect identification, production line instruction generation, data management, and model iteration; it is the system's brain.

[0033] Multi-terminal management module: Multi-device collaboration and status visualization adaptation; supports simultaneous connection of 2-8 iPhone terminals to meet the multi-view acquisition needs of the production line, such as the front, side, and top of the workpiece. Automatic discovery and authentication: After the Mac client starts, it scans all USB ports and identifies the iPhone terminal through preset communication protocols, such as a custom TCP / IP sub-protocol, to verify whether the terminal is running EagleEyeFront software and prevent unauthorized devices from accessing the device. Status visualization display: The main interface uses an array of indicator lights, with one light corresponding to each terminal, to display the status in real time. Green light: online and normal; red light: offline or faulty; yellow light: connecting. Operators can quickly locate abnormal devices. Automatic reconnection after disconnection: After the terminal goes offline, the module sends a connection request every 3 seconds until the reconnection is successful. At the same time, it records the disconnection time and reason, such as USB loosening or terminal power failure, to the system log for easy subsequent operation and maintenance analysis.

[0034] AI Inference Module: Local low-latency defect recognition is based on the YOLO framework, such as YOLOv8, to train a production line-specific defect detection model, which differs from traditional remote cloud inference. Advantages of local deployment: The inference process is completed locally on the MacMini without uploading to the cloud, and the response time is ≤2s. Traditional cloud inference takes 5-10s due to network latency. It is suitable for high-speed production lines with a cycle time of more than 30 pieces per minute. Defect identification range and threshold control: For electronic components, such as mobile phone optimization, it can identify minor defects such as missing screws (diameter ≤ 2mm), crown color deviation (ΔE ≤ 3), and poor battery installation (offset ≥ 0.5mm). The confidence threshold can be set to 80%-95%. For high-precision scenarios, set to 95% to reduce the false positive rate. For mass production scenarios, set to 80% to improve detection efficiency. Threshold adjustments take effect in real time.

[0035] Control command generation module: The execution bridge for production line linkage. Based on AI inference results, it generates standardized control commands that seamlessly integrate with the production line PLC controller. Protocol compatibility: Supports Modbus (serial / Ethernet) or Profinet (industrial Ethernet) protocols, compatible with mainstream PLCs such as Siemens S7 series and Mitsubishi FX series, with instruction transmission delay ≤500ms; Action logic: When the AI ​​determines "OK", it outputs a "release" command, and the PLC controls the production line conveyor belt to continue running; when it determines "NG", it outputs an intercept command, the PLC controls the stop device to rise and intercept the workpiece, and at the same time triggers the production line's audible and visual alarm, such as the LED light flashing and the buzzer sounding, to remind the operator to handle the defective product.

[0036] Data Management Module: Standardized storage and multi-dimensional traceability construct a hierarchical and categorized data storage system to meet quality traceability and compliance requirements. Local storage: Original images are stored in a directory structure of ReceivedImages, station code, camera number, and detection timestamp: YYYYMMDD-HHMMSS, such as "ReceivedImages / Station-05 / Cam-03 / 20240610-143025 / IMG_001.RAW", which facilitates quick location of individual images; Remote synchronization: Synchronize all detection data, such as image paths, AI inference results (defect type / confidence / coordinates), control commands, equipment identification, and detection time, to a remote PC server. It supports searching by time range, station, and defect type through the Apple Insight platform, such as querying all records of "20240610Station-05 missing screw".

