A Machine Vision-Based Full-Function Automated Testing Method and Device for Shipborne Navigation Data Recorders

Abstract

This invention relates to the field of machine vision, and more particularly to a machine vision-based automated testing method and apparatus for the full functionality of a shipborne navigation data recorder. A PLC control unit responds to instructions from a host computer, controlling the signal simulation acquisition module to output multi-source test signals to the device under test (DUT), while simultaneously triggering the vision detection module to synchronously acquire real-time images of the display interface and indicator lights. The PLC receives the electrical signal response data returned by the signal simulation acquisition module and the visual recognition results output by the vision detection module, performs correlation analysis between the two, and if the results do not meet a preset threshold, an abnormal control signal is output and uploaded to the record; if they meet the threshold, the data is summarized to generate a test report. This invention achieves non-contact automated testing of the full functionality of a shipborne navigation data recorder through synchronous linkage of signal output and image acquisition, and dual verification of electrical signals and visual results, improving testing efficiency and judgment accuracy.

Description

Technical Field

[0001] This application relates to the field of machine vision technology, and in particular to a fully automated testing method and testing device for shipborne navigation data recorders based on machine vision. Background Technology

[0002] The shipborne Voyage Data Recorder (VDR) is a core device for ensuring ship navigation safety and accident tracing. Its functions encompass the collection and storage of multiple information sources, including heading, speed, radar imagery, voice recording, and alarms, requiring high accuracy and consistency in testing. Currently, equipment testing primarily relies on manual operation, involving multiple dimensions such as display response, button actions, audio-visual indications, and data interfaces. This traditional method suffers from strong subjectivity in judgment, low testing efficiency, and poor repeatability and consistency. Furthermore, it lacks non-contact testing methods, making it difficult to meet the demands of mass production and standardized testing. While some automated testing equipment has been attempted for ship electronic equipment testing, it primarily performs single-function testing, failing to achieve multi-dimensional, full-function testing, and lacks effective system linkage mechanisms. The testing response speed and recognition accuracy need improvement.

[0003] In terms of visual inspection, some existing technologies have attempted to use image acquisition and recognition techniques to detect the display status of shipboard electronic equipment. However, problems such as insufficient recognition accuracy and poor environmental adaptability still exist. Especially under conditions of complex lighting and reflective display interfaces, the stability of traditional recognition methods is difficult to guarantee, and the judgment mechanism is relatively simple, lacking effective means to handle boundary states. This makes it difficult to fully meet the recognition accuracy requirements of shipborne navigation data recorders in the entire process of inspection. Summary of the Invention

[0004] To address one or more problems in the prior art, the main objective of this application is to provide a fully automated testing method and device for shipborne navigation data recorders based on machine vision.

[0005] To achieve the aforementioned objectives, this application proposes a fully automated testing method for shipborne navigation data recorders based on machine vision, the method comprising:

[0006] In response to the test start command issued by the host computer, the PLC control unit controls the signal analog acquisition module to output multi-source test signals to the shipborne navigation data recorder under test;

[0007] While the signal simulation acquisition module outputs the multi-source test signal, the visual detection module is triggered to acquire images of the shipborne navigation data recorder under test, obtaining real-time images including the display interface and / or indicator light status.

[0008] The system receives electrical signal response data transmitted back by the signal simulation acquisition module, which is acquired by the signal simulation acquisition module from the shipborne navigation data recorder under test.

[0009] Receive the visual recognition result output by the visual detection module;

[0010] The electrical signal response data is correlated with the visual recognition result. If the analysis result does not meet the preset test threshold, an abnormal control signal is output and uploaded to the host computer for recording.

[0011] If the analysis results meet the preset test threshold, the host computer will summarize the test data and generate a test report.

[0012] This application also provides a fully automated testing device for shipborne navigation data recorders based on machine vision, including:

[0013] The response module is used to respond to the test start command issued by the host computer. The PLC control unit controls the signal analog acquisition module to output multi-source test signals to the shipborne navigation data recorder under test.

[0014] The acquisition module is used to trigger the visual detection module to acquire images of the shipborne navigation data recorder under test while the signal simulation acquisition module outputs the multi-source test signal, and to acquire real-time images including the display interface and / or indicator light status.

[0015] The first receiving module is used to receive the electrical signal response data transmitted back by the signal simulation acquisition module, wherein the electrical signal response data is acquired by the signal simulation acquisition module from the shipborne navigation data recorder under test.

[0016] The second receiving module is used to receive the visual recognition result output by the visual detection module;

[0017] The correlation analysis module is used to perform correlation analysis between the electrical signal response data and the visual recognition result. If the analysis result does not meet the preset test threshold, an abnormal control signal is output and uploaded to the host computer for recording.

[0018] The generation module is used to summarize the test data and generate a test report through the host computer if the analysis results meet the preset test threshold.

[0019] This application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the methods described above.

[0020] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0021] This application presents a machine vision-based fully automated testing method and apparatus for shipborne navigation data recorders. Through a hardware-level synchronous triggering mechanism in the PLC control unit, it achieves precise timing matching between signal output and image acquisition, resolving the time lag between signal input and state observation in traditional testing. This ensures that each test corresponds to the device's actual response at a specific moment. By simultaneously acquiring electrical signal response data and visual recognition results, this method constructs a dual verification channel between internal processing logic and the external human-machine interface. This verifies both the normal operation of the core functional modules of the device and the correctness of the user-observable display information, compensating for potential blind spots in a single detection path. The PLC control unit performs correlation analysis and consistency comparison on the two types of data, transforming the judgment process, which originally relied on human experience, into an objective data comparison. This eliminates the interference of human factors on the test results, improving the reliability and repeatability of the detection. The entire process requires no manual intervention, significantly shortening the testing cycle and meeting the standardized testing requirements of mass production scenarios. Compared with traditional manual testing methods, this method has achieved a qualitative improvement in testing efficiency, judgment accuracy, and result consistency, providing a reliable technical solution for the full-function automated testing of shipborne navigation data recorders. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating a fully automated testing method for a shipborne navigation data recorder based on machine vision, according to an embodiment of this application.

