Triggered discrete visual sampling method, apparatus and system

Abstract

The application discloses a kind of trigger discrete visual sampling method, device and system, belong to wearable device and digital twin field.The method includes: real-time monitoring wearer's blink action;The validity discrimination of detected blink action is carried out, and effective blink (such as conscious trigger, or with physiological blink as trigger) is judged according to preset validity determination rule;In response to judging effective blink, trigger image acquisition unit to grab at least one frame discrete image of current time, and discrete image frame is stored and / or transmitted, for visual digital twin construction.The application takes blink action as sampling trigger signal, makes sampling rhythm and human eye visual attention rhythm synchronous, changes continuous video into discrete snapshot, solves the problems of high power consumption, large data redundancy, acquisition visual angle and real perception mismatch of existing wearable device, and can be widely applied to daily life, industrial inspection, driving / flying and other scenes.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence, wearable devices, data acquisition and digital twin technology. Specifically, it relates to a method, apparatus and system for triggering discrete sampling of wearable image acquisition devices through eye movements (especially blinking) to efficiently construct visual digital twins oriented towards the real human eye perspective. Background Technology

[0002] 2.1 Overview of Existing Technology

[0003] Digital twin technology aims to reconstruct physical entities, processes, and scenes in virtual space using digital means. Visual digital twins, as an important branch of digital twins, rely on image / video acquisition devices to continuously acquire visual data of the real environment to support higher-level applications such as 3D reconstruction, scene understanding, and remote collaboration.

[0004] Currently, smart terminal devices used for visual data acquisition—including smart glasses, action cameras, dashcams, and personal recorders—generally adopt a "continuous sampling" working mode, which means recording video streams continuously at a fixed frame rate (such as 30fps or 60fps).

[0005] 2.2 Deficiencies of the Existing Technology

[0006] The above continuous sampling scheme has the following prominent problems in practical applications:

[0007] (1) High power consumption. Continuous video recording keeps the image sensor, encoding chip, storage module and communication module under high load for a long time, resulting in serious heat generation and a significant reduction in battery life. For smart glasses products with extremely limited battery capacity, the battery life for continuous recording is often less than 1 hour, which seriously restricts the feasibility of all-weather wearable applications.

[0008] (2) High data redundancy. Human visual attention is highly selective—only the moment that is “attention” by the brain has recording value. However, continuous sampling records everything at a uniform frame rate, resulting in a large number of meaningless redundant frames (such as a large number of similar road scenes during walking), which leads to a waste of storage resources and a surge in the manual cost of data screening in the later stage.

[0009] (3) The acquisition perspective does not match the actual visual perception of people. Existing devices focus more on "capturing" rather than "seeing". The acquired images may not be consistent with the direction and time that the human eye is actually focusing on, making it difficult to truly reproduce what the wearer sees.

[0010] (4) High bandwidth pressure for data upload and cloud processing. Massive amounts of continuous video data require high bandwidth transmission, which is especially limited in mobile scenarios.

[0011] Therefore, there is an urgent need for a visual sampling scheme that is synchronized with the rhythm of human visual perception, has low power consumption, and low redundancy. Summary of the Invention

[0012] 3.1 Purpose of the Invention

[0013] The present invention aims to solve the problems of high power consumption, large data redundancy, and mismatch between the acquisition perspective and human real visual perception caused by continuous sampling in existing wearable visual acquisition devices, and provides a trigger-based discrete visual sampling method, device and system.

[0014] 3.2 Technical Solution

[0015] To achieve the above objectives, the present invention provides a triggered discrete image sampling method, comprising the following steps:

[0016] - Monitoring steps: Real-time monitoring of the wearer's blinking movements;

[0017] - Identification step: The validity of the detected blinking action is identified, and a valid blink is determined from the blinking action according to the preset validity judgment rules;

[0018] - Sampling step: In response to the determination of a valid blink, the image acquisition unit is triggered to capture at least one discrete image frame at the current moment.

