Navigation method, apparatus, device, storage medium and product

Abstract

The application discloses a navigation method, device, equipment, storage medium and product, relates to the technical field of vehicle navigation, and the navigation method comprises the steps of acquiring multi-source data corresponding to a vehicle and driving scene information, wherein the multi-source sensing data at least comprises navigation data and traffic sign data; based on the driving scene information, adjusting the weight proportion of each data source in the multi-source data in navigation; performing semantic understanding on the traffic sign data to obtain structured data; and adjusting the navigation decision of the vehicle based on the adjusted weight proportion and the structured data. Through the adjusted weight proportion and the structured data after semantic understanding, the navigation decision is timely adjusted, so that the actual road position of the vehicle can be accurately determined.

Description

Technical Field

[0001] This application relates to the field of vehicle navigation technology, and in particular to a navigation method, device, equipment, storage medium and product. Background Technology

[0002] In related technologies, existing in-vehicle navigation and mobile phone navigation systems generally rely on Standard Definition Maps (SD Maps) and satellite navigation signals such as the Global Positioning System (GPS) for vehicle positioning. In normal road scenarios, this solution can basically meet the route guidance needs during driving by matching the received GPS coordinates with road network data. However, due to the limitations of ordinary GPS positioning accuracy, when the vehicle is driving in special road sections with complex structures or high spatial overlap, the positioning results are prone to lateral offset or multipath interference, making it difficult to accurately determine the actual road position of the vehicle. Summary of the Invention

[0003] The main purpose of this application is to provide a navigation method, device, equipment, storage medium and product, which aims to solve the technical problem that in special road sections, the positioning results are prone to lateral deviation or multipath interference, making it difficult to accurately determine the actual road position of the vehicle.

[0004] To achieve the above objectives, this application proposes a navigation method, the navigation method comprising:

[0005] Acquire multi-source data and driving scenario information corresponding to the vehicle, wherein the multi-source perception data includes at least navigation data and traffic sign data;

[0006] Based on the driving scenario information, adjust the weight ratio of each data source in the multi-source data in the navigation;

[0007] Semantic understanding is performed on the traffic sign data to obtain structured data;

[0008] The vehicle's navigation decisions are adjusted based on the adjusted weighting and the structured data.

[0009] In one embodiment, the step of performing semantic understanding on the traffic sign data to obtain structured data includes:

[0010] The traffic sign data is input into a preset semantic understanding model. The preset semantic understanding model performs semantic understanding on the traffic sign data and outputs structured data, wherein the structured data includes at least one of the road's physical attributes, road type attributes, and road function information.

[0011] In one embodiment, the step of inputting the traffic sign data into a preset semantic understanding model, performing semantic understanding on the traffic sign data through the preset semantic understanding model, and outputting structured data includes:

[0012] The traffic sign data is input into a preset semantic understanding model. The preset semantic understanding model is used to extract visual features and recognize text in the traffic sign data to obtain visual features and text features. The text features include traffic-related text information parsed from the traffic sign data.

[0013] The text features and the visual features are linked together to output structured data.

[0014] In one embodiment, the step of inputting the traffic sign data into a preset semantic understanding model, performing semantic understanding on the traffic sign data through the preset semantic understanding model, and outputting structured data includes:

[0015] The traffic sign data is input into a preset semantic understanding model. Through the preset semantic understanding model, multimodal structured feature extraction is performed on the traffic sign data to obtain visual features and traffic element features. The traffic element features include features corresponding to traffic-related detection objects detected from the traffic sign data.

[0016] The visual features and traffic element features are linked together to output structured data.

[0017] In one embodiment, the step of acquiring multi-source data and driving scenario information corresponding to the vehicle includes:

[0018] When a vehicle is detected to be in a preset navigation inaccuracy scenario, a resource request is actively sent to the target sensor, so that the target sensor can collect traffic sign data based on the resource request and transmit the traffic sign data to the local device through a standardized interface;

[0019] Receive traffic sign data sent by the target sensor.