[0037] Model Iteration Module: The core capability of the system's self-evolution addresses the shortcomings of traditional systems, such as long model iteration cycles and reliance on professional personnel, enabling detection capabilities to evolve with the production line. Web annotation platform: Built-in lightweight web annotation tools allow operators to select defective areas in images without the need for professional software, such as using a rectangle to select missing screws, adding labels such as "missing screws - location TOP", and generating standard datasets in PASCALVOC or COCO format; Localized training and deployment: The model training toolchain supports visual configuration of hyperparameters, with learning rates of 0.001-0.01, training epochs of 50-200, and batch sizes of 8-32. The training process is completed locally on the Mac or submitted to the associated server. After training, a model performance report, such as accuracy and recall, is automatically generated. Qualified models can be updated to the AI ​​inference module via a one-click deployment button, shortening the iteration cycle from the traditional several weeks to within 3 days, and quickly adapting to new defects, such as the new screen scratch detection requirement on the production line.

[0038] Production line controller and data storage unit: Supporting units for execution and traceability Production line controller (PLC): The hardware terminal for instruction execution. As a bridge between the Mac terminal and the production line equipment, it receives digital instructions (OK / NG) from the Mac terminal and converts them into analog signals to control the production line actuators, such as conveyor belt motors and stop cylinders. When an interception command is received, in addition to controlling the stop, information on defective products, such as time, location, and defect type, is recorded simultaneously and sent to the production line's MES system to achieve data linkage between quality inspection and production.

[0039] Data storage unit: Full lifecycle data management Local database (Mac): Uses SQLite or PostgreSQL to store frequently accessed image paths and detection results, ensuring fast response to local queries; Remote PC server: Adopts a distributed storage architecture to store massive amounts of historical data, such as one year's worth of testing records, and supports data backup and disaster recovery, such as daily full backups to avoid data loss; AppleInsight platform: As a visual traceability portal, it allows managers to view production line quality inspection reports, such as pass rate trends and defect distribution statistics, providing data support for quality improvement.

[0040] Detailed Workflow Step 1: System Initialization Hardware connection: Connect 2-8 iPhone devices to the Mac control terminal via USB data cable, ensuring that each iPhone corresponds to a unique USB port to avoid confusion; Software Launch and Authentication: Launch the EagleEyeMacross software on Mac and the EagleEyeFront software on iPhone. The Mac will automatically scan the USB port, identify and authenticate the iPhone terminal, and verify the software version and communication protocol compatibility. Identity reporting and status synchronization: The iPhone terminal loads the location code and camera number from the local device and actively reports them to the Mac terminal; after the Mac terminal verifies that the identity is not empty, it updates the indicator light on the main interface to green, completing the preparation for multi-terminal collaboration; Preset parameters: Set the AI ​​inference confidence threshold (e.g., 90%), data storage path, and PLC communication protocol (e.g., ModbusTCP) through the Mac interface. The system will automatically save the parameters to the local configuration file.

[0041] Step 2: Trigger Detection Signal triggering: The photoelectric sensor on the production line detects that the workpiece has arrived at the designated station and sends an equipment arrival signal to the PLC controller; Command forwarding: The PLC controller forwards the arrival signals of the device to the Mac control terminal through industrial protocols such as Profinet; Synchronous photo capture command issuance: The Mac-side multi-terminal management module broadcasts a synchronous photo capture command to all online iPhone terminals, including shooting parameters such as focus mode: automatic and exposure time: 1 / 100s, to ensure simultaneous acquisition from multiple perspectives and avoid image misalignment caused by time differences in acquisition.

[0042] Step 3: Image Acquisition and Transmission Non-destructive acquisition: After receiving the photo capture command on the iPhone, the image acquisition module calls the underlying camera API to capture the workpiece image according to preset parameters. The image is in RAW format and the resolution matches the highest hardware pixel of the camera. Data encapsulation: The identity management module encapsulates the station code, camera number, and shooting timestamp into image metadata, and packages it together with the original image into an image data packet; USB lossless transmission: The USB communication module transmits image data packets to the Mac via a wired link. CRC32 verification is enabled during transmission to ensure data integrity. If the verification fails, the data will be automatically retransmitted, with a maximum of 3 retransmissions.

