Time interval measuring system based on high-speed video collector

Abstract

The invention discloses a time interval measurement system based on a high-speed video collector, belongs to the technical field of time interval measurement and metering traceability, and aims at solving the problems that in the prior art, time interval measurement triggering is not synchronous, errors are large, and visual triggering equipment is difficult to calibrate accurately. The system adopts a synchronous staggered double-high-speed camera to collect visual information of a device under test (DUT) and an external standard time interval instrument; embedding traceable first electronic timestamp information for each frame of acquired image based on the electric signal of the standard instrument; meanwhile, starting and stopping event characteristics of the DUT and visual display readings of the standard instrument are automatically recognized through an artificial intelligence (AI) model, a second visual time information source is obtained, and the system calculates the accurate time when an event occurs through a sub-frame-level time interpolation algorithm; and executing dual redundancy cross validation between the first electronic timestamp information and the second visual time information, calculating a high-precision time interval on the basis of validation, and evaluating the uncertainty of the high-precision time interval.

Description

Technical Field

[0001] The present invention relates to the technical field of time interval measurement and metrological traceability, and specifically relates to a time interval measurement system based on a high-speed video collector. Background Art

[0002] As a basic physical quantity, the accurate measurement of time interval plays a crucial role in many fields such as national defense, scientific research, industrial production, sports competitions, and product quality inspection. In order to ensure the accuracy and reliability of measurement results, regular calibration or verification of various time interval measurement devices (such as stopwatches, timers, oscilloscopes, bullet velocity meters, etc.) and realizing the traceability of the quantity value to the national time and frequency standard are important components of metrological work.

[0003] However, in the prior art, there are significant technical challenges in the quantity value transfer of time intervals, especially when calibrating a wide variety of devices under test (DUTs) with different triggering methods. Traditional calibration methods usually rely on a high-precision time interval measuring instrument as a standard device, and synchronously trigger and compare with the device under test through an electrical signal interface. However, when faced with a large number of devices without a standard electrical interface, whose start / stop states can only be judged by visual observation (such as the start / stop of the pointer of a mechanical stopwatch, the jump of the liquid crystal digits of an electronic stopwatch, the physical phenomenon of a bullet passing through a specific area in a bullet velocity meter, etc.), traditional electrical measurement methods are difficult to apply. The main problems are as follows: Triggering asynchrony introduces huge uncertainties, and it is difficult to achieve ideal synchronous start and stop between the standard device and the device under test. Whether it is manual button triggering or using simple optoelectronic switches, etc., it is difficult to avoid the inconsistency of triggering time delays. This asynchrony will introduce significant measurement errors. Especially when measuring a long time interval or requiring high precision, this error may even reach the second level, seriously affecting the accuracy of the calibration result. The subjectivity and low efficiency of visual interpretation. For devices that rely on visual observation, manual interpretation is often the main means, but this is not only inefficient but also extremely prone to introducing human reading errors and reaction time errors, resulting in poor repeatability and low precision of measurement results. Low automation level. Existing calibrations for visually triggered devices mostly rely on manual operations, with a low automation level and difficult to meet the calibration requirements of large quantities and high efficiency.

[0004] To solve the above problems, researchers have begun to explore the use of image or video technology for time interval measurement or device calibration.

[0005] For example, Chinese invention patent CN103229112B discloses a method for measuring the accuracy of a watch including a visual display device. The method captures a first image of the watch at a first moment t1, determines and stores the corresponding first display value v1; captures a second image at a second moment t2 after a time interval, determines and stores the corresponding second display value v2; and finally calculates the rate of change of the display (v2 - v1) / (t2 - t1). This method uses image capture and time recording to evaluate the running speed of the watch.

[0006] Again, Chinese invention patent CN108549845B discloses a method for determining the position of the pointers on the watch face. The method captures a video of the watch face through a camera, extracts the image, locates the dial area and the pointers, and finally determines the angle by which the pointers deviate from the preset point. This method focuses on real-time determining the spatial position of the pointers through image processing techniques (including asynchronous processing, selectively ignoring frames, etc.), and considers smooth operation on devices with limited computing resources (such as mobile phones).

[0007] Although the above prior arts use image or video technology to process time-related devices, there are still certain limitations, such as: insufficient time resolution and accuracy, difficult to meet the requirements of high-precision time interval measurement, failure to provide a complete, accurate and internally calibrated value traceability scheme, especially how to ensure a high-precision and high-reliability association between the occurrence time of visual events and the standard time reference. A simple time reference source is difficult to guarantee the traceability accuracy and confidence at the microsecond or even nanosecond level, and fails to directly address and effectively solve the significant measurement uncertainty problem caused by trigger asynchronization commonly existing in the metrology calibration scenario. Summary of the Invention

[0008] The object of the present invention is to overcome the deficiencies of the prior art and propose a time interval measurement system based on a high-speed video collector to solve the above problems.