[0020] In one embodiment, the step of actively sending a resource request to the target sensor when the vehicle is detected to be in a preset navigation inaccuracy scenario includes:

[0021] When a vehicle is detected to be in a preset navigation failure scenario, a target sensor is determined from preset visual sensor resources based on the target scene in the preset navigation failure scenario in which the vehicle is located, and a resource call request is actively sent to the target sensor.

[0022] In one embodiment, the target sensor includes a first sensor and a second sensor. The step of determining the target sensor from preset visual sensor resources based on the target scene in the preset navigation inaccuracy scenario in which the vehicle is located, and actively sending a resource access request to the target sensor includes:

[0023] When the target scenario of the preset navigation misalignment scenario is a traffic light scenario, the first sensor is determined from the preset visual sensor resources, and a resource call request is actively sent to the first sensor, the first sensor including a forward-looking camera;

[0024] or,

[0025] When the preset navigation inaccuracy scenario is a main and auxiliary road scenario, the second sensor is determined from the preset visual sensor resources, and a resource call request is actively sent to the second sensor. The second sensor includes a forward-looking camera and a fisheye camera.

[0026] In one embodiment, the step of actively sending a resource request to the target sensor when the vehicle is detected to be in a preset navigation inaccuracy scenario includes:

[0027] When a vehicle meets any of the preset triggering mechanisms, it is determined that the vehicle is in a preset navigation inaccuracy scenario. The preset triggering mechanisms include timed polling triggering, vehicle status triggering, and geofence triggering.

[0028] In one embodiment, the step of determining that the vehicle is in a preset navigation inaccuracy scenario when the vehicle meets any preset triggering mechanism includes:

[0029] When the vehicle meets the polling frequency of the current polling mode, it is determined that the vehicle responds to the timed polling trigger. The polling mode includes cruise mode, city mode and standby mode. The polling frequency of the timed polling trigger is determined based on the road type on which the vehicle is driving.

[0030] In one embodiment, the step of determining the vehicle's response to a timed polling trigger when the vehicle meets the polling frequency of the current polling mode includes:

[0031] Determine the type of road the vehicle is traveling on;

[0032] If the road type is a stable driving type, then the polling mode triggered by the timed polling is set to cruise mode;

[0033] If the road type is a complex driving type, then the polling mode triggered by the timed polling is set to city mode;

[0034] If the road type is a stationary driving type, then the polling mode triggered by the timed polling is set to standby mode.

[0035] In one embodiment, the step of determining that the vehicle is in a preset navigation inaccuracy scenario when the vehicle meets any preset triggering mechanism further includes:

[0036] When the navigation application is running, it acquires vehicle status information and navigation data;

[0037] Based on the vehicle status information and the navigation data, the vehicle response is triggered.

[0038] In one embodiment, the step of determining the vehicle's response to the vehicle state trigger based on the vehicle state information and the navigation data includes at least one of the following:

[0039] If the vehicle speed in the vehicle status information is within a preset threshold range, and based on the navigation data it is determined that the vehicle is within the area of ​​the traffic light geofence, then it is determined that the vehicle responds to the vehicle status trigger.

[0040] If the steering wheel angle in the vehicle status information is greater than a preset angle threshold, and based on the navigation data it is determined that the vehicle is within the area of ​​the ramp geofence, then it is determined that the vehicle responds to the vehicle status trigger.

[0041] If the turn signal in the vehicle status information is active, then it is determined that the vehicle is responding to the vehicle status trigger.

[0042] In one embodiment, the step of determining that the vehicle is in a preset navigation inaccuracy scenario when the vehicle meets any preset triggering mechanism further includes:

[0043] Obtain navigation data, and based on the vehicle location coordinates in the navigation data, determine whether the vehicle is within an area of ​​a preset number of preset geofences or greater.

[0044] If it is in the preset geofence, the weights corresponding to the preset geofence are summed to obtain the comprehensive weight;

[0045] If the overall weight is greater than the weight threshold, then the vehicle response geofence is determined to be triggered.

[0046] In one embodiment, the step of adjusting the vehicle's navigation decision based on the adjusted weighting and the structured data includes:

[0047] Get a predefined set of scene states;

[0048] Based on the adjusted weighting and the structured data, the confidence level of the vehicle in different scenario states within the scenario state set is determined.