[0023] Figure 2 This is a flowchart illustrating a fully automated testing method for a shipborne navigation data recorder based on machine vision, according to an embodiment of this application.

[0024] Figure 3 This is a schematic block diagram of a fully automated testing device for a shipborne navigation data recorder based on machine vision, according to an embodiment of this application.

[0025] Figure 4 This is a schematic block diagram of the structure of a computer device according to an embodiment of this application;

[0026] Figure 5 This is a schematic diagram of the prior art framework of a fully automated testing method for a shipborne navigation data recorder based on machine vision, according to an embodiment of this application.

[0027] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0029] Reference Figure 1 This application provides a fully automated testing method for shipborne navigation data recorders based on machine vision, the method comprising:

[0030] S1. In response to the test start command issued by the host computer, the PLC control unit controls the signal analog acquisition module to output multi-source test signals to the shipborne navigation data recorder under test.

[0031] S2. While the signal simulation acquisition module outputs the multi-source test signal, the visual detection module is triggered to acquire images of the shipborne navigation data recorder under test, and obtain real-time images including the display interface and / or indicator light status.

[0032] S3. Receive the electrical signal response data transmitted back by the signal simulation acquisition module, wherein the electrical signal response data is acquired by the signal simulation acquisition module from the shipborne navigation data recorder under test.

[0033] S4. Receive the visual recognition result output by the visual detection module;

[0034] S5. The electrical signal response data and the visual recognition result are correlated and analyzed. If the analysis result does not meet the preset test threshold, an abnormal control signal is output and uploaded to the host computer for recording.

[0035] S6. If the analysis results meet the preset test threshold, the host computer will summarize the test data and generate a test report.

[0036] As described in steps S1-S3 above, the PLC control unit uniformly schedules each test module, achieving fully automated closed-loop control from signal excitation, image acquisition, data retrieval to result determination. Step 1: When the test begins, the host computer sends a test start command to the PLC control unit. Upon receiving this command, the PLC control unit, based on preset test parameters, controls the signal simulation acquisition module to output multi-source test signals to the shipborne navigation data recorder under test. These signals simulate various data generated during actual ship navigation, such as changes in heading angle, ship speed, and radar image signals. The signal simulation acquisition module transmits these signals to the device under test through a standard interface, putting it into simulated operation mode, thus providing an excitation source for subsequent functional verification. This step establishes the initial test state, ensuring the device under test receives multi-dimensional signal inputs similar to the real navigation environment, laying the foundation for subsequent response acquisition and state determination. Step 2: Simultaneously with the signal simulation acquisition module outputting multi-source test signals, the PLC control unit drives the vision inspection module to acquire images via hardware triggering. This synchronization mechanism is crucial for ensuring accurate test timing. The vision inspection module uses a pre-installed industrial camera to capture real-time images of the display interface and indicator light areas of the device under test (DUT), obtaining real-time images including changes in displayed content and the on / off status of indicator lights. For example, when the signal simulation acquisition module outputs a heading change signal, the PLC simultaneously triggers the camera to capture the change in heading value on the display screen. This hardware-level synchronization can control the time difference between signal output and image acquisition within a very small range, ensuring that the captured image corresponds exactly to the currently output test signal, avoiding misjudgments caused by timing deviations. Step three: In addition to outputting test signals, the signal simulation acquisition module also undertakes the task of acquiring the response data of the DUT. After receiving multi-source test signals, the DUT will generate corresponding electrical signal responses through its internal circuitry, such as data interface output confirmation information or changes in status register values. The signal simulation acquisition module collects these electrical signal response data and sends them back to the PLC control unit. These electrical signal response data reflect the processing results of the DUT at the internal circuit level and are an objective record of whether the core functions of the device are working properly. For example, when a speed signal is output, the device should generate a corresponding data processing flow and output the corresponding value at a specific interface. The change of this value can be accurately obtained through electrical signal response data.

[0037] As described in steps S4-S6 above, after the visual inspection module completes image acquisition in step four, it performs recognition processing on the acquired real-time images. This recognition process includes extracting information such as values, characters, and icons from the display interface from the image, as well as judging the on / off state and color changes of indicator lights. The visual inspection module outputs the results obtained after recognition, i.e., the visual recognition results, to the PLC control unit. Unlike electrical signal response data, visual recognition results reflect the performance of the device under test at the human-machine interaction level, i.e., the interface information that users or inspectors can directly observe. For example, whether the heading value is correctly displayed on the device's screen, whether the alarm indicator lights are lit as expected, etc. This information is a true record of the device's external performance. In step five, after the PLC control unit receives both the electrical signal response data and the visual recognition results, it performs correlation analysis on the two. The core of this analysis is to compare whether the internal processing results of the device under test are consistent with the external display results. Specifically, the PLC parses the electrical signal response data into a first state parameter representing the internal state of the device, and parses the visual recognition results into a second state parameter representing the external display state of the device, and then compares the consistency of these two state parameters. If both are consistent, it indicates that the internal processing of the equipment is correct and the external display is normal, and the test item is judged to have passed. If they are inconsistent, it indicates that there is an abnormality in the equipment, which may be due to a fault in the internal signal processing module causing an output error, or it may be due to an abnormality in the display driver circuit or the display screen itself. Once an abnormality is determined, the PLC control unit immediately outputs an abnormality control signal, uploads the abnormality information to the host computer for storage and display, and takes corresponding measures according to the type of abnormality, such as switching the test channel or triggering an alarm. This correlation analysis mechanism is one of the core innovations of this method. It performs dual verification of the internal state and external performance of the equipment, effectively avoiding missed detections or misjudgments that may occur with a single detection path. Step Six: If the analysis results of all test items meet the preset test thresholds, that is, all functions are normal, the PLC control unit uploads all test data to the host computer. The host computer summarizes, statistically analyzes, and finally generates a complete test report. The test report includes test parameters, response data, judgment results, and timestamps for each test item, which can be used for subsequent quality traceability and data analysis.