[0019] This invention uses the wearer's blinking action as a sampling trigger signal, enabling the image acquisition unit to perform discrete frame sampling only during effective blinking moments. This achieves synchronization between the sampling rhythm and the rhythm of human visual attention, thereby transforming continuous video stream recording into triggering single-frame / few-frame captures only at meaningful moments. This significantly reduces power consumption, eliminates redundant data, and faithfully reproduces what the wearer sees.

[0020] The preset validity determination rules include at least the following two implementation methods, which can be configured or switched based on the application scenario:

[0021] (I) Physiological Trigger (Active Sampling Mode): All detected blinks are considered valid blinks, with the wearer's natural physiological blinking rhythm (on average once every 2-10 seconds, approximately 15-20 times per minute) serving as the sampling beat. Research shows that blinking and attention shifting are synchronous in neural mechanisms, and blinking moments often correspond to the cognitive process of "attention-memory." Therefore, discrete sampling with physiological blinking as the beat can also capture meaningful visual moments, and the sampling density is naturally consistent with the rhythm of human visual attention.

[0022] (ii) Conscious Triggering: This means that conscious blinking is judged as valid blinking; for example, conscious blinking caused by visual attention switching or visual memory encoding is distinguished from unconscious physiological blinking, and only conscious blinking that meets the preset validity conditions is judged as valid blinking. This method uses the wearer's active visual intention as the sampling basis and is suitable for scenarios that require precise control of the sampling timing (such as power saving mode, safety-critical scenarios).

[0023] The two implementation methods described above are merely examples. The preset validity determination rules are not limited to these and may also include determination rules for specific blinking patterns such as intentional double blinking and prolonged eye closure.

[0024] The following steps and their specific implementation methods (such as field-of-view configuration, blink detection method, validity identification rules, spatiotemporal metadata, cache management, digital twin construction, etc.) are all preferred embodiments of the present invention and do not constitute a limitation on the present invention. Those skilled in the art will understand that the specific implementation methods of the above components and method steps are merely preferred embodiments and do not constitute a limitation on the present invention. For example: the wearable device includes, but is not limited to, smart glasses, smart helmets, augmented reality head-mounted displays, and other head-mounted electronic devices; the sensor used for blink detection is not limited to microelectromechanical sensors, but can be any sensor capable of detecting physiological signals around the wearer's eyes; the image sensor is not limited to a miniature camera; the posture information can be obtained by a posture sensor, and the posture sensor is not limited to an inertial measurement unit; the communication module can be any wireless communication module. Those skilled in the art can make equivalent substitutions or modifications to the above preferred embodiments according to actual application scenarios, all of which fall within the protection scope of the present invention.

[0025] Step 1: Image Acquisition Preparation. A wearable image acquisition device is provided, wherein the field of view of its image acquisition unit is configured to match the field of view of the human eye or be located in the blind spot of the human eye (such as the side or rear) to acquire images that match or complement the environment seen by the human eye. The wearable image acquisition device includes, but is not limited to, smart glasses, smart helmets, AR / VR headsets, etc.

[0026] Step 2: Blink Detection. The wearer's blinking movements are monitored in real time using a blink detection unit. The blink detection unit can be implemented using at least one of the following methods:

[0027] - Physical sensing-driven detection (2a): Microelectromechanical sensors (including but not limited to pressure sensors, electromyography sensors, accelerometers, etc.) are placed at the site where the wearable device contacts the skin around the eyes. Blinking actions are identified by detecting skin deformation / pressure changes caused by the contraction of the orbicularis oculi muscle. When the intensity of the detected mechanical signal exceeds a preset threshold I and the duration is within the blink characteristic duration range, it is determined as a blinking action.

[0028] - Pattern Recognition-Driven Detection (2b): A miniature camera facing the wearer's eyes is placed inside the wearable device to identify blinks by analyzing the opening and closing state of the eyelids in real time. A lightweight neural network model is used to infer the eye image sequence. When a complete cycle of the eyelid opening, closing and opening is detected, and the duration of eye closure T is within a preset range, it is determined as a blink.

