Diving goggles based on multi-source data fusion algorithm and adaptive navigation method thereof

Abstract

This invention proposes a diving goggle based on a multi-source data fusion algorithm and its adaptive navigation method. It belongs to the field of intelligent wearable devices and underwater positioning and navigation technology. The method includes: acquiring IMU-depth-geomagnetic multi-source sensor data of the diving environment to generate an initial positioning dataset; performing fusion positioning calculations based on the initial positioning dataset to construct a semantic topology map in a GPS-free environment; performing ViT semantic segmentation processing on the semantic topology map to identify and mark key underwater landmarks and dangerous areas, generating a semantic navigation layer; simultaneously, collecting the user's blink signals through the eye-tracking module built into the diving goggle to achieve hands-free target marking and command input, updating the user's focus points in the semantic navigation layer; through the multi-source data fusion algorithm, this diving goggle adaptive navigation method significantly improves the accuracy and stability of underwater positioning, eliminates dependence on external beacons, and achieves centimeter-level positioning in a GPS-free environment.

Description

Technical Field

[0001] This invention proposes a diving goggles based on a multi-source data fusion algorithm and its adaptive navigation method, belonging to the field of intelligent wearable devices and underwater positioning and navigation technology. Background Technology

[0002] Currently, underwater navigation technology faces multiple challenges. Traditional methods mainly rely on external beacons for positioning. For example, the Ultra Short Baseline (USBL) system requires surface base stations, while inertial regression (DR) technology suffers from large cumulative errors, resulting in inaccurate positioning and numerous limitations during underwater operations.

[0003] Meanwhile, existing technologies have significant shortcomings in underwater environmental perception, failing to effectively identify and utilize underwater natural landmarks for navigation. They can only record simple geometric trajectories, lacking a deep understanding of the underwater environment and semantic context. Furthermore, interaction methods during diving are extremely limited; manual operation is very inconvenient underwater, and there is a lack of hands-free target marking and command input methods. More importantly, existing navigation strategies are static and simplistic, often merely reversing along the original path during return, failing to fully consider ocean current changes, obstacle distribution, and the diver's physiological state. Against this backdrop, there is an urgent need for a novel adaptive navigation method for diving goggles that can deeply integrate multi-source data and achieve underwater semantic understanding and physiological adaptive decision-making. This would break through the limitations of traditional navigation technologies and provide divers with a safer, more accurate, and intelligent navigation experience. Summary of the Invention

[0004] This invention provides diving goggles based on a multi-source data fusion algorithm and an adaptive navigation method thereof to solve the problems mentioned in the background section above: The present invention proposes an adaptive navigation method for diving goggles based on a multi-source data fusion algorithm, the method comprising: S1. Collect IMU-depth-Geomagnetic multi-source sensor data in the diving environment to generate an initial positioning dataset; perform fusion positioning calculation based on the initial positioning dataset to construct a semantic topology map in a GPS-free environment; S2. Perform ViT semantic segmentation based on the semantic topology map to identify and mark key underwater landmarks and dangerous areas, and generate a semantic navigation layer; at the same time, collect the user's blink signals through the eye-tracking module built into the diving goggles to realize hands-free target marking and command input, and update the user's focus points in the semantic navigation layer. S3. Based on the semantic navigation layer and user physiological status monitoring data, perform adaptive return path planning to generate a dynamic return route that takes into account ocean currents, obstacles and user physiological status. S4. Continuously collect safety parameters through the environmental perception module built into the diving goggles to generate real-time safety warning data; combine the dangerous area markings in the semantic navigation layer to perform dual verification and warning of potential safety risks; S5. Based on real-time safety warning data and the execution status of dynamic return routes, a weighted navigation stability assessment is performed to generate a diving navigation stability index; the navigation guidance strategy is dynamically adjusted based on this index to generate the final intelligent diving navigation data.

[0005] The present invention proposes a diving goggle based on a multi-source data fusion algorithm, the diving goggle comprising: One or more processors; Memory, used to store one or more programs; Wherein, when the one or more programs are executed by the one or more processors, the one or more processors are made to implement the method described in any one of the above.