[0038] refer to Figure 5In existing technologies, the testing of shipborne navigation data recorders mainly relies on manual methods. Operators visually observe the display screen and indicator light status of the device under test, while manually operating the buttons on the device to trigger different functions, and manually recording the observed results in a test table. This testing method depends entirely on the operator's visual judgment and manual operation. Verification of each function requires manual completion one by one, resulting in a long testing cycle. Furthermore, the accuracy of the test results is greatly affected by subjective factors such as the operator's experience and fatigue level. As mentioned above, this invention achieves precise timing matching between signal output and image acquisition through a hardware-level synchronous triggering mechanism of the PLC control unit, solving the problem of time difference between signal input and status observation in traditional testing, and ensuring that each test corresponds to the actual response of the device at a specific moment. By simultaneously acquiring electrical signal response data and visual recognition results, this method constructs a dual verification channel of internal processing logic and external human-machine interface, verifying both whether the core functional modules of the device are working properly and whether the user-observable display information is correct, thus compensating for the blind spots that may exist in a single testing path. The PLC control unit performs correlation analysis and consistency comparison on the two types of data, transforming the judgment process, which originally relied on human experience, into an objective data comparison. This eliminates the interference of human factors on the test results and improves the reliability and repeatability of the inspection. The entire process requires no manual intervention, significantly shortening the testing cycle and meeting the standardized testing requirements of mass production scenarios. Compared with traditional manual testing methods, this method achieves a qualitative improvement in testing efficiency, judgment accuracy, and result consistency, providing a reliable technical solution for the fully automated testing of shipborne navigation data recorders.

[0039] In one embodiment, after the step of correlation analysis between the electrical signal response data and the visual recognition results, the method further includes:

[0040] The electrical signal response data is parsed into a first state parameter characterizing the internal state of the shipborne navigation data recorder under test.

[0041] The visual recognition result is interpreted as a second state parameter that characterizes the display status of the external human-machine interface of the shipborne navigation data recorder under test.

[0042] Perform a consistency comparison between the first state parameter and the second state parameter;

[0043] If the first state parameter is inconsistent with the second state parameter, the analysis result is determined to be inconsistent with the preset test threshold.

[0044] The output of the abnormal control signal and its uploading to the host computer for recording includes:

[0045] The PLC control unit generates channel switching instructions and alarm trigger instructions based on the inconsistent comparison results.

[0046] In response to the channel switching command, the drive relay switches to the backup test channel to continue executing the unfinished test items;

[0047] In response to the alarm trigger command, the alarm device is driven to output an audible, visual, and / or electrical signal alarm;

[0048] The inconsistent comparison results and corresponding abnormal control signals are uploaded to the host computer for storage and display.

[0049] As described above, this embodiment constructs a consistency verification mechanism between internal state and external display, and designs differentiated processing procedures for abnormal situations. Specifically, when the PLC control unit receives electrical signal response data and visual recognition results, it first parses and processes these two types of data. The electrical signal response data reflects the processing results at the internal circuit level of the device under test, such as the values ​​output by the data interface, changes in the flag bits of the status register, etc. The PLC parses it into a first state parameter representing the internal state of the device. The visual recognition results reflect the actual content presented on the device's display interface, such as the heading value displayed on the screen, the on / off state of the alarm indicator, etc. The PLC parses it into a second state parameter representing the display state of the device's external human-machine interface. Through this parsing, the originally complex raw data is transformed into a state description that can be directly compared. Subsequently, the PLC control unit performs a consistency comparison between the first state parameter and the second state parameter. The core logic of this comparison is to verify whether the internal processing results of the device match the external display results. For example, when the signal simulation acquisition module outputs a speed signal, the device should generate a corresponding speed value and output it through the data interface, and the display screen should also display the speed value. If the internal output value matches the value displayed on the screen, it indicates that the equipment is processing correctly and displaying normally; if they do not match, it indicates that the equipment is malfunctioning. This inconsistency may stem from a fault in the internal signal processing module causing an output error, or from an abnormality in the display driver circuit or the display screen itself. When the comparison results are inconsistent, the PLC control unit determines that the test item does not meet the preset test threshold and takes corresponding measures based on the abnormality. Specifically, this includes three levels: First, generating a channel switching command to drive the relay to switch the test channel to a backup channel, allowing the current abnormal item to be skipped and subsequent unfinished test items to be executed, avoiding interruption of the entire test process due to a single abnormal test item. Second, generating an alarm trigger command to drive the alarm device to output an audible and visual alarm or an electrical signal alarm, promptly alerting the operator to the abnormality. Third, uploading the inconsistent comparison results and corresponding abnormal control signals to the host computer for storage and display, facilitating subsequent quality traceability and fault analysis.

[0050] It's worth noting that traditional automated testing solutions often simply record and continue execution when an anomaly is detected, or directly interrupt the test, lacking the ability to handle different anomaly types differently. This method introduces a channel switching mechanism, ensuring that non-critical anomalies do not block the entire testing process, effectively improving testing efficiency; it introduces an alarm mechanism, ensuring that critical anomalies are promptly perceived by operators; and it uploads complete anomaly information, providing data support for subsequent fault location and quality analysis. More importantly, this method uses the consistency comparison between internal state and external display as the core basis for anomaly judgment. This design directly addresses the interlocking illusion problem existing in current technologies. In manual testing or single signal detection scenarios, faults such as internal processing errors but normal display, or display freezes but internal normal operation, are difficult to detect because testers can only observe the external display and cannot know the internal state. However, this method uses double verification, only determining normal operation when the internal state is consistent with the external display. Anomalies on either side will be accurately captured, effectively avoiding missed detections caused by false consistency. This internal and external linkage detection approach fundamentally guarantees the reliability of test results.