[0043] Step 4: AI Defect Identification and Result Judgment Data parsing and storage: The Mac-side data management module receives image data packets, parses metadata and original images, stores images according to a standardized directory, and simultaneously writes metadata to a local database; Local AI Inference: The AI ​​inference module reads the stored original image, inputs it into the local YOLO model for defect detection, and outputs the inference results, such as defect type: missing screw, confidence level: 96%, location coordinates (x1, y1, x2, y2): (120, 150, 180, 210); Result determination: The controller compares the inference confidence with a preset threshold, such as 90%. If the confidence is greater than or equal to the threshold, it is determined as "qualified (OK)"; otherwise, it is determined as "unqualified (NG)". If multiple perspectives are collected, such as 3 iPhones taking pictures of the same workpiece, the majority rule is adopted. For example, if 2 iPhones determine OK and 1 iPhone determines NG, the final result is OK.

[0044] Step 5: Control Feedback and Data Traceability Command output: Based on the judgment result, the control command generation module sends a "allow" or "intercept" command to the PLC controller through a preset industrial protocol; Production line execution: After receiving the instruction, the PLC controller controls the production line to execute the action within 1 second. OK: the conveyor belt continues to run; NG: the workpiece is stopped and an audible and visual alarm is triggered. Data synchronization: The data management module will synchronize all data from this test, including image paths, AI results, command records, device identifiers, and timestamps, to a remote PC server and the Apple Insight platform to complete data traceability and recording.

[0045] Step 6: Exception Handling and Model Iteration Real-time exception handling: If the iPhone connection monitoring module detects a USB disconnection, it immediately initiates automatic reconnection, scans the port once per second, the corresponding terminal indicator light on the Mac turns red, and logs the USB disconnection, time, and port number. If the iPhone permission monitoring submodule detects that camera permission is denied, a guidance pop-up window will appear, and data collection will automatically restart after the user grants permission. If AI inference on the Mac fails, such as due to image corruption, a pause command is immediately sent to the PLC, and an alarm is triggered, such as a flashing yellow light on the Mac, awaiting manual investigation.

[0046] Periodic model iteration: Each week, quality inspectors use a Mac-based web annotation platform to annotate images of defective products for the week, generating a dataset. Call the model training toolchain, configure the hyperparameters, start local training, and generate a new model; Verify the performance of the new model, such as improving the accuracy from 98.5% to 99.2%. Once it passes the verification, deploy it to the AI ​​inference module with one click to complete the upgrade of the system's detection capabilities.

[0047] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A production line quality inspection system based on iPhone-Mac collaboration, characterized in that, Includes iPhone image acquisition terminal, Mac control terminal, production line controller and data storage unit; The iPhone image acquisition terminal is connected to the Mac control terminal via a USB data cable. The Mac control terminal is electrically connected to the production line controller via an industrial protocol. The Mac control terminal is also connected to the data storage unit via a network. The iPhone image acquisition terminal is used to acquire original images of workpieces on the production line and report the equipment identification. The Mac control terminal is used for multi-terminal collaborative management, local AI defect identification and control command generation. The production line controller is used to receive commands and execute workpiece release or interception actions.

2. The production line quality inspection system based on iPhone-Mac collaboration according to claim 1, characterized in that, The iPhone image acquisition terminal runs EagleEyeFront software, which includes a USB communication module, an image acquisition module, an identity management module, and an exception handling module. The USB communication module is used to establish a full-duplex wired connection with the Mac control terminal. The image acquisition module calls the iPhone's underlying camera API to obtain lossless original images. The identity management module stores and reports the station code and camera number. The exception handling module monitors the USB connection status and camera permissions and triggers automatic reconnection or authorization guidance.

3. The production line quality inspection system based on iPhone-Mac collaboration according to claim 2, characterized in that, The image acquisition module acquires raw images in RAW or uncompressed bitmap format, with a pixel resolution no lower than the highest hardware pixel count of the iPhone camera. The identity management module persistently stores the station code and camera number on the iPhone. The default value is loaded when it is used for the first time, and users can modify it through the graphical interface and perform non-empty verification.