[0006] The beneficial effects of this invention are as follows: Through a multi-source data fusion algorithm, this adaptive navigation method for diving goggles significantly improves the accuracy and stability of underwater positioning, eliminating reliance on external beacons and achieving centimeter-level positioning in GPS-free environments. Simultaneously, this method enhances underwater environmental perception, enabling the identification and utilization of underwater natural landmarks for semantic navigation, providing divers with richer environmental information. This reduces navigation errors caused by a lack of environmental awareness, avoiding safety hazards during diving. Furthermore, through eye-interaction technology, divers can hands-free mark targets and input commands, greatly improving the convenience of underwater operations. It can adaptively adjust the return route based on the diver's physiological state to ensure diving safety, and provides intuitive navigation through AR optical flow guidance, enhancing the immersive and engaging diving experience. Overall, this method avoids many limitations of traditional navigation technologies, providing divers with a more intelligent, safe, and efficient navigation solution. Attached Figure Description

[0007] Figure 1 This is a diagram illustrating the steps of the method described in this invention. Detailed Implementation

[0008] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0009] One embodiment of the present invention, such as Figure 1 As shown, an adaptive navigation method for diving goggles based on a multi-source data fusion algorithm is described, the method comprising: S1. Collect IMU-depth-Geomagnetic multi-source sensor data in the diving environment to generate an initial positioning dataset; perform fusion positioning calculation based on the initial positioning dataset to construct a semantic topology map in a GPS-free environment. This map contains semantic information and spatial coordinates of underwater natural landmarks (such as coral reefs and rock textures). S2. Perform ViT (Visual Transformer) semantic segmentation processing based on the semantic topology map to identify and mark key underwater landmarks and dangerous areas, and generate a semantic navigation layer; at the same time, collect the user's blink signals through the eye-tracking module built into the diving goggles to realize hands-free target marking and command input, and update the user's focus points in the semantic navigation layer. S3. Based on the semantic navigation layer and user physiological status monitoring data (including heart rate, blood oxygen saturation, and dive duration), adaptive return path planning is performed to generate a dynamic return route that takes into account ocean currents, obstacles, and user physiological status. This route is projected into the field of view of the diving goggles in real time through AR optical flow guidance technology to provide intuitive navigation guidance. S4. Continuously collect safety parameters through the environmental perception module built into the diving goggles, including water temperature, water depth, and oxygen cylinder content, and generate real-time safety warning data; combine with the danger zone markings in the semantic navigation layer to perform dual verification and warning of potential safety risks, ensuring the safety of the diving process; S5. Based on real-time safety warning data and the execution status of dynamic return routes, perform a weighted navigation stability assessment to generate a diving navigation stability index; dynamically adjust navigation guidance strategies (such as optical flow velocity and voice prompt frequency) according to the index to achieve adaptive optimization of the navigation system and generate the final intelligent diving navigation data.

[0010] The working principle and effects of the above technical solution are as follows: By fusing multi-source data to construct an underwater semantic topology map, combined with AR optical flow guidance, navigation accuracy in GPS-free environments is improved, the intuitiveness of return navigation is enhanced, and the possibility of getting lost while diving is reduced. Multi-dimensional safety parameter collection and dual early warning enhance diving safety, avoiding dangers caused by abnormal water temperature, insufficient oxygen, and other hazards, thus reducing the probability of accidents. Hands-free blink control accommodates both shooting and command input, reducing the tedium of manual operation and improving ease of use. AI recognition and motion guidance functions help in recognizing marine life, optimize diving posture, and record motion data, enriching the diving experience. The navigation strategy dynamically adjusts according to the stability index, enhancing system adaptability and reducing navigation difficulties caused by environmental or physiological changes, making diving safer and smoother.

[0011] In one embodiment of the present invention, S1 includes: S11. Activate the built-in IMU sensor, depth sensor, and geomagnetic sensor in the diving goggles to collect comprehensive data on the diving environment and generate multi-dimensional raw environmental data. S12. Receive multi-dimensional raw environmental data, perform noise filtering and data calibration, remove abnormal data points, and generate a clean initial positioning dataset. S13. Extract the feature information corresponding to underwater natural landmarks in the clean initial positioning dataset, analyze the semantic attributes of landmarks such as coral reefs and rock textures, and simultaneously record the spatial location data of various landmarks. S14. Use a multi-source data fusion algorithm to process the initial clean positioning dataset, integrate the advantages of different sensor data, and generate accurate fused positioning data. S15. Based on the precise fusion of positioning data and landmark semantics and spatial location data, a semantic topology map is built in a GPS-free environment to fully present the semantic relationships and spatial distribution of the underwater environment.