[0051] In one embodiment, the visual recognition result output by the visual detection module includes the following steps:

[0052] The real-time image is preprocessed to obtain a preprocessed enhanced image;

[0053] Feature extraction is performed on the preprocessed enhanced image to obtain template matching similarity S1 and convolution feature similarity S2;

[0054] A weighted similarity determination function is used to fuse the template matching similarity S1 and the convolutional feature similarity S2 to obtain a comprehensive feature.

[0055] Based on the comprehensive features, the comprehensive similarity S is calculated, where the calculation formula is: S = 0.6 × S1 + 0.4 × S2;

[0056] State classification is performed based on the comprehensive similarity S, including:

[0057] If S≥0.8, then the status is considered normal;

[0058] If 0.6 ≤ S < 0.8, then proceed to the review and judgment channel;

[0059] If S < 0.6, the state is determined to be abnormal;

[0060] Based on state judgment classification, the visual recognition result is determined and output.

[0061] As described above, after the visual inspection module acquires real-time images of the device under test, it first performs image preprocessing. This stage includes converting the color image to grayscale to reduce the data volume, removing noise interference introduced during image acquisition using Gaussian filtering, separating the target area from the background using adaptive threshold segmentation, and finally extracting the clear edges of button outlines, indicator light boundaries, and displayed characters using an edge detection algorithm. After this series of preprocessing operations, the original image is transformed into an enhanced image with more prominent features. Subsequently, the visual inspection module performs feature extraction on the preprocessed enhanced image. Feature extraction adopts a dual-path parallel approach: on the one hand, the target area in the image is compared with a preset standard template using a template matching algorithm to calculate the pixel-level similarity and obtain the template matching similarity S1. The advantage of this path is that it can quickly capture image areas that are highly consistent with the standard template and has a stable recognition capability for display content under normal conditions. On the other hand, a lightweight convolutional neural network is used to extract deep features from the image. This network employs a three-layer convolutional structure and can learn local illumination-invariant features. Even when there are slight reflections, brightness changes, or font differences on the display interface, it can still extract features that reflect the essential structure of the characters, obtaining a convolutional feature similarity S2. After obtaining the two similarity indices, the visual detection module uses a weighted similarity judgment function to fuse the features of the two to obtain a comprehensive feature. During the fusion process, the weight of template matching similarity S1 is set to 0.6, and the weight of convolutional feature similarity S2 is set to 0.4. This weight allocation reflects the complementary relationship between the two feature extraction methods: template matching results are stable and reliable, and perform well in standardized testing environments, so it is given a higher weight; convolutional features have stronger adaptability under complex lighting conditions, but the computational cost is relatively large, so a moderate weight is given. Through weighted fusion, the stability of traditional methods is preserved while introducing the robustness of deep learning. Based on the comprehensive feature, the visual detection module calculates the comprehensive similarity S, with the formula S equal to 0.6 multiplied by S1 plus 0.4 multiplied by S2. This comprehensive similarity measure quantifies the degree of matching between the current image and the expected state. Based on different ranges of the S-value, the system performs a three-level classification judgment: when the S-value is greater than or equal to 0.8, the judgment state is normal, indicating that the image highly matches the expectation, and the recognition result can be directly output; when the S-value is between 0.6 and 0.8, the judgment enters the verification judgment channel, indicating that the image is in a state of blurred boundaries and requires further confirmation; when the S-value is less than 0.6, the judgment state is abnormal, indicating that the image differs significantly from the expectation, and an abnormal result is directly output. Finally, the visual detection module determines and outputs the final visual recognition result based on the state judgment classification result. This method introduces a three-level classification judgment mechanism and sets up a verification judgment channel.In traditional methods, recognition results are often binary classifications of "normal" or "abnormal." For images in boundary states, the system either reluctantly classifies them as normal, leading to missed detections, or arbitrarily classifies them as abnormal, resulting in false alarms. This method's verification channel design provides a buffer zone for boundary states. When the recognition result is uncertain, the system can enter a verification process, obtaining clearer images by adjusting the light source, changing the shooting angle, or taking multiple samples, thus making a more accurate judgment. This solves the problem of false judgments caused by recognition uncertainty in existing technologies, significantly improving the reliability and fault tolerance of visual inspection.

[0062] In one embodiment, after the step of determining if the analysis result does not meet the preset test threshold, the method further includes:

[0063] If the anomaly is a skippable, non-related test item anomaly, a channel switching command is generated to drive the relay to switch to the backup test channel, skip the current test item and continue to execute the next test item;

[0064] If the anomaly is a related test item anomaly or a system-level anomaly that affects subsequent tests, an alarm trigger command is generated to drive the alarm device to output an audible and visual alarm, and the test process is paused to await manual intervention.