[0009] The object of the present invention is achieved by the following technical solutions: A time interval measurement system based on a high-speed video collector, comprising: A high-speed video acquisition module configured to acquire an image sequence including the visual start-stop event characteristics of the device under test (DUT) and the visual display of an external standard time interval measurement instrument. The high-speed video acquisition module includes at least two high-speed cameras; A synchronization control and time information processing module connected to the high-speed video acquisition module. The synchronization control and time information processing module includes: an external signal receiving unit configured to receive an electrical signal output by an external standard time interval measurement instrument as a time reference; A synchronous trigger signal generation unit, configured to generate a synchronous trigger signal based on an electrical signal to drive at least two high-speed cameras to perform image acquisition by exposing in a synchronous manner with a preset time stagger; A timestamp generation and embedding unit, configured to generate and embed traceable first electronic timestamp information for each frame of the acquired image sequence based on the electrical signal; An intelligent image analysis module, configured to process the image sequence acquired by the high-speed video acquisition module and perform analysis using an artificial intelligence (AI) model. The intelligent image analysis module includes: A DUT event feature recognition unit, configured to analyze the image sequence using the AI model to identify the visual features of the start event and stop event of the DUT; A standard instrument visual reading unit, configured to analyze the image sequence using the AI model or optical character recognition (OCR) technology to identify the visual display time value of the standard time interval measuring instrument, so as to obtain a second visual time information source; A sub-frame level time interpolation calculation unit, configured to calculate the sub-frame level accuracy moments at which the start and stop events of the DUT and the changes in the visual display time value occur based on the first electronic timestamp information and the relevant image frames including the start and stop events of the DUT and the changes in the visual display time value; A data processing and verification module, configured to communicate with the intelligent image analysis module and the synchronous control and time information processing module. The data processing and verification module includes: A dual-source time information cross-verification unit, configured to verify the accuracy and consistency of the time measurement link by comparing and analyzing the time based on the first timestamp information and the time based on the second visual time information; A time interval calculation unit, configured to calculate the time interval of the DUT based on the verification result of the cross-verification unit and the sub-frame level accuracy moments of the start event and stop event of the DUT.

[0010] The synchronous trigger signal generated by the synchronous trigger signal generation unit is an interleaved trigger signal used to drive the first and second cameras among at least two high-speed cameras to perform time-division multiplexing acquisition with a preset time stagger.

[0011] The timestamp generation and embedding unit writes the first electronic timestamp information into the metadata of each frame of the image, and the resolution of the first electronic timestamp information reaches the microsecond level or the nanosecond level.

[0012] The intelligent image analysis module further includes an image preprocessing unit, which is arranged on the data input paths of the DUT event feature recognition unit and the standard instrument visual reading unit, and is configured to perform at least one preprocessing operation of denoising, image enhancement or geometric correction on the acquired image sequence.

[0013] The sub-frame level time interpolation calculation unit is configured to calculate the sub-frame level accuracy time by analyzing at least one of the pixel-level displacement information, optical flow vector information, or brightness / gray scale intensity distribution change information of the target features related to the event in several frames of images before and after the event occurrence moment.

[0014] The dual-source time information cross-verification unit is configured to compare the identified visual time value at a specific sub-frame level accuracy time when the standard instrument vision display time value determined by the standard instrument vision reading unit and the sub-frame level time interpolation calculation unit changes, with the theoretical time value represented by the first electronic time stamp information corresponding to the specific sub-frame level accuracy time, so as to perform cross-verification.

[0015] It further includes an adaptive optical control module, and the adaptive optical control module includes: an optical state evaluation unit configured to evaluate the current imaging quality index or the target position state according to the output of the intelligent image analysis module or directly analyze the image; an optical component driving unit configured to control the electro-optical components associated with the high-speed video acquisition module in response to the evaluation result of the optical state evaluation unit, so as to automatically adjust the focus parameters or the field of view pointing of at least one of at least two high-speed cameras.

[0016] The AI models adopted by the DUT event feature recognition unit and the standard instrument vision reading unit are deep learning models trained for specific DUT event types and specific standard instrument display styles.

[0017] The preset time interleaving amount is set to be half of the frame period of the high-speed camera.

[0018] The data processing and verification module further includes an uncertainty evaluation unit configured to evaluate the measurement uncertainty of the calculated DUT time interval by comprehensively considering at least a part of the following factors: the uncertainty of the first electronic time stamp information, the confidence level of the AI model recognition result, the uncertainty introduced by the sub-frame level time interpolation calculation, and the time information consistency deviation obtained by the cross-verification unit.

[0019] The beneficial effects of the present invention are: 1. Using high-speed video acquisition as a basic means, combined with accurate timestamp information, it can capture the moment of event occurrence more accurately than traditional manual or simple electrical signal triggering methods, effectively reducing the huge uncertainty caused by trigger asynchrony. By adopting the technology of time-division multiplexing and interleaving acquisition with dual high-speed cameras, the sampling density on the time axis is effectively improved, which is equivalent to doubling the effective frame rate of the system, further enhancing the ability to capture high-speed transient events and the potential for time resolution. The present invention introduces a sub-frame-level time interpolation algorithm, which can deeply analyze the pixel-level information (such as displacement, optical flow, grayscale change, etc.) of several frames of images before and after the event, and calculate the precise moment of the event that far exceeds the physical frame rate limit of the camera, and improve the time resolution to the microsecond level or even higher, thereby fundamentally improving the ultimate accuracy of time interval measurement.