[0012] The working principle and effects of the above technical solution are as follows: By collecting underwater environmental data from multiple sensors in a comprehensive manner, coupled with noise filtering and calibration processing, the cleanliness of the initial positioning data is improved, reducing the interference of abnormal data on subsequent processing. Natural landmark features are extracted and their spatial locations are recorded. Combining the advantages of multiple-source data fusion algorithms enhances positioning accuracy in GPS-free environments and reduces the impact of errors from single-sensor data. The constructed semantic topology map fully presents the semantic relationships and spatial distribution of the underwater environment, avoiding the problem of underwater positioning failure. This provides a reliable data foundation for subsequent navigation and allows users to clearly perceive the layout of the underwater environment, reducing the risk of getting lost and making diving positioning more stable and accurate.

[0013] In one embodiment of the present invention, step S14 includes: Receive the clean initial positioning dataset, separate the positioning data categories corresponding to the IMU sensor, depth sensor, and geomagnetic sensor, and generate a classified sensor positioning dataset. Perform data consistency verification on the classification sensor localization dataset to eliminate the temporal deviation of data from different sensors and generate a time-synchronized localization dataset. The multi-source data fusion algorithm is invoked to perform weight allocation calculations on the time-series synchronous positioning dataset, highlighting the influence of high-confidence sensor data and generating weighted fusion intermediate data; A second screening for outliers is performed on the weighted fusion intermediate data to eliminate biased data generated during the fusion process and generate fusion calibration positioning data. Integrate and fuse calibration positioning data, optimize data accuracy thresholds, and generate accurate fused positioning data.

[0014] The working principle and effects of the above technical solution are as follows: By separating the positioning data categories corresponding to different sensors, interference caused by data confusion is reduced, and the clarity of data classification is improved. Consistency verification of the classified data eliminates temporal deviations and avoids fusion errors caused by asynchronous data from different sensors. Weight allocation calculations highlight the influence of high-reliability data, enhancing the reliability of fused data and reducing deviations caused by the limitations of a single data source. Secondary screening for outliers eliminates biased data generated during the fusion process, further improving data purity. Integrating calibration data and optimizing accuracy thresholds generates accurate fused positioning data, which not only provides reliable data support for subsequent semantic topology map construction but also improves positioning accuracy in GPS-free environments, avoids navigation problems caused by positioning deviations, and makes underwater positioning more stable and accurate.

[0015] In one embodiment of the present invention, S2 includes: S21. Retrieve the constructed semantic topology map, and start the ViT semantic segmentation technology to perform layered processing on the map data, separating the feature differences between key landmarks and dangerous areas. S22. Based on the feature difference analysis results, key underwater landmarks are identified, dangerous areas are clearly delineated, and an initial semantic navigation layer is generated. S23. Activate the eye-tracking module built into the diving goggles to capture the user's eye movement trajectory in real time and filter out valid blink signals that meet the preset standards; S24. Convert effective blink signals into hands-free target marking instructions and input instructions to clarify the user's focus and operational needs; S25. Update the marked content and attention area in the initial semantic navigation layer according to the hands-free instructions to form an updated semantic navigation layer that meets the user's needs.

[0016] The working principle and effects of the above technical solution are as follows: ViT semantic segmentation processes map data, separating the feature differences between key landmarks and dangerous areas, improving the accuracy of underwater environment identification and preventing accidental entry into unknown and dangerous areas. Dedicated marking and boundary definition for landmarks and dangerous areas enhances the practicality of the navigation layer, allowing users to clearly distinguish key environmental elements. The eye-tracking module captures effective blink signals and converts them into hands-free commands, reducing the tediousness of manual operation during diving and improving ease of use. The navigation layer is updated according to the commands, ensuring that the navigation content aligns with the user's focus, reducing interference from irrelevant information. This not only assists in accurate navigation but also meets users' personalized marking needs and allows for easy blink-based photo taking, making underwater exploration safer, smoother, and more enjoyable.

[0017] In one embodiment of the present invention, step S21 includes: Retrieve the constructed semantic topology map, perform data normalization processing, remove redundant spatial correlation information in the map, and generate normalized semantic topology map data; Feature dimensions are extracted from the normalized semantic topology map data to extract the core feature indicators of terrain, landmarks and regions in the underwater environment, and generate multi-dimensional map feature data. Start the ViT semantic segmentation technology, import multi-dimensional map feature data, adapt the segmentation model parameters, and generate the adapted segmentation data. Based on the adapted segmented data, the multi-dimensional map feature data is hierarchically divided to distinguish environmental feature levels of different depths and types, and to generate multi-level map feature subsets. By comparing the attribute differences of feature subsets of multi-level maps, unique feature identifiers of key landmarks and dangerous areas are selected, the feature information corresponding to the two types of areas is separated, and a feature difference dataset is generated.