[0065] As mentioned above, when the PLC control unit determines that the analysis results do not meet the preset test threshold, the system first identifies and classifies the anomaly type. The core basis for this classification is the degree of correlation between the abnormal test item and other test items. Specifically, in the preset test sequence, different test items have state dependencies. Some test items are independent of each other; for example, there is no data or state dependency between the button response test and the voice recording / playback test, and an anomaly in one item will not affect the normal execution of other items. Other test items, however, are dependent; for example, a failure in the system initialization test will cause all subsequent functional tests to lose their baseline state, and an anomaly in the power stability test may affect the overall operating state of the equipment. Once such a test item is abnormal, subsequent tests cannot be performed in the correct equipment state. Based on this classification, differentiated processing logic is executed. For skippable, non-correlated test item anomalies, the PLC control unit generates a channel switching instruction, drives a relay to switch the test channel to a backup test channel, skips the current abnormal item, and continues to execute the next test item. The advantage of this approach is that an anomaly in a single test item will not block the entire test process; it can automatically bypass the fault point and continue to complete the testing of other functions. For example, in mass production testing, if a non-core button on the equipment malfunctions, tests on other critical functions such as heading, speed, and radar imaging can still be completed, thus obtaining the most complete test data possible. For anomalies in related test items or system-level anomalies that affect subsequent tests, the PLC control unit takes a more cautious approach. It generates alarm trigger commands, drives the alarm device to output audible and visual alarms, and simultaneously suspends the test process, awaiting manual intervention. These anomalies often indicate problems with the equipment's basic functions or system environment. Continuing to execute subsequent tests not only fails to obtain valid data but may also lead to misjudgments due to the abnormal equipment status or even further damage to the equipment. For example, if the equipment's system clock synchronization test malfunctions, all subsequent time-related functional tests will lose their reference point. In this case, suspending the test and having manual intervention for troubleshooting is a more reasonable choice. This method provides a practical implementation path for anomaly skipping through the hardware design of a backup test channel. Traditional software-level skipping methods often only record anomalies but cannot truly bypass the anomaly point because subsequent tests are still executed on the same hardware channel, and the residual state of the faulty equipment may continue to affect subsequent results. This method uses a relay to physically switch the test channel to a backup channel, thereby resetting and isolating the test environment and eliminating the interference of abnormal conditions on subsequent tests.

[0066] Reference Figure 2 In one embodiment, the method further includes: periodically performing synchronization calibration in the test sequence, the synchronization calibration step including:

[0067] S71. Output a calibration electrical signal to the signal analog acquisition module and record the time when the calibration electrical signal is emitted;

[0068] S72. While outputting the calibration electrical signal, trigger the visual detection module to perform image acquisition;

[0069] S73. The receiving time of the calibration electrical signal returned by the receiving signal analog acquisition module, and the time of image feature change corresponding to the calibration electrical signal returned by the visual detection module.

[0070] S74. Calculate the time delay difference between the signal transmission path and the visual acquisition path based on the transmission time, reception time, and the time of image feature change;

[0071] S75. If the delay difference exceeds a preset threshold, a compensation strategy is generated based on the delay difference.

[0072] As described above, in the automated testing process, the PLC control unit synchronizes signal output and image acquisition through hardware triggering, which is fundamental to ensuring the accuracy of the test timing. However, in actual operation, due to the inherent characteristics of the PLC scan cycle, the slight jitter of the hardware trigger signal, and the differences in transmission paths between different modules, the relative delay between signal output and image acquisition is not constant. As the test sequence lengthens, this slight timing deviation gradually accumulates. When the accumulated delay exceeds the design tolerance range, the originally synchronously triggered signal output and image acquisition will become misaligned, causing the acquired image to correspond to the signal response of the previous test item rather than the state of the current test item, thus leading to misjudgment. To solve this problem, this method periodically inserts a synchronous calibration step into the test sequence. The calibration process is executed as follows: First, the PLC control unit outputs a well-defined calibration electrical signal to the signal simulation acquisition module and records the time of its issuance. This calibration electrical signal has unique waveform characteristics, which can be clearly distinguished from the test signal when the device under test is working normally, avoiding interference with the normal test process. Second, while outputting the calibration electrical signal, the PLC control unit triggers the vision inspection module to acquire images. This synchronous triggering mechanism is consistent with the triggering method during normal testing, ensuring that the calibration process accurately reflects the actual timing relationships during testing. The image acquired by the vision inspection module contains changes in image features corresponding to the calibration electrical signal, such as the flashing of a specific indicator light or the appearance of a preset mark on the display screen. Subsequently, the PLC control unit receives the calibration electrical signal reception time returned by the signal simulation acquisition module, and the image feature change time corresponding to the calibration electrical signal returned by the vision inspection module. These two sets of times reflect the time it takes for the calibration electrical signal to reach the signal simulation acquisition module through the signal transmission path, and the time it takes for the visual feature change corresponding to the calibration electrical signal to be acquired and recognized by the image, respectively. Based on the emission time of the calibration electrical signal, the reception time of the signal simulation acquisition module, and the time when the vision inspection module recognizes the image feature change, the PLC control unit calculates the time delay difference between the signal transmission path and the vision acquisition path. This time delay difference quantifies the relative deviation in timing between the two paths, reflecting the actual state of synchronization accuracy in the current testing system. Finally, if the calculated time delay difference exceeds a preset threshold, it indicates that the timing deviation has accumulated to a level that may affect the accuracy of the test, and the PLC control unit generates a compensation strategy based on this time delay difference. The specific form of the compensation strategy is to adjust the timing of triggering the visual detection module in subsequent test items, such as advancing or delaying the triggering time of image acquisition by a corresponding amount of time, so that the actual time difference between signal output and image acquisition is restored to the design tolerance range.