[0029] - Hybrid detection (2c): Combine physical sensing signals and pattern recognition signals for joint judgment. When both indicate a blinking event, sampling is triggered to reduce the probability of false triggering.

[0030] Step 3: Blink Validity Identification. To avoid oversampling caused by natural, unconscious blinking (humans blink an average of once every 2-10 seconds), this method further includes a blink type identification sub-step:

[0031] - Compare the time interval between two consecutive blinks with a preset threshold. If the interval is less than the first threshold (e.g., 0.5 seconds), it is determined as an "intentional double blink" and sampling is triggered.

[0032] - Compare the duration of eye closure during a single blink with a baseline threshold. If the duration of eye closure exceeds the second threshold (e.g., 500ms), it is judged as a "long eye closure" and sampling is triggered.

[0033] In one preset configuration (such as active sampling mode), all blinks are considered valid triggers; in another preset configuration (such as power saving mode), only double blinks and long eye closures trigger sampling, or sampling is triggered once for every preset number of blinks. Other optional preset configurations include single eye blinking, alternating blinking between the left and right eyes, and specific blinking patterns (such as "blink-pause-blink").

[0034] Step 4: Respond to a valid blink by triggering discrete sampling. When a valid blink is determined in Step 3, the image acquisition unit is immediately triggered to capture K frames of images at the current moment (K≥1, default K=1, but can also be configured to capture 3-5 frames in succession to capture dynamic scenes). The trigger delay should be controlled within N ms (e.g., N=50) to ensure that the acquired images are highly synchronized with the wearer's visual perception at the moment of blinking.

[0035] Step 5: Image Data Storage and Transmission. Acquired image frames, along with timestamps, GPS location information, and device attitude information (attitude information can be obtained from attitude sensors, not limited to IMU inertial measurement units), are stored in the device's local storage unit. Simultaneously, data can be transmitted in real-time or in batches to associated devices (such as smartphones, tablets, or cloud servers) via communication modules (such as Bluetooth, Wi-Fi, UWB, or cellular networks).

[0036] - Step 6: Cache Management and Overwrite Strategy. When the local storage space is lower than the preset threshold, the cache management strategy is activated: the earliest sampled data is automatically deleted in chronological order to make room for new samples (first-in, first-out overwrite strategy); or based on content analysis, high-value frames containing text, faces, or specific objects are retained first, while low-value frames that are blurry or have duplicate content are deleted.

[0037] - Step 7 (Optional): Digital Twin Construction. The discrete image frame set (with spatiotemporal metadata) acquired in Step 5 is input into a 3D reconstruction engine or neural rendering model to generate a visual digital twin scene from the wearer's first-person perspective. Specifically, the cloud server 109 receives discrete image frames and their associated spatiotemporal metadata uploaded by the wearable device via the associated terminal device 108. Using the spatiotemporal metadata as a spatial registration reference, the scene is reconstructed using a 3D reconstruction engine or neural rendering model (such as NeRF or 3DGS). When new discrete image frames arrive, an incremental update strategy can be adopted, only locally reconstructing and updating the spatial regions in the scene that have changed. The visual digital twin platform 501 can be used for at least one application such as scene retrospection, remote collaboration, operation training, accident analysis, or behavior auditing. Optionally, the visual digital twin platform 501 can also be generated locally on the associated terminal device 108 or on the wearable device itself 100, and is not limited to the cloud server 109.

[0038] 3.3 Beneficial Effects

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] (1) Significantly reduced power consumption. By transforming continuous video streaming recording into capturing single frames / a small number of frames only at meaningful moments, the duty cycle of the image sensor, encoder, and storage module is significantly reduced. Taking a normal person blinking 15-20 times per minute and triggering 1 frame each time as an example, compared to continuous recording at 30fps (1800 frames per minute), the amount of data generated is reduced by more than 95%, and the overall battery life is expected to be improved by 5-10 times. The improvement is even greater when intentional triggering is used.