4. The production line quality inspection system based on iPhone-Mac collaboration according to claim 3, characterized in that, The Mac control terminal runs EagleEyeMacross software, including a multi-terminal management module, an AI inference module, a control command generation module, and a data management module. The multi-terminal management module automatically scans and authenticates USB-connected iPhone terminals, displaying online / offline status using a status indicator array. The AI ​​inference module deploys a defect detection model trained on the local YOLO framework. The control command generation module outputs release / interception commands based on the AI ​​results. The data management module stores images and detection data according to a standardized directory.

5. The production line quality inspection system based on iPhone-Mac collaboration according to claim 4, characterized in that, The multi-terminal management module supports simultaneous connection of at least 2-8 iPhone terminals and has a built-in automatic reconnection mechanism after disconnection; the defect identification range of the AI ​​inference module includes missing screws, crown color deviation, and improper battery installation, and the inference confidence threshold can be set to 80%-95%.

6. The production line quality inspection system based on iPhone-Mac collaboration according to claim 5, characterized in that, The production line controller is a PLC controller that communicates with the Mac control terminal via Modbus or Profinet protocol. When the Mac control terminal outputs a qualified instruction, the production line controller controls the production line to release the workpiece; when it outputs a non-qualified instruction, it controls the production line to intercept the workpiece and triggers an audible and visual alarm.

7. The production line quality inspection system based on iPhone-Mac collaboration according to claim 6, characterized in that, The data storage unit includes a local database on the Mac control terminal and a remote PC server. The local database stores the original images in a directory structure of ReceivedImages, station code, camera number, and detection timestamp. The remote PC server synchronously stores the full detection data, including image data, AI inference results, control commands, and device identification, and supports access and traceability through the Apple Insight platform.

8. The production line quality inspection system based on iPhone-Mac collaboration according to claim 6, characterized in that, The exception handling module includes a permission monitoring submodule and a connection monitoring submodule; When the permission monitoring submodule detects that camera permission has been denied, a graphical pop-up window will appear to guide the user to the system settings for authorization. When the connection monitoring submodule detects a USB disconnection, it continuously scans the USB port and attempts to reconnect automatically, while displaying a red fault indicator light and a connection failure message on the interface.

9. The production line quality inspection system based on iPhone-Mac collaboration according to claim 6, characterized in that, The Mac control terminal also includes a model iteration module, which has a built-in Web annotation platform and model training toolchain. The web annotation platform allows users to select image defects and add labels to generate standard datasets. The model training toolchain supports configuring hyperparameters such as learning rate and training epochs. Newly trained models can be deployed to the AI ​​inference module with one click.

10. The production line quality inspection method based on iPhone-Mac collaboration applied to any of the production line quality inspection systems based on iPhone-Mac collaboration according to claims 1-9, characterized in that, Includes the following steps: System initialization: The Mac control terminal scans the USB-connected iPhone terminal, completes identity authentication and connection establishment, and the iPhone terminal loads and reports the station location code and camera number; Trigger detection: The production line controller sends a device arrival signal to the Mac control terminal, and the Mac control terminal broadcasts a synchronous photo-taking command to all online iPhone terminals; Image acquisition and transmission: The iPhone terminal uses the camera to acquire raw images and transmits them to the Mac control terminal without loss via USB data cable; AI Defect Recognition: The Mac control terminal calls the local AI model to analyze the image and determines whether the workpiece is qualified or unqualified based on the confidence threshold; Control feedback: The Mac control terminal outputs release / interception commands to the production line controller and simultaneously synchronizes the detection data to the data storage unit; 6) Anomaly Handling: If a USB disconnection or lack of permissions is detected, the iPhone terminal will trigger automatic reconnection or authorization boot, and the Mac control terminal will record the fault log.

Images