[0020] 2. The system not only relies on the electrical signal generated by an external high-precision standard time interval instrument to drive the generation and embedding of high-precision electronic timestamp information in the image as the main time reference, but also captures the visual display information of the standard instrument in real time through a high-speed camera and uses artificial intelligence technology to read it in real time, forming a second independent time information path. The system has a built-in cross-validation function, which can compare and analyze the consistency of the time information obtained by the two independent paths in real time or afterwards, and can instantly discover and evaluate possible deviations, delays or faults in the time measurement link, greatly enhancing the credibility of the measurement results and providing reliability assurance and self-diagnosis capabilities that cannot be matched by existing single traceability path systems.

[0021] 3. Apply advanced artificial intelligence (AI) models (especially deep learning models) to automatically identify the start and stop visual event characteristics of the device under test. This enables the system to process various complex and non-standardized visual trigger signals (such as pointers passing over scales, specific digital combinations appearing, physical phenomena occurring, etc.), get rid of the reliance on simple thresholds or manual interpretation, and realize intelligent event detection. Similarly, AI or OCR technology is used to realize automatic reading of the visual display of external standard time interval instruments, avoiding manual reading errors. The integrated adaptive optical control function can automatically track the target and automatically adjust the focal length without cumbersome manual settings, ensuring that high-quality image data can be obtained under different conditions, significantly improving the system's automation level and ease of use.

[0022] 4. The system adopts a non-contact high-speed vision acquisition method, and its applicability is almost not limited by the interface type or triggering method of the device under test. As long as the device or process under test has observable start and stop characteristics, measurement or calibration can be carried out. Especially importantly, it can effectively solve the problem of high-precision value traceability of vision-triggered devices (such as various mechanical stopwatches, bullet speed measuring instruments, certain timers in industrial processes, etc.) that have long troubled the metrology field. These devices often do not have a standard electrical signal output, and it is difficult to accurately calibrate them using traditional methods. The present invention provides a practical and accurate calibration method for such devices.

[0023] 5. By connecting to an external traceable standard time interval measuring instrument and adopting precise electronic timestamp technology, the system establishes a clear and stable metrological traceability chain to ensure that the measurement results can ultimately be traced back to the national time and frequency reference. The built-in uncertainty evaluation function follows the internationally accepted Guide to the Expression of Uncertainty in Measurement (GUM) and can comprehensively evaluate all significant uncertainty components including timestamp accuracy, AI recognition confidence, interpolation algorithm error, and cross-validation deviation, and finally outputs a complete measurement result including the measured value and its corresponding combined standard uncertainty (or expanded uncertainty). This makes the measurement results not only accurate but also scientifically rigorous and complete, meeting the requirements of high-standard metrology applications.

[0024] 6. The image preprocessing steps help to cope with adverse imaging conditions such as noise and uneven illumination. The adaptive optical function enables the system to have better tolerance to factors such as small changes in the installation position and environmental vibration. The overall automated design reduces errors that may be introduced by human operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is the system architecture of the present invention Figure 1 ; Figure 2 is the system architecture of the present invention Figure 2 ; Figure 3 is the system architecture of the present invention Figure 3 . DETAILED DESCRIPTION OF THE INVENTION

[0026] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present invention.

[0027] It should be noted that in the following solutions, the orientation concepts of "left", "right", "up", "down", "front", "back", "inside", and "outside" are all relative directions, and will not be listed one by one here.

[0028] Embodiment 1 As Figures 1 to 3 shown, this embodiment provides a time interval measurement system based on a high-speed video collector. This system constructs a modular architecture, including a core high-speed video acquisition module 10, a synchronization control and time information processing module 20, an intelligent image analysis module 30, and a data processing and verification module 40.

[0029] The core task of the high-speed video acquisition module 10 is to capture visual information. In this embodiment, this module specifically includes a first high-speed camera and a second high-speed camera. The performance parameters of these two cameras are matched. For example, a frame rate not lower than 1000 fps is selected (a high-frame-rate camera such as 3500 fps can be used in experiments), and it has sufficient resolution to clearly image the key features of the device under test (DUT) (such as pointers, digital displays, physical markers) and the visual display screen of the external standard time interval measurement instrument used as the time reference. These two cameras are precisely installed and positioned to ensure that they can observe the visualization start-stop event feature area of the DUT and the display area of the external standard time interval measurement instrument simultaneously or as needed.

[0030] The synchronization control and time information processing module 20 is the core of the system's time control and traceability. It is connected through its internal external signal receiving unit 21 to an external standard time interval measuring instrument that is external to the system, verified or calibrated by an authoritative institution, and traceable to the national time and frequency reference (for example, a high-resolution microsecond meter, such as a standard with a 1µs resolution, or a time and frequency counter with higher precision). The external signal receiving unit 21 receives the stable and accurate electrical signal (such as a 10MHz frequency signal or a 1PPS second pulse signal) output by the standard instrument as the unified time reference for the entire system. The synchronization trigger signal generation unit 22 generates two trigger signals with precisely controlled phases based on this input electrical signal. According to claim 2, these two trigger signals are interleaved trigger signals, which respectively drive the first high-speed camera and the second high-speed camera, making their exposure moments evenly interleaved on the time axis. For example, the start moment of the exposure of the second camera is precisely located at the midpoint between the start moments of two adjacent exposures of the first camera, thus realizing time-division multiplexing acquisition and effectively improving the time sampling density. At the same time, the timestamp generation and embedding unit 23 also generates high-precision and traceable first electronic timestamp information for each frame of the image collected by the high-speed video acquisition module 10 based on the input standard electrical signal. According to claim 3, this timestamp information has a high resolution at the microsecond or nanosecond level and is written in real time into the metadata of the corresponding image frame, ensuring that each frame of the image has an accurate and traceable absolute time mark.