[0041] (2) Significantly reduced data redundancy. The sampling rhythm is naturally synchronized with the rhythm of human visual attention—the blink of an eye often completes the cognitive process of "attention-memory." Therefore, whether consciously or physiologically triggered, each sampling corresponds to a meaningful visual moment. There is no need to sift through massive amounts of video clips afterward. Based on an estimated 8-hour continuous operation, the storage capacity of 30fps continuous recording is approximately 10GB, while the storage capacity of the discrete sampling mode of the physiological triggering mode of this invention is less than 1GB, reducing storage usage by more than an order of magnitude; when using conscious triggering, it can be further reduced by about an order of magnitude.

[0042] (3) Realistically reproduce what the human eye sees. The sampling trigger time is synchronized with the wearer's blinking action, so that the collected data faithfully reflects the wearer's real vision at the moment of visual attention, providing a high-quality "first-person visual log" for digital twins; under the preferred implementation method of matching the field of view of the image acquisition unit with the field of view of the human eye, the collected data can be further consistent with the wearer's real visual experience.

[0043] (4) Strong scene adaptability. Triggering strategies and validity judgment rules can be flexibly configured for different scenarios—power-saving mode for daily wear (only triggered by conscious actions such as double blinking), active mode for workshop inspection (recording with physiological blinking as the beat, or combined with voice commands for compound triggering), and long-closed-eye suppression and double-blink confirmation mechanism for driving / flight scenarios. (5) Communication bandwidth friendly. The data volume of discrete frame images is much smaller than that of continuous video streams, which can be quickly uploaded through low-bandwidth channels, making it suitable for mobile and edge computing scenarios.

[0044] --- Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the overall system architecture provided in an embodiment of the present invention;

[0046] Figure 2 This is a schematic diagram of data flow and digital twin construction provided in an embodiment of the present invention;

[0047] Figure 3 The diagram illustrates the effectiveness identification provided in this embodiment of the invention, showing (a) physiological triggering and (b) conscious triggering.

[0048] Figure 4 The diagram illustrates the implementation of the blink detection unit provided in this embodiment of the invention, including (a) physical sensing-driven detection and (b) pattern recognition-driven detection.

[0049] --- Detailed Implementation

[0050] like Figure 1 As shown, the wearable device body 100 provided in this embodiment of the invention includes a blink detection unit 101, an effectiveness identification unit 102, a main control unit 103, an image acquisition unit 104, a positioning module 105, a storage unit 106, and a communication module 107. The wearable device 100 is communicatively connected to an associated terminal device 108 through the communication module 107, and the communication module 107 or the associated terminal device 108 uploads data to a cloud server 109.

[0051] Example 1: Smart Glasses Form – Everyday Life Scenarios

[0052] This embodiment provides a smart glasses system with integrated blink-triggered discrete sampling function, comprising the following components:

[0053] Hardware components:

[0054] - The main body of the eyeglass frame;

[0055] - Front-facing camera, matching the binocular field of view of the human eye;

[0056] - Eye movement / blink detection module: An EMG electromyography sensor (physical sensing solution) can be set on the nose pad or the inside of the temple, or a miniature infrared camera (pattern recognition solution) can be set on the inside of the frame.

[0057] - Inertial Measurement Unit (IMU) (accelerometer + gyroscope);

[0058] - GPS / BeiDou positioning module;

[0059] - Bluetooth Low Energy 5.3 + Wi-Fi 7 communication module;

[0060] - Local storage unit;

[0061] -Power supply unit;

[0062] - Main control chip, used for blink detection inference.

[0063] Work process as follows Figure 2 As shown:

[0064] 1. After the user puts on the glasses, the system enters a low-power standby state, and the blink detection module polls the sensor at a low frequency (such as 10Hz).

[0065] 2. When the user blinks, the main controller wakes up the image acquisition unit 104 and takes a picture.

[0066] 3. Image frames and data such as timestamps, GPS coordinates, and IMU attitude collected by the positioning module 105 are packaged and cached in the local storage unit 106.