[0073] It's worth noting that the design of calculating the comprehensive similarity S through template matching and deep feature fusion, and then performing three-level classification based on the S value, effectively solves the stability problem of traditional visual recognition in complex environments such as lighting changes, display offsets, and glare interference. However, as the algorithm is applied more deeply in actual testing scenarios, a more subtle problem gradually emerges: when the comprehensive similarity S reaches 0.8 or higher, the system determines the status as normal, but this determination is based only on pixel-level morphological similarity, rather than semantic understanding of the displayed content. Taking the heading test as an example, the test project requires verifying whether the display screen of the shipborne navigation data recorder accurately displays the heading 090 degrees. Due to internal equipment failure, the actual value displayed on the screen may be the temperature sensor value of 090 degrees, or it may be the value of 090 in radians, or it may be garbled characters generated by abnormal display drivers, but these garbled characters are extremely similar to 090 in shape. In the above scenario, template matching algorithms will yield a similarity score close to 0.9 due to the high degree of pixel distribution matching, and convolutional neural networks will also output high scores because the extracted local edge features are similar to the standard template. The weighted fusion-based comprehensive similarity S could easily reach over 0.85. According to the original judgment logic, the system will output a conclusion that the status is normal. However, the actual function of the device is abnormal, that is, the heading information is not displayed correctly. This false positive phenomenon of high similarity but semantic error stems from the fact that the hybrid recognition algorithm is essentially still a morphological comparison algorithm. It can only judge whether the images look similar, but cannot understand the meaning expressed by the images. In the detection scenario of shipborne navigation data recorders, the semantic correctness of the displayed content is precisely the core of functional verification. Whether the heading value is correct, whether the units match, and whether the field positions correspond, all these require semantic understanding to accurately judge. Simply relying on morphological similarity cannot distinguish between displayed content with the same value but different meanings, nor can it identify hidden faults caused by display misalignment or field disorder. This issue is not prominent in traditional manual testing because testers can directly understand the meaning of the displayed content; seeing a temperature of 0.90 degrees Celsius will naturally make them realize the discrepancy with the heading test. However, in automated testing systems, enabling machines to possess this semantic understanding capability is crucial for visual recognition technology to truly replace manual labor. If this problem cannot be solved, automated testing systems will generate false pass conclusions in certain specific fault scenarios, and these faults are precisely those that need to be detected and intercepted through automated testing. Therefore, introducing a semantic understanding layer on top of morphological recognition to perform secondary verification of visual recognition results becomes a necessary approach to improve the reliability of automated testing.

[0074] To address the above problems, in one embodiment, the step of determining and outputting the visual recognition result further includes:

[0075] Extract the text content from the display interface from the real-time image;

[0076] Obtain the expected display value of the multi-source test signal corresponding to the current test item;

[0077] Perform a semantic comparison between the text content and the expected display value;

[0078] If the text content is inconsistent with the expected display value, the status is directly determined to be abnormal.

[0079] As mentioned above, the visual detection module calculates the comprehensive similarity S through template matching and deep feature fusion, and classifies the image state based on the S value. This process is mainly based on pixel-level morphological comparison, which can accurately determine the degree of similarity between the image and the standard template in appearance. However, in practical applications, there is a special anomaly: the content displayed by the device is highly similar to the expected form, even exceeding the similarity threshold of 0.8, but its semantic meaning is completely inconsistent with the actual test requirements. For example, the test project requires verifying whether the heading display is 090 degrees. Due to internal device failure, the display screen actually displays the temperature value 090 degrees, or the radian unit 090, or garbled characters due to display driver abnormality but with an appearance extremely similar to 090. In these cases, the template matching algorithm will give a high score due to high pixel similarity, and the convolutional feature extraction will also output a high similarity due to local feature similarity. The comprehensive similarity S may well reach 0.8 or higher, causing the system to misjudge the state as normal, while the actual function is abnormal. To solve this problem of morphological and semantic disconnect, this method adds a semantic verification step before the recognition result is output. Semantic verification follows this process: First, the visual inspection module extracts the text content from the real-time images captured. This extraction uses optical character recognition (OCR) technology to convert visual elements such as numbers, letters, and unit symbols in the image into character sequences that can be understood by a computer. For example, the image content captured on the screen is converted into the character sequence 090 degrees, temperature 090 degrees, or 090 rad. Second, the visual inspection module obtains the expected display value of the multi-source test signal corresponding to the current test item. This expected display value comes from the test signal content recorded by the PLC control unit. Since the test signal is output by the signal analog acquisition module, the PLC control unit clearly knows the type of test item being performed and the corresponding expected value. For example, if the current test item is a heading test, the expected display value should be a heading of 090 degrees; if the current test item is a speed test, the expected display value should be the speed value. This expected display value represents the display content that the equipment should present under normal operating conditions. Subsequently, the visual inspection module performs a semantic comparison between the extracted text content and the expected display value. Semantic comparison is fundamentally different from morphological comparison; it focuses on the meaning of the content rather than the degree of similarity in appearance. During semantic comparison, the system ignores subtle differences in format, such as treating 090 degrees and 90 degrees as having the same meaning, and treating heading 090 and 090 degrees as consistent, combining numbers and units for overall understanding. Semantic comparison is considered successful only when the semantic meaning of the text content completely matches the semantic meaning of the expected displayed value.If the semantic comparison result shows that the text content is inconsistent with the expected display value, it means that although the content displayed by the device may be similar to the expected one in form, the actual meaning it expresses is incorrect. The system directly determines that the test item is abnormal and no longer relies on the classification result of the comprehensive similarity S. This judgment logic directly covers the false positive results that may be generated by morphological recognition, ensuring that a normal result is output only when both morphology and semantics are matched.

[0080] In one embodiment, before performing a consistency comparison between the electrical signal response data and the visual recognition result, the method further includes:

[0081] The visual recognition result is compared with the multi-source test signal currently output by the signal simulation acquisition module;

[0082] If the visual recognition result is inconsistent with the multi-source test signal, the analysis result is directly determined to be inconsistent with the preset test threshold and recorded as a visual path or display device abnormality.