[0018] The working principle and effects of the above technical solution are as follows: By removing redundant information from the semantic topology map, the data processing burden is reduced, the efficiency of subsequent calculations is improved, and redundant data is avoided from slowing down the recognition speed. Core feature indicators of terrain landmark areas are extracted to enhance the relevance of map features and reduce the interference of irrelevant features on the recognition results. ViT semantic segmentation technology adapts to multi-dimensional feature data, improving the adaptability of the segmentation model and making feature processing more consistent with the actual underwater environment. Different types of environmental features are hierarchically divided, making feature classification clearer and reducing recognition errors caused by confusion between different features. Unique features of key landmarks and dangerous areas are screened and separated to improve the distinction between the two types of areas, avoiding the risk of mistaking dangerous areas for safe landmarks. This provides accurate feature support for subsequent navigation layer generation and makes underwater environment recognition more reliable, adding to the safety of diving.

[0019] In one embodiment of the present invention, step S3 includes: S31. Activate the physiological monitoring component of the diving goggles to collect physiological status data, including user heart rate, blood oxygen saturation and diving time, and generate user physiological status dataset. S32. Retrieve the updated semantic navigation layer, extract the environmental association data therein, including ocean current direction and obstacle distribution, and generate an environmental reference dataset. S33. Integrate user physiological state dataset and environmental reference dataset to perform preliminary adaptive return path planning and generate multiple candidate return routes; S34. Based on the user's physiological tolerance and the complexity of the environment, the candidate return routes are prioritized and optimized to generate dynamic return routes; S35. Activate AR optical flow guidance technology to transform the dynamic return route into a visual optical flow image, which is projected onto the field of view area of ​​the diving goggles in real time to provide intuitive navigation reference.

[0020] The working principle and effects of the above technical solution are as follows: By collecting physiological data such as user heart rate and blood oxygen saturation, and extracting environmental correlation data such as ocean current direction and obstacle distribution, the return route planning is made more realistic, improving route adaptability and avoiding dangers caused by ignoring physical load or environmental changes. Multiple candidate routes are generated by integrating these two types of data, and priority ranking and optimization are performed based on physiological tolerance and environmental complexity to enhance route rationality and reduce detours, frequent obstacle avoidance, or exceeding the body's tolerance. AR optical flow technology transforms the dynamic return route into a visualized image projection field of view, improving navigation intuitiveness and reducing the probability of getting lost due to poor underwater visibility. This approach balances diving safety and route efficiency, allowing users to clearly perceive the navigation direction, reducing anxiety caused by unknown environments, making the dive return journey smoother and safer, and further enhancing the safety and experience of underwater exploration.

[0021] In one embodiment of the present invention, S33 includes: Receive user physiological state dataset and environmental reference dataset, perform data format unification processing, integrate the core information of the two types of data, and generate a comprehensive planning basic dataset; Weighted comprehensive planning data is generated by weighting factors such as physiological tolerance indicators, ocean current influence, and obstacle density in the comprehensive planning basic dataset. Based on the weighted comprehensive planning data, core constraints for path planning are set, including user physiological safety thresholds, obstacle avoidance ranges, and ocean current adaptation intervals, and a set of constraint planning parameters is generated. Import the constraint programming parameter set, use the multi-path search algorithm to mine feasible return paths that meet the conditions, and generate an initial candidate route set; The initial candidate route set is deduplicated and filtered to remove redundant routes with excessive path overlap, retaining differentiated and effective routes, and generating multiple candidate return routes.

[0022] The working principle and effects of the above technical solution are as follows: By unifying the format of user physiological state and environmental reference data, the core information of the two types of data is integrated, reducing planning obstacles caused by data incompatibility and improving the usability of basic data. Weighting calculations are performed on factors such as physiological tolerance indicators, ocean current influence, and obstacle density to enhance the targeting of path planning and avoid deviations from actual needs due to overlooking certain factors. Constraints such as physiological safety thresholds, obstacle avoidance ranges, and ocean current adaptation intervals are set to reduce the risk of routes exceeding the body's tolerance or encountering sudden obstacles, avoiding potential safety hazards. A multi-path search algorithm is used to discover routes that meet the conditions, increasing the return journey options and reducing the frustration of a single route failure. Initial candidate routes are deduplicated and redundant options with excessive overlap are eliminated, improving route planning efficiency and avoiding invalid routes consuming decision-making time. This ensures that the return journey path balances safety and adaptability, and provides a high-quality foundation for subsequent priority ranking, making the route planning for the dive return trip more reliable and efficient, reducing anxiety caused by unknown factors.