[0067] 4. The stored data is connected to the terminal device 108 (mobile app) via the communication module 107 or directly uploaded to the cloud server 109 to generate a digital twin platform 501 (e.g., inputting a set of discrete image frames into a NeRF / 3D Gaussian Splatting model to reconstruct a visual digital twin scene from the wearer's first-person perspective). During the reconstruction process, the spatiotemporal metadata is used as a spatial registration reference; when a new discrete image frame arrives, incremental local reconstruction is performed only on the spatial regions in the scene that have changed, in order to reduce computational overhead.

[0068] 5. When the storage utilization rate of local storage unit 106 exceeds the preset value, perform cache management 401: automatically delete the earliest discrete image frames in the order of acquisition time, and / or calculate the content value score of each stored image frame based on image content analysis, prioritize deleting image frames with blurry content, duplicate content, or content value score lower than the preset low value threshold, and retain image frames with clear content, unique content, or content value score higher than the preset high value threshold.

[0069] Example 2: Workshop Inspection Scenario

[0070] This embodiment is designed for factory workers, and the hardware can be smart glasses or integrated into a helmet. The key difference lies in the triggering strategy:

[0071] - Employs "active sampling mode": sampling is triggered with each natural blink of the wearer (approximately 15-20 frames per minute); such as Figure 3 As shown in (a), after the blink detection unit 101 detects the blink action 201 in this embodiment, it does not distinguish between conscious and unconscious blinking, but directly triggers the image acquisition unit 104 to perform K-frame sampling 204 (K≥1).

[0072] - Additional "voice command compound trigger": When the wearer says a preset keyword (such as "record" or "abnormal"), it triggers a burst of K-frames in combination with blinking and automatically marks it as "key focus";

[0073] - The collected data is uploaded in real time to the workshop digital twin platform 501 via the communication module 107 for real-time monitoring and post-event traceability.

[0074] In this embodiment, since all natural blinks directly trigger sampling, there is no need for blink type identification. This results in the lowest computational overhead for blink detection and the lowest overall power consumption among all embodiments, making it suitable for scenarios requiring long-term continuous recording of the work process. Based on 15–20 natural blinks per minute, approximately 900–1200 frames of data are generated per hour, a reduction of about 99% compared to continuous recording at 30fps (108,000 frames per hour). When local storage space is insufficient, the earliest frames are overwritten in chronological order to ensure the integrity of the most recent work process recordings.

[0075] Example 3: Driving / Flight Scenario

[0076] Visual sampling requirements for drivers or pilots:

[0077] - Triggering strategy: Employs a "long-term eye-closing inhibition + double blink confirmation" mechanism. For example... Figure 3As shown in (b), after the blink detection unit 101 detects the blink action 201 in this embodiment, the validity identification unit 102 determines it as intentional double blink 202 according to the preset validity judgment rule, and triggers the image acquisition unit 104 to perform K-frame sampling 204; since it is necessary to be vigilant about long eye closure caused by fatigue driving in the driving scenario, this embodiment regards long eye closure 203 as invalid trigger and issues an alarm 205.

[0078] - Sampling field of view: In addition to the front camera, a side and rear camera is added to cover blind spots of human vision (such as the side and rear of the vehicle), and multiple images are captured simultaneously each time it is triggered.

[0079] - Data usage: To build a "visual digital twin" of the entire driving / flight process for accident retrospective, driving behavior analysis, and training review.

[0080] In this embodiment, the preset time interval threshold is set to 0.5 seconds, and the duration threshold for closed eyes is set to 500 milliseconds: when the time interval between two consecutive blinks of the wearer is less than 0.5 seconds, it is determined to be intentional double blinking and sampling is triggered; when a single eye closure lasts for more than 500 milliseconds, it is determined to be long eye closure, sampling is not triggered, and acoustic and tactile alarms are triggered. The trigger delay between the determination of a valid blink and the start of exposure of the image acquisition unit is controlled within 50 milliseconds to ensure that the capture time is synchronized with the visual attention time; when it is necessary to record emergency or key scenes, it can be configured to capture 3 frames in a single valid blink (K=3) to capture dynamic information.