[0031] The intelligent image analysis module 30 is responsible for processing the image sequence from the high-speed video acquisition module 10. In the basic framework of this embodiment, this module at least includes units that perform core functions: the DUT event feature recognition unit 32 uses an AI model to recognize the start and stop events of the DUT; the standard instrument visual reading unit 33 uses AI / OCR technology to recognize the display readings of the standard instrument, constituting the second visual time information source; the sub-frame level time interpolation calculation unit 34 is used to calculate the precise moment when the event occurs.

[0032] The data processing and verification module 40 is the data final processing and verification center of the system. In this embodiment, the focus is on its core verification and calculation functions. The dual-source time information cross-verification unit 42 in this module plays a key role. According to claim 6, it is configured to execute a key verification process: comparing the second visual time information (i.e., the identified visual time value) obtained at a specific moment (this moment is identified by the standard instrument vision reading unit 33 and accurately located by the sub-frame level time interpolation calculation unit 34, such as the exact moment when the standard instrument display number jumps) with the theoretical time value represented by the first electronic timestamp information embedded in the image metadata corresponding to this exact moment. Through this comparison, the accuracy and consistency of the entire time measurement link (from the standard electrical signal to the timestamp generation and embedding, and then to the visual information acquisition, identification, and timing) can be verified. Finally, based on the confirmation that the time link is reliable (i.e., the cross-verification passes or the deviation is within the allowable range), the time interval calculation unit 43 mainly calculates the final time interval of the device under test (DUT) according to the start and stop event moments of the DUT provided by the timestamp generation and embedding unit 23 and refined by the sub-frame level time interpolation calculation unit 34.

[0033] Each module is connected through a high-speed data interface (such as CoaXPress or CameraLink, GigEVision, etc.) and a control bus, and is coordinated and scheduled by the central processing and control unit (including part or all of modules 30 and 40).

[0034] Working process: System initialization and connection: Connect the external standard time interval measurement instrument to the external signal receiving unit 21 of the synchronization control and time information processing module 20, align the two cameras of the high-speed video acquisition module 10 with the key area of the DUT and the display screen of the standard instrument, and start the system software.

[0035] Synchronous acquisition start: The operator issues a start measurement instruction or the system waits for a DUT event to trigger, and the synchronization control and time information processing module 20 starts to work: The external signal receiving unit 21 receives the standard electrical signal; the synchronous trigger signal generation unit 22 generates interleaved trigger pulses based on this signal and drives the first and second cameras in the high-speed video acquisition module 10 respectively; the timestamp generation and embedding unit 23 generates and embeds the first electronic timestamp information for each frame of the acquired image, and the high-speed video acquisition module 10 continuously acquires the image sequence according to the interleaved time sequence.

[0036] Image Information Processing (Parallel): The intelligent image analysis module 30 receives and processes the image sequence. The DUT event feature recognition unit 32 monitors and recognizes the start and stop event features of the DUT. The standard instrument vision reading unit 33 recognizes the real-time display readings of the standard instrument (the second vision time information source). The sub-frame level time interpolation calculation unit 34 processes the identified key event frames and calculates the exact occurrence time.

[0037] Cross-validation: After the dual-source time information cross-validation unit 42 in the data processing and validation module 40 obtains the standard instrument vision readings and their exact occurrence times, it compares and validates them with the first electronic timestamp information at the corresponding times.

[0038] Result Calculation: After the cross-validation passes, the time interval calculation unit 43 uses the exact start and stop times of the DUT obtained through verification and sub-frame level interpolation to calculate the time interval T = T_stop - T_start.

[0039] Result Output: The system outputs the calculated time interval value.

[0040] The traceable external standard time interval measuring instrument signal is received by the external signal receiving unit 21, and the first electronic timestamp information with high resolution (microsecond level or nanosecond level) is generated and embedded by the timestamp generation and embedding unit 23, providing an accurate and reliable time reference for the entire measurement and laying the foundation for high-precision measurement.

[0041] To improve the time resolution and event capture ability, the high-speed video acquisition module 10 uses two cameras, and the time-division multiplexing interleaved acquisition is realized by the synchronous trigger signal generation unit 22, effectively improving the time sampling rate and enhancing the capture ability and time resolution ability for fast transient events.

[0042] Extremely high measurement reliability and self-checking ability. The core of this embodiment lies in establishing a dual-redundancy traceability mechanism. The system not only relies on the high-precision first electronic timestamp information, but also obtains the second vision time information source through the standard instrument vision reading unit 33, and the dual-source time information cross-validation unit 42 compares and validates the two. This design can real-time check the accuracy of the time measurement link, discover potential deviations, greatly enhance the credibility of the measurement results and the reliability of the system, and has the self-checking ability that the traditional single-traceability path method does not have.

[0043] Solve the problem of accurate traceability of vision trigger devices. The system directly captures the visual features of the DUT through the high-speed video acquisition module 10, regardless of whether it has an electrical signal output, providing an accurate and reliable value transfer path for devices such as mechanical stopwatches and bullet speed measuring instruments that are difficult to calibrate by traditional electrical measurement methods.