[0023] In one embodiment of the present invention, S34 includes: Extract the core attributes of multiple candidate return routes, including path length and turning frequency. Combine the user physiological state dataset to extract core indicators of physiological tolerance. Integrate the environmental reference dataset to determine key parameters of environmental complexity and generate a comprehensive evaluation index set. The data in the comprehensive evaluation index set are quantified and transformed, and physiological tolerance and environmental complexity are converted into calculable numerical data to generate a quantitative evaluation dataset. Based on the diving safety priority rules, weights are assigned to physiological and environmental indicators in the quantitative assessment dataset to highlight the influence weight of physiological safety indicators and generate weighted assessment parameters. By substituting the weighted evaluation parameters into the route evaluation model, a comprehensive score is calculated for each candidate return route, and the routes are arranged in order of their scores to generate a route priority sequence. For routes at the top of the priority sequence, check for potential conflicts or room for optimization, adjust route nodes to avoid environmental risks caused by temporary changes, adapt to the user's real-time physiological state, and generate dynamic return routes.

[0024] The working principle and effects of the above technical solution are as follows: By extracting core attributes such as path length and turning frequency of candidate routes, and combining them with physiological tolerance indicators and environmental complexity parameters, a comprehensive evaluation index set is generated, making route evaluation more comprehensive and avoiding route deviations caused by single-dimensional considerations. The various evaluation data are quantified and converted to improve the accuracy of the evaluation and reduce errors caused by subjective judgment. Weights are allocated according to diving safety priority rules, highlighting the impact of physiological safety indicators, enhancing the safety orientation of the route, and avoiding ignoring diving risks caused by physical strain. The comprehensive score is calculated and ranked by substituting the data into the evaluation model, making route selection more evidence-based, improving decision-making efficiency, and reducing the cost of screening invalid routes. Potential conflicts of leading routes are checked and nodes are adjusted to adapt to real-time physiological states and environmental changes, avoiding temporary risks affecting return safety. This ensures both route optimization and a more stable return process, further enhancing the safety and adaptability of underwater navigation, making the dive return journey safer and smoother.

[0025] In one embodiment of the present invention, step S4 includes: S41. Activate the diving goggles environmental perception module to continuously collect safety-related parameters, including water temperature, water depth, oxygen cylinder content, etc., and generate multi-dimensional safety parameter raw data. S42. Filter and integrate the raw data of multi-dimensional safety parameters, remove invalid data, and generate raw data for real-time safety warnings. S43. Retrieve the danger zone marking information from the updated semantic navigation layer and cross-compare it with the original real-time safety warning data; S44. Based on the cross-comparison results, potential safety risks are double-verified, triggering the corresponding level of early warning mechanism to ensure diving safety; S45. The eye-tracking module continuously captures the user's blinking signals. When a preset number of consecutive blinks are detected, the camera function is activated to capture and store the current field of vision image.

[0026] The working principle and effects of the above technical solution are as follows: By continuously collecting safety parameters such as water temperature, water depth, and oxygen cylinder content, invalid data is filtered, integrated, and eliminated to improve the purity of safety data and reduce the impact of interference information on early warning judgment. Danger zone marking information is retrieved and cross-checked with the original safety early warning data for double verification of potential risks, enhancing the accuracy of early warnings and avoiding missed or false alarms. Corresponding early warning mechanisms are triggered to reduce the probability of accidents during diving, building a solid defense for diving safety. An eye-tracking module captures a preset number of consecutive blinks to initiate photography, reducing the tediousness of manual operation during diving, improving shooting convenience, and preventing the loss of underwater beauty or disruption of the diving state due to manual operation. This comprehensively protects diving safety while meeting users' underwater photography needs, making the diving experience safer, smoother, and more enjoyable.