[0081] Example 4: Hybrid Scheme for Blink Detection

[0082] To balance detection accuracy and power consumption, this embodiment employs a hybrid approach combining physical sensing and pattern recognition, such as... Figure 4 As shown:

[0083] 1. The physical sensor 301 (EMG / pressure) continuously monitors with ultra-low power consumption (<1mW) to detect skin deformation / pressure changes caused by orbicularis oculi muscle contraction to identify blinking movements. The sensor described above is not limited to microelectromechanical systems (MEMS) types and can be any physiological signal sensor.

[0084] 2. When the intensity of the detected mechanical signal exceeds the preset threshold I, the wearable device 100 is activated by a miniature camera 302 located inside the wearer's eye to perform image-level pattern recognition and blink confirmation.

[0085] 3. If the duration T1 for which the signal strength detected by the physical sensor 301 exceeds the preset threshold I is within the preset range, and the duration T2 for the eyelid to open, close, and then open again detected by the camera 302 is also within the preset range, it is determined as a blinking action.

[0086] In this embodiment, the first stage uses a low-power mechanical sensor for constant-on monitoring, with an average power consumption of less than 1 milliwatt. When the mechanical signal meets the initial blinking condition, a wake-up signal is generated, and the second-stage image sensor switches from sleep mode to working mode to perform image-level blink confirmation. After confirmation, it quickly returns to sleep mode. Since the image sensor only works briefly during the suspected blinking period, its duty cycle is less than 5%, and the overall detection power consumption is reduced by more than an order of magnitude compared to the scheme that only uses an image sensor. At the same time, through the joint confirmation of the two-stage signals, interference such as head shaking and muscle twitching can be effectively eliminated, and the false trigger rate is significantly reduced.

[0087] Example 6: Outdoor Sports / Smart Helmet Scenario

[0088] This embodiment is designed for outdoor sports (such as cycling, hiking, and skiing) recording scenarios, and the hardware form is a smart helmet with integrated blink-triggered discrete sampling function:

[0089] Hardware components: The helmet body 100 integrates a front-facing camera 104, a side and rear blind spot camera, a blink detection unit 101, a positioning module 105 (GPS / BeiDou), a storage unit 106, and a communication module 107; the optical axis of the front-facing camera is consistent with the wearer's line of sight, and the field of view covers the wearer's binocular field of vision.

[0090] Triggering strategy: By default, physiological triggering (active sampling mode) is used to record the entire movement process; when the wearer needs to mark key moments (such as dangerous sections of road or exciting actions), conscious triggering (intentional double blinking) is used to take K frames in succession (such as K=3) and add a "key focus" mark;

[0091] Motion data linkage trigger: Optionally, blink trigger can be linked with motion status data (such as heart rate, speed, altitude, and slope). When the motion status data meets preset conditions (such as heart rate exceeding a preset heart rate threshold), the subsequently detected valid blinks will be automatically marked as "high-value events".

[0092] Field of vision compensation: The side and rear blind spot camera captures images synchronously with the front camera each time it is triggered, compensating for the blind spot of the human eye and creating a more complete record of the motion scene;

[0093] Data usage: After the motion ends, discrete image frames and spatiotemporal metadata are uploaded to the associated terminal device 108 or cloud server 109 via communication module 107 to generate a first-person visual digital twin platform 501 corresponding to the motion trajectory, which is used for motion review, route sharing or safety incident tracing.

Claims

1. A trigger-based discrete image sampling method, applied to a wearable device, characterized in that, include: Real-time monitoring of the wearer's blinking movements; The validity of detected blinking actions is identified, and valid blinks are determined from the blinking actions according to preset validity judgment rules; In response to a valid blink, an image acquisition unit is triggered to capture at least one discrete image frame at the current moment.

2. The method according to claim 1, characterized in that, The preset validity determination rule is as follows: conscious blinking is determined as valid blinking. Conscious blinking includes, but is not limited to, intentional double blinking where the time interval between two consecutive blinks is less than a preset time interval threshold, and / or long blinking where the duration of eye closure exceeds a preset duration threshold. Blinking includes single-eye blinking, double-eye blinking, alternating blinking of the left and right eyes, or blinking sequences with a preset timing pattern.