[0049] Based on the confidence level, the target scene state of the vehicle is determined;

[0050] The vehicle's navigation decisions are compensated based on the target scene state.

[0051] In one embodiment, the step of determining the confidence level of the vehicle in different scenario states within the scenario state set based on the adjusted weighting and the structured data includes:

[0052] The visual score is determined based on the number of overlaps between each feature in the structured data and the visual features corresponding to different scene states in the scene state set.

[0053] Based on the navigation data, the degree of matching between the vehicle's position coordinates and the map network corresponding to the different scene states is calculated to obtain a positioning score;

[0054] Based on the adjusted weight ratio, the positioning score and the visual score are weighted and summed to obtain the confidence level under different scene states.

[0055] Furthermore, to achieve the above objectives, this application also proposes a navigation device, the navigation device comprising:

[0056] The acquisition module is used to acquire multi-source data and driving scenario information corresponding to the vehicle, wherein the multi-source perception data includes at least navigation data and traffic sign data;

[0057] The weight adjustment module is used to adjust the weight ratio of each data source in the multi-source data in the navigation based on the driving scenario information.

[0058] The semantic understanding module is used to perform semantic understanding on the traffic sign data to obtain structured data;

[0059] The navigation adjustment module is used to adjust the vehicle's navigation decisions based on the adjusted weight ratios and the structured data.

[0060] In addition, to achieve the above objectives, this application also proposes a navigation device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the navigation method as described above.

[0061] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and which, when executed by a processor, implements the steps of the navigation method described above.

[0062] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the navigation method described above.

[0063] One or more technical solutions proposed in this application have at least the following technical effects:

[0064] Compared to related technologies, where the positioning accuracy of ordinary GPS is limited, and the positioning results are prone to lateral shifts or multipath interference when a vehicle is driving in a special road segment with complex structure or high spatial overlap, making it difficult to accurately determine the actual road position of the vehicle, this application obtains multi-source data and driving scenario information corresponding to the vehicle. The multi-source perception data includes at least navigation data and traffic sign data. Based on the driving scenario information, the weight ratio of each data source in the multi-source data is adjusted in the navigation. Based on the adjusted weight ratio and the structured data, the navigation decision of the vehicle is adjusted. This application automatically obtains multi-source data including at least navigation data and traffic sign data, as well as driving scenario information, and adjusts the weight ratio of each data source in the navigation according to the driving scenario information, allowing the data source that best matches the current driving scenario to dominate the navigation. Then, semantic understanding is performed on the traffic sign data to obtain structured data. Through the adjusted weight ratio and the semantically understood structured data, the navigation decision is adjusted in a timely manner to accurately determine the actual road position of the vehicle. Attached Figure Description

[0065] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0066] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0067] Figure 1 This is a flowchart illustrating an embodiment of the navigation method of this application.

[0068] Figure 2 This is a flowchart of the in-vehicle navigation system based on the navigation method described in this application;

[0069] Figure 3 Automatic correction diagrams are provided for scenes under and on the viaduct;

[0070] Figure 4 This is a flowchart illustrating Embodiment 2 of the navigation method of this application;

[0071] Figure 5 This is a schematic diagram illustrating the automatic correction of the main and auxiliary road scenarios in the navigation method of this application;

[0072] Figure 6 This is a flowchart illustrating Embodiment 4 of the navigation method of this application;

[0073] Figure 7 This is a schematic diagram of the module structure of the navigation device according to an embodiment of this application;

[0074] Figure 8 This is a schematic diagram of the device structure of the hardware operating environment involved in the navigation method in the embodiments of this application.

[0075] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0076] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0077] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0078] The main solution of this application embodiment is: to acquire multi-source data and driving scenario information corresponding to the vehicle, wherein the multi-source perception data includes at least navigation data and traffic sign data; based on the driving scenario information, to adjust the weight ratio of each data source in the multi-source data in navigation; to perform semantic understanding on the traffic sign data to obtain structured data; and to adjust the vehicle's navigation decision based on the adjusted weight ratio and the structured data.