[0027] In one embodiment of the present invention, step S5 includes: S51. Collect real-time safety warning data and various feedback information during the execution of the dynamic return route, and generate a navigation execution feedback dataset; S52. Perform weighted calculations on the navigation execution feedback dataset, conduct navigation stability assessments, and generate a diving navigation stability index; S53. Based on the diving navigation stability index, dynamically adjust the optical flow velocity and voice prompt frequency of the navigation guidance to optimize the navigation guidance strategy; S54. Activate the AI ​​recognition module, receive underwater visual acquisition data, match it with the marine biological feature database, and present the biological name and related basic information; record motion data during the diving process, including trajectory, speed, etc., analyze swimming or diving posture related features, and generate motion state analysis data. S55. Based on motion state analysis data, extract posture optimization direction and generate personalized motion guidance data; integrate the optimized navigation guidance strategy, AI recognition results, and personalized motion guidance data to generate the final intelligent diving navigation data.

[0028] The working principle and effects of the above technical solution are as follows: By collecting real-time safety warning data and navigation execution feedback information, a diving navigation stability index is generated through weighted calculation, making the navigation status assessment more accurate and avoiding discomfort caused by blindly adjusting the guidance strategy. Based on the index, the optical flow velocity and voice prompt frequency are dynamically adjusted to optimize the navigation guidance method, enhance the adaptability of navigation to the actual scene, and reduce navigation difficulties caused by visual or auditory interference. The AI ​​recognition module matches marine life characteristics, presenting the names and basic information of the organisms, enriching the knowledge of the diving process and reducing confusion when facing unknown creatures. Track, speed, and other motion data are recorded and posture characteristics are analyzed to extract and optimize directions, generating personalized guidance, improving the standardization of diving or swimming postures, and reducing fatigue or injury caused by improper postures. The optimized navigation strategy, AI recognition results, and motion guidance data are integrated to generate final intelligent navigation data, which can comprehensively ensure diving safety while also taking into account learning and exercise needs, making the diving experience richer, smoother, and more practical, reducing the limitations of a single function.

[0029] In one embodiment of the present invention, S53 includes: Receive the diving navigation stability index, extract the core information corresponding to the index, including the numerical range and fluctuation amplitude, and simultaneously retrieve the optical flow velocity parameter and voice prompt frequency parameter of the current navigation guidance to generate a basic adjustment dataset; The stability index values ​​in the basic adjustment dataset are divided into intervals to define three adjustment intervals: low stability, medium stability, and high stability, generating index interval division data. Based on the exponential interval division data, optical flow velocity adjustment gradients are set for different intervals. The optical flow velocity is reduced in the low stable interval, the normal velocity is maintained in the medium stable interval, and the velocity is increased in the high stable interval, thus generating an optical flow velocity adjustment scheme. Based on the data divided into index intervals, a voice prompt frequency adjustment rule is formulated: the prompt frequency is increased in the low-stability interval, the standard frequency is maintained in the medium-stability interval, and the frequency is reduced in the high-stability interval, thus generating a voice prompt adjustment scheme. By integrating optical flow velocity adjustment schemes and voice prompt adjustment schemes, and adapting to the navigation needs of the current diving environment, an optimized navigation guidance strategy is generated.

[0030] The working principle and effects of the above technical solution are as follows: By extracting core information such as the numerical range and fluctuation amplitude of the navigation stability index, and simultaneously retrieving the current optical flow velocity and voice prompt frequency parameters, the adjustment is supported by solid data, avoiding navigation discomfort caused by blind adjustments. The index is divided into three ranges: low stability, medium stability, and high stability, making the adjustment direction more precise and reducing guidance confusion caused by adjustment deviations. In the low stability range, the optical flow velocity is reduced and the voice prompts are encrypted to ensure that users can clearly receive guidance in complex situations; the medium stability range maintains a normal state, balancing guidance effectiveness and user experience; the high stability range increases the optical flow velocity and reduces the prompt frequency to reduce unnecessary interference. Integrating the two types of adjustment schemes to adapt to the actual underwater environment enhances the fit between navigation guidance and real-time status, avoiding the disconnect between optical flow or voice rhythm and navigation needs. It can provide solid safety guidance in unstable situations and maintain smooth navigation in stable situations, reducing confusion or operational errors caused by improper guidance, making underwater navigation more in line with actual needs.

[0031] One embodiment of the present invention provides diving goggles based on a multi-source data fusion algorithm, the diving goggles comprising: One or more processors; Memory, used to store one or more programs; Wherein, when the one or more programs are executed by the one or more processors, the one or more processors are made to implement the method described in any one of the above.