3. The method according to claim 1, characterized in that, The preset validity determination rule is as follows: all detected blinking actions are determined to be valid blinks, and the wearable device performs discrete sampling based on the wearer's physiological natural blinking rhythm.

4. The method according to claim 1, characterized in that, Also includes: The captured discrete image frames are saved to the local storage unit, and / or the captured discrete image frames are transmitted to the associated terminal device or the cloud; and when the remaining space of the local storage unit is lower than a preset threshold, the earliest discrete image frames are automatically deleted in the order of acquisition time, and / or the content value score of each stored image frame is calculated based on image content analysis, and image frames with blurry content, duplicate content, or content value score lower than a preset low value threshold are deleted first, while image frames with clear content, unique content, or content value score higher than a preset high value threshold are retained.

5. The method according to claim 1, characterized in that, The blinking action detection adopts a graded wake-up and joint confirmation method: the low-power sensor is always on to monitor at the first power level. When the detected signal meets the preset preliminary blinking conditions, a wake-up signal is generated. In response to the wake-up signal, the image sensor switches from the sleep state to the working state to perform image-level blink confirmation. Only when both the preliminary judgment of the low-power sensor and the image-level blink confirmation of the image sensor indicate that a blinking action exists is it determined to be a valid blink. The image sensor remains in the sleep state during the non-confirmation stage.

6. The method according to claim 1, characterized in that, The image acquisition unit is an environmental image acquisition unit, whose optical axis direction is consistent with the wearer's line of sight or within a preset angle deviation range, so that the acquired environmental image matches the wearer's visual perception content; the field of view of the environmental image acquisition unit covers the wearer's binocular field of vision, or the environmental image acquisition unit includes multiple cameras, respectively covering the wearer's frontal and lateral rearal fields of vision to compensate for blind spots of human vision; the wearable device is a head-mounted electronic device.

7. The method according to claim 1, characterized in that, Also includes: Spatiotemporal metadata is associated with each discrete image frame. The spatiotemporal metadata includes at least one of the following: the timestamp of the image acquisition time, the geographical location information of the acquisition time, and the attitude information of the wearable device at the acquisition time. The attitude information is acquired by an attitude sensor and includes at least one of pitch angle, yaw angle, and roll angle. The discrete image frames and their associated spatiotemporal metadata are input into a 3D reconstruction engine or neural rendering model to generate a visual digital twin scene from the wearer's first-person perspective.

8. The method according to claim 1, characterized in that, The triggering includes: the triggering delay between the determination of a valid blink and the start of exposure of the image acquisition unit does not exceed a preset triggering delay threshold; the number of discrete image frames K captured is a configurable parameter, K≥1, and when K>1, K frames are captured quickly within a preset time window at a burst shooting rate higher than the normal video recording frame rate of the wearable device; it also includes: detecting the wearer's voice command, and when the voice command contains preset keywords, in response to determining a valid blink, triggering the capture of K frames, and automatically marking the captured K frames as a preset attention category.

9. A trigger-based discrete visual sampling device, characterized in that, include: A wearable device; A blink detection unit is installed on the wearable device to monitor the wearer's blinking movements in real time; A validity identification unit is used to identify the validity of detected blinking actions and determine valid blinks from the blinking actions according to preset validity determination rules; An image acquisition unit, installed on the wearable body, is used to capture at least one discrete image frame at the current moment in response to a determination of a valid blink. A storage unit and / or communication module for storing and / or transmitting captured discrete image frames; A main control unit is electrically connected to the blink detection unit, the validity identification unit, the image acquisition unit, the storage unit, and / or the communication module.

10. A trigger-based discrete visual sampling system, characterized in that, include: The discrete visual sampling device according to claim 9; An associated terminal device is connected to the discrete visual sampling device via a communication module, and is used to receive, store, display the discrete image frames and / or upload the discrete image frames to a cloud server; A cloud server is used to receive the discrete image frames and construct a visual digital twin scene from the wearer's first-person perspective based on the discrete image frames.

11. 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 method of any one of claims 1 to 8.

Images