[0044] The uncertainty introduced by triggering asynchrony is reduced. Compared with the problem that it is difficult to synchronously trigger the standard device and the device under test in the traditional method, the uncertainty in the triggering link is significantly reduced (can be reduced to the millisecond level or even lower) in this system through unified visual observation and accurate timestamp recording.

[0045] In summary, the core system framework and dual redundant traceability mechanism demonstrated in Embodiment 1 lay the foundation for the high precision and high reliability of the entire invention and effectively solve the key problems faced by the prior art.

[0046] Embodiment Two As Figures 1 to 3 shown, in this embodiment, based on the core system framework described in Embodiment 1, the functions of the intelligent image analysis module 30 and the data processing and verification module 40 are enhanced and refined to achieve a higher level of intelligent analysis and time measurement accuracy.

[0047] First, the enhancement of the intelligent image analysis module 30 is reflected in the following aspects: Before the data stream enters the core recognition unit, an image preprocessing unit 31 is added. This unit is located on the path where the high-speed video acquisition module 10 outputs image data to the DUT event feature recognition unit 32 and the standard instrument visual reading unit 33. The image preprocessing unit 31 is configured to perform necessary optimization operations on the original image sequence. For example, it uses Gaussian filtering or median filtering algorithms for denoising to reduce the impact of image sensor noise on subsequent recognition; performs image enhancement through histogram equalization or contrast stretching to improve the clarity and contrast of DUT event features or standard instrument displays; or performs geometric corrections such as lens distortion correction and affine transformation to correct the deformation introduced by the optical system and ensure the accuracy of feature positioning. This unit performs at least one of the above operations, and its purpose is to provide higher-quality and more standardized image input for subsequent AI analysis.

[0048] In this embodiment, the DUT event feature recognition unit 32 and the standard instrument visual reading unit 33 adopt a specifically trained deep learning model as their core recognition engine. For example, the DUT event feature recognition unit 32 can use an object detection model based on a convolutional neural network (CNN) (such as YOLO, FasterR-CNN) to accurately locate key areas on the DUT (such as the pointer tip, digital jump position), and combine a recurrent neural network (RNN) or a long short-term memory network (LSTM) to analyze the changes in this area over time series, so as to accurately identify the occurrence of start and stop events. For the standard instrument visual reading unit 33, an OCR (optical character recognition) deep learning model specifically trained for digital or dial reading, or a general scene text recognition model, can be used to achieve high-precision automatic reading of the time value on the standard instrument display screen. These AI models are all trained with a large amount of labeled data including various specific DUT event types (such as the mechanical stopwatch pointer crossing the line, the digital change on the electronic display screen, the visual characteristics at the moment of the bullet leaving the barrel) and specific standard instrument display styles (different fonts, different background lights, different tilt angles) to obtain high recognition accuracy and good generalization ability.

[0049] In this embodiment, the sub-frame level time interpolation calculation unit 34 adopts a more advanced calculation method to achieve ultra-high time resolution. It not only satisfies the determination of which frame interval the event occurs in, but also deeply analyzes the pixel-level information of the target features related to the event in the images of several frames before and after the event occurrence moment (such as the frame before the event occurrence, the frame of the event occurrence, the frame after the event occurrence). Specifically, the sub-frame level time interpolation calculation unit 34 is configured to calculate the sub-frame level accuracy moment through at least one of the following methods: Analyze the pixel-level displacement information of key features (such as the pointer edge, digital pen segment edge, moving object contour), and deduce the exact moment when it crosses a specific boundary (such as the scale line, display area boundary) by establishing its motion model between frames (such as a uniform or uniformly accelerated model).

[0050] Calculate the optical flow vector information of the event area, and judge the exact time point of the state mutation by analyzing the change of the pixel velocity field over time.

[0051] Analyze the change information of the brightness or gray scale intensity distribution of event-related pixels (such as the jumping digital pen segment, the background swept by the pointer) over time. For example, the function fitting (such as S-shaped curve fitting) can be performed on the pixel intensity change curve, and the inflection point of the curve or the moment when it reaches a specific threshold is determined as the sub-frame level accuracy moment of the event occurrence. Through these methods, the sub-frame level time interpolation calculation unit 34 can improve the positioning accuracy of the event occurrence moment to a level far exceeding the physical frame interval of the camera (even the effective frame interval Δt / 2), reaching the microsecond level.

[0052] Secondly, the enhancement of the data processing and verification module 40 is reflected in the addition of a dedicated uncertainty evaluation function: in addition to the dual-source time information cross-verification unit 42 and the time interval calculation unit 43 described in the first embodiment, the module also clearly includes an uncertainty evaluation unit 44. The function of this unit is to give the measurement uncertainty that conforms to metrological specifications (such as the GUMB type evaluation) to the DUT time interval calculated by the time interval calculation unit 43. The uncertainty evaluation unit 44 is configured to comprehensively consider the uncertainty components introduced by multiple key factors in the entire measurement link. These factors at least include: The uncertainty of the first electronic timestamp information provided by the timestamp generation and embedding unit 23 of the synchronization control and time information processing module 20 itself (derived from external standard sources and internal processing circuits).

[0053] The confidence level of the recognition result of the AI model adopted by the DUT event feature recognition unit 32 and the standard instrument vision reading unit 33 in the intelligent image analysis module 30, or the classification / recognition error probability.