[0032] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An adaptive navigation method for diving goggles based on a multi-source data fusion algorithm, characterized in that, The method includes: S1. Collect IMU-depth-Geomagnetic multi-source sensor data in the diving environment to generate an initial positioning dataset; perform fusion positioning calculation based on the initial positioning dataset to construct a semantic topology map in a GPS-free environment; S2. Perform ViT semantic segmentation based on the semantic topology map to identify and mark key underwater landmarks and dangerous areas, and generate a semantic navigation layer; at the same time, collect the user's blink signals through the eye-tracking module built into the diving goggles to realize hands-free target marking and command input, and update the user's focus points in the semantic navigation layer. S3. Based on the semantic navigation layer and user physiological status monitoring data, perform adaptive return path planning to generate a dynamic return route that takes into account ocean currents, obstacles and user physiological status. S4. Continuously collect safety parameters through the environmental perception module built into the diving goggles to generate real-time safety warning data; combine the dangerous area markings in the semantic navigation layer to perform dual verification and warning of potential safety risks; S5. Based on real-time safety warning data and the execution status of dynamic return routes, a weighted navigation stability assessment is performed to generate a diving navigation stability index; the navigation guidance strategy is dynamically adjusted based on this index to generate the final intelligent diving navigation data.

2. The adaptive navigation method for diving goggles based on a multi-source data fusion algorithm according to claim 1, characterized in that, S1 includes: S11. Activate the built-in IMU sensor, depth sensor, and geomagnetic sensor in the diving goggles to collect comprehensive data on the diving environment and generate multi-dimensional raw environmental data. S12. Receive multi-dimensional raw environmental data, perform noise filtering and data calibration, remove abnormal data points, and generate a clean initial positioning dataset. S13. Extract the feature information corresponding to underwater natural landmarks in the clean initial positioning dataset, analyze the semantic attributes of the landmarks, and simultaneously record the spatial location data of various landmarks. S14. Use a multi-source data fusion algorithm to process the initial clean positioning dataset, integrate the advantages of different sensor data, and generate accurate fused positioning data. S15. Based on the precise fusion of positioning data and landmark semantics and spatial location data, a semantic topology map is built in a GPS-free environment to fully present the semantic relationships and spatial distribution of the underwater environment.

3. The adaptive navigation method for diving goggles based on a multi-source data fusion algorithm according to claim 2, characterized in that, S14 includes: Receive the clean initial positioning dataset, separate the positioning data categories corresponding to the IMU sensor, depth sensor, and geomagnetic sensor, and generate a classified sensor positioning dataset. Perform data consistency verification on the classification sensor localization dataset to eliminate the temporal deviation of data from different sensors and generate a time-synchronized localization dataset. The multi-source data fusion algorithm is invoked to perform weight allocation calculations on the time-series synchronous positioning dataset, highlighting the influence of high-confidence sensor data and generating weighted fusion intermediate data; A second screening for outliers is performed on the weighted fusion intermediate data to eliminate biased data generated during the fusion process and generate fusion calibration positioning data. Integrate and fuse calibration positioning data, optimize data accuracy thresholds, and generate accurate fused positioning data.

4. The adaptive navigation method for diving goggles based on a multi-source data fusion algorithm according to claim 1, characterized in that, The S2 includes: S21. Retrieve the constructed semantic topology map, and start the ViT semantic segmentation technology to perform layered processing on the map data, separating the feature differences between key landmarks and dangerous areas. S22. Based on the feature difference analysis results, key underwater landmarks are identified, dangerous areas are clearly delineated, and an initial semantic navigation layer is generated. S23. Activate the eye-tracking module built into the diving goggles to capture the user's eye movement trajectory in real time and filter out valid blink signals that meet the preset standards; S24. Convert effective blink signals into hands-free target marking instructions and input instructions to clarify the user's focus and operational needs; S25. Update the marked content and attention area in the initial semantic navigation layer according to the hands-free instructions to form an updated semantic navigation layer that meets the user's needs.