[0079] In related technologies, due to limitations in the accuracy of ordinary GPS positioning, when a vehicle is traveling on a special road section with a complex structure or highly overlapping spaces, the positioning results are prone to lateral shifts or multipath interference, making it difficult to accurately determine the actual road position of the vehicle, leading to navigation errors.

[0080] This application automatically acquires multi-source data, including at least navigation data and traffic sign data, as well as driving scenario information. Based on the driving scenario information, it adjusts the weight ratio of each data source in the navigation, allowing the data source that best matches the current driving scenario to dominate the navigation. Then, it performs semantic understanding on the traffic sign data to obtain structured data. Through the adjusted weight ratio and the semantically understood structured data, it adjusts the navigation decision in a timely manner to accurately determine the actual road location of the vehicle.

[0081] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or navigation device capable of performing the above functions. The following description uses a navigation device as an example to illustrate this embodiment and the subsequent embodiments.

[0082] Based on this, the embodiments of this application provide a navigation method, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the navigation method of this application.

[0083] In this embodiment, the navigation method includes steps S10 to S40:

[0084] Step S10: Obtain multi-source data and driving scenario information corresponding to the vehicle, wherein the multi-source perception data includes at least navigation data and traffic sign data;

[0085] It should be noted that the execution entity in this embodiment is the navigation device. Traffic sign data includes images or video streams related to navigation captured by high-precision sensors (such as a high-frame-rate forward-facing camera) through a request from the intelligent driving domain. Driving scenario information includes GPS positioning accuracy, visual confidence level, vehicle speed, ambient light intensity, and other information. When the navigation device detects that the vehicle is in a preset navigation inaccuracy scenario, it will automatically acquire navigation data from the cockpit domain (navigation) and traffic sign data from the intelligent driving domain through a standardized interface, achieving direct transmission and parsing of sensor data, and updating the driving scenario information every second.

[0086] Step S20: Based on the driving scenario information, adjust the weight ratio of each data source in the multi-source data in the navigation.

[0087] Understandably, navigation devices dynamically adjust the weight of each data source in navigation based on driving scenario information, ensuring that the decision-making model always adapts to the current scenario. This breaks the limitations of traditional navigation systems that rely on fixed priorities or a single data source, enabling navigation decisions to adapt to complex road environments and changing sensor conditions. Based on driving scenario information, it determines the credibility of the current environment (such as GPS positioning accuracy and visual clarity) and assigns decision-making authority to the most reliable data source, thereby effectively addressing the problem of a single data source failing or becoming less reliable in specific scenarios.

[0088] For example, when a vehicle enters a tunnel during the day, the navigation device obtains real-time data showing that positioning accuracy drops sharply due to signal obstruction (HDOP horizontal accuracy attenuation factor value increases), visual confidence remains high due to uniform lighting inside the tunnel, the vehicle speed is 80 km / h, and ambient light intensity drops sharply upon entering the tunnel. Based on these real-time parameters, the device dynamically adjusts the GPS weight while maintaining or increasing the visual weight, thereby relying on structured data output from visual perception (such as lane line recognition) to compensate for navigation and positioning failures during GPS outages, maintaining the continuity of lane-level guidance.

[0089] This application abandons the traditional static weights (such as a fixed visual weight of 60% and GPS weight of 40%) and proposes a scene-adaptive weight iteration mechanism. It relies on visual structured data for lane-level positioning compensation, which avoids the positioning jumps or loss caused by GPS failure in traditional systems and ensures the continuous and stable navigation guidance.

[0090] Step S30: Perform semantic understanding on the traffic sign data to obtain structured data;

[0091] It should be noted that since most current navigation systems are based on high-precision maps and static POI data, they lack the ability to acquire and compensate for real-time dynamic traffic elements. In actual driving, drivers need to rely on traffic light colors and countdown information, road construction notices, temporary speed limit signs, mobile traffic control signs, etc., to make decisions. Traditional navigation systems cannot obtain this temporary information through pure GPS and maps, resulting in navigation prompts being out of sync with the actual traffic environment. Therefore, navigation devices perform semantic understanding of traffic sign data to obtain structured data to supplement these traffic elements and improve navigation reliability.