[0054] The model uncertainty or goodness of fit introduced by the interpolation algorithm adopted by the sub-frame level time interpolation calculation unit 34.

[0055] The consistency deviation or verification residual between the two time information chains obtained by the dual-source time information cross-verification unit 42. The uncertainty evaluation unit 44 synthesizes these components to obtain the total uncertainty of the final time interval measurement result.

[0056] Working process: The working process of this embodiment adds several key steps on the basis of the first embodiment: Initialization and setting: (Same as the first embodiment) Adaptive optimization: (Optional) Synchronous data acquisition: (Same as the first embodiment) Image preprocessing: The collected image sequence is first sent to the image preprocessing unit 31 in the intelligent image analysis module 30 for denoising, enhancement or correction.

[0057] AI intelligent analysis: The preprocessed image sequence is sent to the DUT event feature recognition unit 32 and the standard instrument vision reading unit 33, and the trained deep learning model is used to automatically identify the DUT start / stop events and the standard instrument vision readings.

[0058] Sub-frame level precise timing: The sub-frame level time interpolation calculation unit 34 processes the key event frame sequence recognized by the AI, and calculates the sub-frame level precise moments of the DUT start / stop events and the changes in the standard instrument readings by using information such as pixel-level displacement, optical flow or gray level change.

[0059] Cross-validation: (Same as Example 1, but the time values for comparison are from the sub-frame level calculation results with higher precision at this time) Result calculation: (Same as Example 1) Uncertainty evaluation: The uncertainty evaluation unit 44 collects the uncertainty information of each link (timestamp, AI confidence level, interpolation error, verification deviation, etc.) and calculates the total uncertainty of the final time interval.

[0060] Output: Output the measurement result of the time interval and its combined standard uncertainty (or expanded uncertainty).

[0061] Breakthrough of the precision limit. Through the sub-frame level time interpolation calculation unit 34, the time resolution of the system is no longer limited by the physical frame rate or effective frame rate of the camera and can reach the microsecond level or even higher, thus greatly improving the final precision of the time interval measurement.

[0062] Enhancement of analysis robustness and accuracy. The image preprocessing unit 31 optimizes the quality of the input image, reduces the influence of noise and distortion on subsequent recognition. The DUT event feature recognition unit 32 and the standard instrument vision reading unit 33 using the specifically trained deep learning model enable the system to process complex backgrounds, lighting changes, and diverse DUT and standard instrument types more accurately and reliably, improving the robustness and accuracy of the intelligent analysis.

[0063] Deepening of the intelligent level. The application of the deep learning model makes the event recognition and reading processes more intelligent and automated, capable of handling complex visual tasks that previously required manual intervention or were difficult to solve with traditional algorithms.

[0064] Integrity and scientific nature of the measurement result. The addition of the uncertainty evaluation unit 44 enables the system to not only give the measurement value but also scientifically and comprehensively evaluate and report the uncertainty of the measurement result according to metrological specifications. This uncertainty comprehensively considers the contributions of all key links such as timestamps, AI recognition, interpolation calculation, and cross-validation, making the output result a complete and credible metrological measurement result, greatly enhancing its scientific value and application value.

[0065] In summary, through the deep optimization at the intelligent analysis and precise calculation levels in Example 2, the entire time interval measurement system has been significantly improved in terms of precision, intelligent level, robustness, and the scientific nature of the result.

[0066] Example 3 As Figures 1 to 3As shown, this embodiment further integrates technical features for enhancing the system's automation, robustness, and optimizing core working parameters on the basis of the systems described in Embodiment 1 and Embodiment 2. The overall system architecture still includes a high-speed video acquisition module 10, a synchronization control and time information processing module 20, an intelligent image analysis module 30, and a data processing and verification module 40, and includes the image preprocessing unit 31 described in Embodiment 2, a DUT event feature recognition unit 32 using a deep learning model, a standard instrument vision reading unit 33, a sub-frame level time interpolation calculation unit 34 based on a specific principle, and the data processing and verification module 40 including an uncertainty evaluation unit 44.

[0067] The core enhancement of this embodiment lies in the addition of an adaptive optical control module 50 to the system. This module aims to dynamically optimize the image acquisition process, reduce the dependence on manual fine-tuning, and improve the system's adaptability under different environments and DUTs. The adaptive optical control module 50 specifically includes: Optical state evaluation unit 51: This unit continuously receives real-time image data from the high-speed video acquisition module 10 or the analysis results of image quality (such as clarity, contrast, signal-to-noise ratio) or target position (such as ROI center coordinates, target offset) from the intelligent image analysis module 30. The optical state evaluation unit 51 is configured with corresponding algorithms (for example, a clarity evaluation function based on image gradients, a position detection algorithm based on template matching or feature point tracking) to continuously evaluate whether the current imaging quality meets the preset standard or whether the region of interest (ROI) remains at the ideal observation position.

[0068] Optical component drive unit 52: This unit generates corresponding control instructions according to the evaluation results output by the optical state evaluation unit 51 (for example, the current clarity value, target offset vector). These instructions are sent to the electro-optical components associated with the high-speed video acquisition module 10 (for example, an electric focus motor, an electric zoom controller mounted on the camera lens, or an electric precision pan-tilt head carrying the camera). The optical component drive unit 52 is configured to respond to the evaluation results and precisely control these electro-optical components for adjustment. For example, it drives the focus motor to fine-tune the focal length to maximize the clarity index, or drives the pan-tilt head to make small translation / pitch adjustments to re-center the ROI in the field of view. This adjustment can be performed for at least one high-speed camera in the system.