[0083] As described above, a pre-verification step is added before the consistency comparison between the internal state and the external display. This step directly compares the visual recognition result with the multi-source test signal currently output by the signal simulation acquisition module, constructing a triangular verification architecture among the test signal, external display, and internal state. In the aforementioned verification logic, the system parses the electrical signal response data into internal state parameters and the visual recognition result into external display state parameters, then performs a consistency comparison between the two. The core assumption of this design is that the visual recognition result can accurately reflect the device's display state, and the electrical signal response data can accurately reflect the device's internal processing state; if the two are consistent, the device is normal. However, this assumption faces challenges in specific fault scenarios: when the visual detection module itself malfunctions, such as a dirty lens causing unclear image acquisition, an aging light source causing insufficient brightness, or improper camera parameter settings causing abnormal imaging, the visual recognition result may not accurately reflect the device's true display state. In this case, even if the device's internal processing is correct and the display screen actually displays normally, the visual recognition result may still be incorrect. A more subtle scenario is that the device's internal processing is incorrect, but the display screen actually displays correctly, and the visual recognition result, due to its own malfunction, happens to match the internal error state, resulting in a false pass conclusion. This interlocking artifact problem is difficult to detect in traditional dual-verification architectures because the system lacks an independent third-party reference benchmark to verify the correctness of the visual recognition results. To address this issue, this method introduces a third-party benchmark verification before consistency comparison. Specifically, the PLC control unit first acquires the multi-source test signals currently output by the signal simulation acquisition module. Since the test signals are output by the PLC control unit, the PLC knows the specific content of the test signals currently being sent, such as the heading value, speed value, and whether an alarm command has been sent. These test signals, as the most basic external inputs, are third-party benchmarks independent of the device's internal processing and visual acquisition. Subsequently, the PLC control unit compares the visual recognition results with this third-party benchmark. The visual recognition results reflect the content on the device's display interface, while the test signals reflect the content that the device should display. Under the premise that the device is working normally and the visual acquisition is error-free, the visual recognition results should be consistent with the test signals. If the two are inconsistent, it indicates either a fault in the visual acquisition path causing distorted recognition results, or the display device itself has failed to correctly display the content corresponding to the test signals. Both situations belong to anomalies in the visual path or the display device. When the comparison results show inconsistencies, the PLC control unit directly determines that the test item does not meet the preset test threshold, eliminating the need for further consistency comparison between the internal state and the external display. Simultaneously, the system records this anomaly as a visual path or display device malfunction, facilitating subsequent fault location and maintenance.

[0084] Reference Figure 3This application also provides a fully automated testing device for shipborne navigation data recorders based on machine vision, comprising:

[0085] Response module 1 is used to respond to the test start command issued by the host computer. The PLC control unit controls the signal analog acquisition module to output multi-source test signals to the shipborne navigation data recorder under test.

[0086] Acquisition module 2 is used to trigger the visual detection module to acquire images of the tested shipborne navigation data recorder while the signal simulation acquisition module outputs the multi-source test signal, and acquire real-time images including the display interface and / or indicator light status.

[0087] The first receiving module 3 is used to receive the electrical signal response data transmitted back by the signal simulation acquisition module. The electrical signal response data is acquired by the signal simulation acquisition module from the shipborne navigation data recorder under test.

[0088] The second receiving module 4 is used to receive the visual recognition result output by the visual detection module;

[0089] The correlation analysis module 5 is used to perform correlation analysis between the electrical signal response data and the visual recognition result. If the analysis result does not meet the preset test threshold, an abnormal control signal is output and uploaded to the host computer for recording.

[0090] The generation module 6 is used to summarize the test data and generate a test report through the host computer if the analysis results meet the preset test threshold.

[0091] As described above, it is understood that each component of the fully automated testing device for a shipborne navigation data recorder based on machine vision proposed in this application can realize the function of any one of the fully automated testing methods for a shipborne navigation data recorder based on machine vision as described above, and the specific structure will not be described in detail.

[0092] Reference Figure 4 This application also provides a computer device, which may be a server, and its internal structure may be as follows: Figure 4As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores monitoring data and other data. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a fully automated testing method for a shipborne navigation data recorder based on machine vision.

[0093] The processor described above executes the fully automated testing method for a shipborne navigation data recorder based on machine vision, including: responding to a test start command issued by a host computer, the PLC control unit controls the signal simulation acquisition module to output multi-source test signals to the shipborne navigation data recorder under test; simultaneously with the output of the multi-source test signals by the signal simulation acquisition module, the vision detection module is triggered to acquire images of the shipborne navigation data recorder under test, obtaining real-time images including the display interface and / or indicator light status; receiving electrical signal response data returned by the signal simulation acquisition module, the electrical signal response data being acquired by the signal simulation acquisition module from the shipborne navigation data recorder under test; receiving the visual recognition result output by the vision detection module; performing correlation analysis between the electrical signal response data and the visual recognition result; if the analysis result does not meet a preset test threshold, an abnormal control signal is output and uploaded to the host computer for recording; if the analysis result meets the preset test threshold, the host computer summarizes the test data and generates a test report.

[0094] One embodiment of this application also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements a fully automated testing method for a shipborne navigation data recorder based on machine vision, comprising the following steps: responding to a test start command issued by a host computer, a PLC control unit controls a signal simulation acquisition module to output multi-source test signals to the shipborne navigation data recorder under test; simultaneously with the output of the multi-source test signals by the signal simulation acquisition module, a vision detection module is triggered to acquire images of the shipborne navigation data recorder under test, obtaining real-time images including the display interface and / or indicator light status; receiving electrical signal response data returned by the signal simulation acquisition module, the electrical signal response data being acquired by the signal simulation acquisition module from the shipborne navigation data recorder under test; receiving the visual recognition result output by the vision detection module; performing correlation analysis between the electrical signal response data and the visual recognition result; if the analysis result does not meet a preset test threshold, an abnormal control signal is output and uploaded to the host computer for recording; if the analysis result meets the preset test threshold, the host computer summarizes the test data and generates a test report.