5. The adaptive navigation method for diving goggles based on a multi-source data fusion algorithm according to claim 4, characterized in that, S21 includes: Retrieve the constructed semantic topology map, perform data normalization, remove redundant spatial association information in the map, and generate normalized semantic topology map data; Feature dimensions are extracted from the normalized semantic topology map data to extract the core feature indicators of terrain, landmarks and regions in the underwater environment, and generate multi-dimensional map feature data. Start the ViT semantic segmentation technology, import multi-dimensional map feature data, adapt the segmentation model parameters, and generate the adapted segmentation processing data. Based on the adapted segmented data, the multi-dimensional map feature data is hierarchically divided to distinguish environmental feature levels of different depths and types, and to generate multi-level map feature subsets. By comparing the attribute differences of feature subsets of multi-level maps, unique feature identifiers of key landmarks and dangerous areas are selected, the feature information corresponding to the two types of areas is separated, and a feature difference dataset is generated.

6. The adaptive navigation method for diving goggles based on a multi-source data fusion algorithm according to claim 1, characterized in that, The S3 includes: S31. Activate the diving goggles physiological monitoring component to collect physiological state data and generate a user physiological state dataset; S32. Retrieve the updated semantic navigation layer, extract the environmental association data from it, and generate an environmental reference dataset; S33. Integrate user physiological state dataset and environmental reference dataset to perform preliminary adaptive return path planning and generate multiple candidate return routes; S34. Based on the user's physiological tolerance and the complexity of the environment, prioritize and optimize the candidate return routes to generate dynamic return routes. S35. Activate AR optical flow guidance technology to transform the dynamic return route into a visual optical flow image, which is projected onto the field of view area of ​​the diving goggles in real time to provide intuitive navigation reference.

7. The adaptive navigation method for diving goggles based on a multi-source data fusion algorithm according to claim 6, characterized in that, S34 includes: Extract the core attributes of multiple candidate return routes, refine the core indicators of physiological tolerance by combining user physiological state datasets, integrate environmental reference datasets to determine key parameters of environmental complexity, and generate a comprehensive evaluation index set. The data in the comprehensive evaluation index set are quantified and transformed, and physiological tolerance and environmental complexity are converted into calculable numerical data to generate a quantitative evaluation dataset. Based on the diving safety priority rules, weights are assigned to physiological and environmental indicators in the quantitative assessment dataset to highlight the influence weight of physiological safety indicators and generate weighted assessment parameters. By substituting the weighted evaluation parameters into the route evaluation model, a comprehensive score is calculated for each candidate return route, and the routes are arranged in order of score to generate a route priority sequence. For routes at the top of the priority sequence, check for potential conflicts or room for optimization, adjust route nodes to avoid environmental risks caused by temporary changes, adapt to the user's real-time physiological state, and generate dynamic return routes.

8. The adaptive navigation method for diving goggles based on a multi-source data fusion algorithm according to claim 1, characterized in that, The S4 includes: S41. Activate the diving goggles environmental perception module to continuously collect safety-related parameters and generate multi-dimensional safety parameter raw data. S42. Filter and integrate the raw data of multi-dimensional safety parameters, remove invalid data, and generate raw data for real-time safety warnings. S43. Retrieve the danger zone marking information from the updated semantic navigation layer and cross-compare it with the original real-time safety warning data; S44. Based on the cross-comparison results, potential safety risks are double-verified, triggering the corresponding level of early warning mechanism to ensure diving safety; S45. The eye-tracking module continuously captures the user's blinking signals. When a preset number of consecutive blinks are detected, the camera function is activated to capture and store the current field of vision image.

9. The adaptive navigation method for diving goggles based on a multi-source data fusion algorithm according to claim 1, characterized in that, The S5 includes: S51. Collect real-time safety warning data and various feedback information during the execution of the dynamic return route, and generate a navigation execution feedback dataset; S52. Perform weighted calculations on the navigation execution feedback dataset, conduct navigation stability assessments, and generate a diving navigation stability index; S53. Based on the diving navigation stability index, dynamically adjust the optical flow velocity and voice prompt frequency of the navigation guidance to optimize the navigation guidance strategy; S54. Activate the AI ​​recognition module, receive underwater visual acquisition data, match it with the marine biological feature database, and present the biological name and related basic information; record motion data during the diving process, analyze the relevant features of swimming or diving postures, and generate motion state analysis data. S55. Based on motion state analysis data, extract posture optimization direction and generate personalized motion guidance data; integrate the optimized navigation guidance strategy, AI recognition results, and personalized motion guidance data to generate the final intelligent diving navigation data.

10. Diving goggles based on a multi-source data fusion algorithm, characterized in that, The diving goggles include: One or more processors; Memory, used to store one or more programs; Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method of any one of claims 1 to 9.

Images