[0069] Through the closed-loop feedback control of the optical state evaluation unit 51 and the optical component drive unit 52, the adaptive optical control module 50 can achieve automatic focusing and target auto-tracking during the measurement process, ensuring that the system can continuously obtain high-quality and stably aligned image data even when there are minor changes in the DUT position or the initial settings are not perfect.

[0070] In addition, in this embodiment, specific optimization settings are made for the preset time interleaving amount. The synchronization trigger signal generation unit 22 inside the synchronization control and time information processing module 20 is specifically configured such that the time delay between the interleaved trigger signals generated to drive the first and second high-speed cameras is exactly equal to half of the frame period of the high-speed camera itself (Δt / 2). For example, if the camera frame rate is 2000 fps (frame period Δt = 0.5 ms), the trigger moment of the second camera will be exactly 0.25 ms later than that of the first camera.

[0071] Working process: Based on Embodiment 2, the working process of this embodiment mainly incorporates adaptive control in the initialization and acquisition phases and defines the interleaving parameters: Initialization and setting: (Same as Embodiment 2) In this phase, in addition to defining the ROI, the operator also needs to enable the autofocus and / or auto-tracking functions of the adaptive optics control module 50.

[0072] Initial optimization of adaptive optics: Before or at the beginning of formal acquisition, the optical state evaluation unit 51 of the adaptive optics control module 50 analyzes the initial field of view image, and the optical component driving unit 52 controls the electro-optical components to perform necessary focusing and field of view adjustments to ensure the best observation conditions from the start.

[0073] Synchronous data acquisition (optimized parameters): 1. The external signal receiving unit 21 receives the standard electrical signal. 2. The synchronization trigger signal generation unit 22 generates a trigger signal with an accurate time interleaving amount of Δt / 2 based on the standard signal to drive the dual high-speed cameras for acquisition. 3. The high-speed video acquisition module 10 acquires images, and the timestamp generation and embedding unit 23 embeds timestamps. 4. (In parallel) Continuous optimization of adaptive optics: During the entire acquisition process, the adaptive optics control module 50 keeps working. The optical state evaluation unit 51 continuously evaluates the image quality and the target position, and the optical component driving unit 52 fine-tunes the optical components as needed to maintain the best observation state.

[0074] Image preprocessing: (Same as Embodiment 2) Executed by the image preprocessing unit 31.

[0075] AI intelligent analysis: (Same as Embodiment 2) Executed by the DUT event feature recognition unit 32 and the standard instrument vision reading unit 33 using a deep learning model.

[0076] Sub-frame level precise timing: (Same as Embodiment 2) Executed by the sub-frame level time interpolation calculation unit 34.

[0077] Cross-validation: (Same as Embodiment 2) Executed by the dual-source time information cross-validation unit 42.

[0078] Result calculation: (same as in Embodiment 2) Executed by the time interval calculation unit 43.

[0079] Uncertainty evaluation: (same as in Embodiment 2) Executed by the uncertainty evaluation unit 44.

[0080] Output: (same as in Embodiment 2) Output the result and uncertainty.

[0081] Significantly improve the automation level and usability. The introduction of the adaptive optical control module 50 enables the system to automatically complete the focusing and target tracking tasks, greatly reducing the burden on the operator during the initial setup and measurement processes, lowering the requirements for operation skills, and improving the overall usability and automation level of the system.

[0082] Enhance the system robustness and environmental adaptability. The autofocus and tracking functions enable the system to better handle actual working conditions such as slight movement, vibration of the object under test, or changes in environmental light, ensuring that high-quality image data can be continuously obtained even under non-ideal conditions, thereby improving the stability and reliability of the measurement.

[0083] Optimize the time sampling uniformity and efficiency. The time interleaving amount of the dual cameras is precisely set to half of the frame period (Δt / 2), ensuring the most uniform distribution of sampling points for events on the time axis, maximizing the use of the dual-camera configuration to improve the effective time resolution, and having the best effect on capturing and accurately measuring high-speed transient events.

[0084] Guarantee the best performance of the subsequent processing module. By continuously optimizing the imaging quality through the adaptive optical control module 50, it provides the best input data for the DUT event feature recognition unit 32, the standard instrument vision reading unit 33, and the sub-frame level time interpolation calculation unit 34 in the intelligent image analysis module 30, helps these advanced algorithms to exert their maximum efficiency, and indirectly improves the accuracy and reliability of the final measurement.

[0085] In summary, through the introduction of adaptive optical control and optimization of the core acquisition parameters in Embodiment 3, the entire high-precision time interval measurement system is more perfect in terms of automation, usability, robustness, and time resolution optimization, and the comprehensive performance is further improved.

[0086] The above description is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications, and environments, and can be changed within the scope of the concept described herein through the above teachings or the technology or knowledge in related fields. Any changes and variations made by those skilled in the art without departing from the spirit and scope of the present invention shall fall within the protection scope of the appended claims of the present invention.