[0095] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in this application and in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0096] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0097] The above description is only a preferred embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural changes made based on the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A fully automated testing method for shipborne navigation data recorders based on machine vision, characterized in that, The method includes: In response to the test start command issued by the host computer, the PLC control unit controls the signal analog acquisition module to output multi-source test signals to the shipborne navigation data recorder under test; While the signal simulation acquisition module outputs the multi-source test signal, the visual detection module is triggered to acquire images of the shipborne navigation data recorder under test, obtaining real-time images including the display interface and / or indicator light status. The system receives electrical signal response data transmitted back by the signal simulation acquisition module, which is acquired by the signal simulation acquisition module from the shipborne navigation data recorder under test. Receive the visual recognition result output by the visual detection module; The electrical signal response data is correlated with the visual recognition result. If the analysis result does not meet the preset test threshold, an abnormal control signal is output and uploaded to the host computer for recording. If the analysis results meet the preset test threshold, the host computer will summarize the test data and generate a test report.

2. The fully automated testing method for a shipborne navigation data recorder based on machine vision according to claim 1, characterized in that, After the step of correlating and analyzing the electrical signal response data with the visual recognition results, the method further includes: The electrical signal response data is parsed into a first state parameter characterizing the internal state of the shipborne navigation data recorder under test. The visual recognition result is interpreted as a second state parameter that characterizes the display status of the external human-machine interface of the shipborne navigation data recorder under test. Perform a consistency comparison between the first state parameter and the second state parameter; If the first state parameter is inconsistent with the second state parameter, the analysis result is determined to be inconsistent with the preset test threshold. The output of the abnormal control signal and its uploading to the host computer for recording includes: The PLC control unit generates channel switching instructions and alarm trigger instructions based on the inconsistent comparison results. In response to the channel switching command, the drive relay switches to the backup test channel to continue executing the unfinished test items; In response to the alarm trigger command, the alarm device is driven to output an audible, visual, and / or electrical signal alarm; The inconsistent comparison results and corresponding abnormal control signals are uploaded to the host computer for storage and display.

3. The fully automated testing method for a shipborne navigation data recorder based on machine vision according to claim 1, characterized in that, The visual recognition result output by the visual detection module includes the following steps: The real-time image is preprocessed to obtain a preprocessed enhanced image; Feature extraction is performed on the preprocessed enhanced image to obtain template matching similarity S1 and convolution feature similarity S2; A weighted similarity determination function is used to fuse the template matching similarity S1 and the convolutional feature similarity S2 to obtain a comprehensive feature. Based on the comprehensive features, the comprehensive similarity S is calculated, where the calculation formula is: S = 0.6 × S1 + 0.4 × S2; State classification is performed based on the comprehensive similarity S, including: If S≥0.8, then the status is considered normal; If 0.6 ≤ S < 0.8, then proceed to the review and judgment channel; If S < 0.6, the state is determined to be abnormal; Based on state judgment classification, the visual recognition result is determined and output.

4. The fully automated testing method for a shipborne navigation data recorder based on machine vision according to claim 1, characterized in that, After the step of determining if the analysis result does not meet the preset test threshold, the method further includes: If the anomaly is a skippable, non-related test item anomaly, a channel switching command is generated to drive the relay to switch to the backup test channel, skip the current test item and continue to execute the next test item; If the anomaly is a related test item anomaly or a system-level anomaly that affects subsequent tests, an alarm trigger command is generated to drive the alarm device to output an audible and visual alarm, and the test process is paused to await manual intervention.

5. The fully automated testing method for a shipborne navigation data recorder based on machine vision according to claim 1, characterized in that, The method further includes: periodically performing synchronous calibration in the test sequence, the synchronous calibration step including: Output a calibration electrical signal to the signal analog acquisition module and record the time when the calibration electrical signal is emitted; Simultaneously with outputting the calibration electrical signal, the visual detection module is triggered to acquire an image; The receiving signal received by the analog acquisition module and the image feature change time corresponding to the calibration signal received by the visual detection module. Based on the transmission time, reception time, and image feature change time, calculate the time delay difference between the signal transmission path and the visual acquisition path; If the delay difference exceeds a preset threshold, a compensation strategy is generated based on the delay difference.

6. The fully automated testing method for a shipborne navigation data recorder based on machine vision according to claim 3, characterized in that, The step of determining and outputting the visual recognition result further includes: Extract the text content from the display interface from the real-time image; Obtain the expected display value of the multi-source test signal corresponding to the current test item; Perform a semantic comparison between the text content and the expected display value; If the text content is inconsistent with the expected display value, the status is directly determined to be abnormal.

7. The fully automated testing method for a shipborne navigation data recorder based on machine vision according to claim 2, characterized in that, Before performing a consistency comparison between the electrical signal response data and the visual recognition result, the method further includes: The visual recognition result is compared with the multi-source test signal currently output by the signal simulation acquisition module; If the visual recognition result is inconsistent with the multi-source test signal, the analysis result is directly determined to be inconsistent with the preset test threshold and recorded as a visual path or display device abnormality.

8. A fully automated testing device for shipborne navigation data recorders based on machine vision, characterized in that, include: The response module is used to respond to the test start command issued by the host computer. The PLC control unit controls the signal analog acquisition module to output multi-source test signals to the shipborne navigation data recorder under test. The acquisition module is used to trigger the visual detection module to acquire images of the shipborne navigation data recorder under test while the signal simulation acquisition module outputs the multi-source test signal, and to acquire real-time images including the display interface and / or indicator light status. The first receiving module is used to receive the electrical signal response data transmitted back by the signal simulation acquisition module, wherein the electrical signal response data is acquired by the signal simulation acquisition module from the shipborne navigation data recorder under test. The second receiving module is used to receive the visual recognition result output by the visual detection module; The correlation analysis module is used to perform correlation analysis between the electrical signal response data and the visual recognition result. If the analysis result does not meet the preset test threshold, an abnormal control signal is output and uploaded to the host computer for recording. The generation module is used to summarize the test data and generate a test report through the host computer if the analysis results meet the preset test threshold.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images