Claims

1. A time interval measurement system based on a high-speed video collector, characterized in that: include: A high-speed video acquisition module (10) configured to acquire an image sequence including visual start-stop event features of a device under test (DUT) and a visual display of an external standard time interval measuring instrument, wherein the high-speed video acquisition module (10) comprises at least two high-speed cameras; A synchronization control and time information processing module (20) is connected to the high-speed video acquisition module (10), and the synchronization control and time information processing module (20) comprises: an external signal receiving unit (21) configured to receive an electrical signal output by the external standard time interval measuring instrument as a time reference; A synchronization trigger signal generating unit (22) is configured to generate a synchronization trigger signal based on the electrical signal to drive the at least two high-speed cameras to expose and collect images in a synchronous manner with a preset time staggered amount; A timestamp generation and embedding unit (23), configured to generate and embed traceable first electronic timestamp information for each frame of the acquired image sequence based on the electrical signal; The intelligent image analysis module (30) is configured to process the image sequence acquired by the high-speed video acquisition module (10) and analyze it using an artificial intelligence (AI) model, wherein the intelligent image analysis module (30) comprises: A DUT event feature recognition unit (32) configured to analyze the image sequence using the AI ​​model to recognize visual features of the start event and stop event of the DUT; A standard instrument visual reading unit (33) is configured to analyze the image sequence using the AI ​​model or optical character recognition (OCR) technology to identify the visual display time value of the standard time interval measuring instrument, thereby obtaining a second visual time information source; A sub-frame time interpolation calculation unit (34) is configured to calculate the sub-frame precision time when the DUT start / stop event and the visual display time value change occur based on the first electronic time stamp information and the relevant image frame containing the DUT start / stop event and the visual display time value change; A data processing and verification module (40) is configured to communicate with the intelligent image analysis module (30) and the synchronization control and time information processing module (20), and the data processing and verification module (40) comprises: A dual-source time information cross-verification unit (42) is configured to verify the accuracy and consistency of the time measurement link by comparing and analyzing the time based on the first timestamp information and the time based on the second visual time information; The time interval calculation unit (43) is configured to calculate the time interval of the DUT based on the verification result of the cross-verification unit (42) and the sub-frame level precision moments of the start event and the stop event of the DUT.

2. The time interval measurement system based on high-speed video acquisition device according to claim 1, characterized in that: The synchronization trigger signal generated by the synchronization trigger signal generating unit (22) is an interleaved trigger signal used to drive the first camera and the second camera of the at least two high-speed cameras to perform time division multiplexing acquisition at the preset time interleaving amount.

3. The time interval measurement system based on high-speed video acquisition device according to claim 1, characterized in that: The timestamp generation and embedding unit (23) writes the first electronic timestamp information into the metadata of each frame of image, and the resolution of the first electronic timestamp information reaches the microsecond level or nanosecond level.

4. The time interval measurement system based on high-speed video acquisition device according to claim 1, characterized in that: The intelligent image analysis module (30) further comprises an image preprocessing unit (31), which is arranged on a data input path of the DUT event feature recognition unit (32) and the standard instrument visual reading unit (33), and is configured to perform at least one preprocessing operation of denoising, image enhancement or geometric correction on the acquired image sequence.

5. The time interval measurement system based on high-speed video acquisition device according to claim 1, characterized in that: The sub-frame time interpolation calculation unit (34) is configured to calculate the sub-frame precision moment by analyzing at least one of pixel-level displacement information, optical flow vector information, or brightness / grayscale intensity distribution change information of target features related to the event in a plurality of frame images before and after the event occurs.

6. The time interval measurement system based on high-speed video acquisition device according to claim 1, characterized in that: The dual-source time information cross-validation unit (42) is configured to compare the visual time value identified at a specific sub-frame-level precision moment when the standard instrument visual display time value determined by the standard instrument visual reading unit (33) and the sub-frame-level time interpolation calculation unit (34) changes with the theoretical time value represented by the first electronic timestamp information corresponding to the specific sub-frame-level precision moment, so as to perform the cross-validation.

7. The time interval measurement system based on high-speed video acquisition device according to claim 1, characterized in that: Also included is an adaptive optical control module (50), the adaptive optical control module (50) comprising: an optical state evaluation unit (51), configured to evaluate a current imaging quality index or a target position state according to the output of the intelligent image analysis module (30) or directly analyzing the image; An optical component driving unit (52) is configured to respond to an evaluation result of the optical state evaluation unit (51) and control an electric optical component associated with the high-speed video acquisition module (10) to automatically adjust a focus parameter or a field of view orientation of at least one of the at least two high-speed cameras.

8. The time interval measurement system based on high-speed video acquisition device according to claim 1, characterized in that: The AI ​​model used by the DUT event feature recognition unit (32) and the standard instrument visual reading unit (33) is a deep learning model trained for a specific DUT event type and a specific standard instrument display style.

9. The time interval measurement system based on high-speed video acquisition device according to claim 1, characterized in that: The preset time stagger amount is set to half of the frame period of the high-speed camera.

10. The time interval measurement system based on high-speed video acquisition device according to claim 1, characterized in that: The data processing and verification module (40) further includes an uncertainty assessment unit (44) configured to assess the measurement uncertainty of the calculated DUT time interval by comprehensively considering at least part of the following factors: the uncertainty of the first electronic timestamp information, the confidence of the AI ​​model recognition result, the uncertainty introduced by the sub-frame level time interpolation calculation, and the time information consistency deviation obtained by the cross-validation unit (42).

Images