[0092] Step S40: Adjust the vehicle's navigation decision based on the adjusted weight ratio and the structured data.

[0093] It should be noted that navigation decision-making refers to the final determination of the vehicle's current situation (e.g., on / under a bridge, main road / auxiliary road, lane). The navigation device, based on adjusted weights and semantic understanding, uses structured data to flexibly adapt to complex and changing driving environments (e.g., signal obstruction, weather changes). In cases of conflicting data sources or decreased reliability of a particular data source, the navigation prioritizes the most reliable evidence. Figure 2 , Figure 2 A flowchart of an in-vehicle navigation system based on visual recognition compensation is provided.

[0094] In one feasible implementation, step S30 includes:

[0095] The traffic sign data is input into a preset semantic understanding model. The preset semantic understanding model performs semantic understanding on the traffic sign data and outputs structured data. The structured data includes at least one of the road's physical attributes, road type attributes, and road function information.

[0096] It is understandable that traffic sign data refers to real-time images and high-definition video streams collected by vehicle-mounted sensors (such as forward-facing cameras and surround-view cameras), used to describe visual information about the environment surrounding the vehicle and related to navigation. The pre-defined semantic understanding model is a multi-task AI neural network model. This model is not limited to existing CNN convolutional neural networks, YOLO series models, or Transformer series VLM models, and has scene recognition field output. Structured data is integrated visual scene information. Road physical attributes include lane lines (such as solid lines and dashed lines), road boundaries (such as guardrails and curbs), and ground markings, describing the physical structure of the road. Road type attributes refer to road category information, such as main road, auxiliary road, on bridge, under bridge, etc. Road function information refers to the road's purpose or guidance information, such as highway entrances, auxiliary road exits, traffic light countdowns, etc. Structured data refers to data with a fixed format, clearly defined fields, and predefined models, typically organized in rows and columns.

[0097] It should be noted that the navigation device extracts structured features from traffic sign data through a preset semantic understanding model, generating structured data containing road physical attributes, type attributes, and functional information. It can obtain rich and semantically clear road environment descriptions from visual input, providing an accurate and reliable evidence basis for subsequent multi-source data fusion and navigation, thereby effectively solving the positioning ambiguity problem caused by GPS errors or complex environments in traditional navigation.

[0098] Specifically, the elements to be identified in the preset semantic understanding model include: main road and auxiliary road identification, which accurately determines whether a vehicle is on the main road or auxiliary road by analyzing features such as road boundaries, markings, and road signs; bridge and under-bridge identification, which accurately identifies the position on or under bridges by utilizing bridge structural features and changes in road markings; traffic light identification, which detects the status and countdown of traffic lights in real time with an accuracy of ±1 second; temporary traffic sign identification, which identifies temporary traffic signs such as temporary construction, speed limits, and no-turn signs; and lane-level positioning, which combines road markings and the relative position of vehicles to achieve centimeter-level lane positioning and output the lane where the vehicle is located.

[0099] Specifically, the navigation device inputs traffic sign data into a preset semantic understanding model after preprocessing steps such as distortion correction and high dynamic range (HDR) synthesis.

[0100] In one feasible implementation, the steps of inputting the traffic sign data into a preset semantic understanding model, performing semantic understanding on the traffic sign data through the preset semantic understanding model, and outputting structured data include:

[0101] The traffic sign data is input into a preset semantic understanding model. The preset semantic understanding model is used to extract visual features and recognize text in the traffic sign data to obtain visual features and text features. The text features include traffic-related text information parsed from the traffic sign data.

[0102] It should be noted that visual features refer to non-textual semantic information parsed from image pixels, including but not limited to lane line type (solid / dashed lines), traffic sign shape and color, traffic light status, road boundaries (guardrails, curbs), drivable areas, and the outlines, positions, and attributes of traffic participants such as vehicles and pedestrians. Text features specifically refer to the character sequence information identified and parsed from specific areas (such as within the bounding box of a traffic sign) in the image of traffic sign data by the optical character recognition module integrated into the model, such as road names ("G15 Expressway"), directional instructions ("auxiliary road exit"), traffic control information ("construction ahead", "speed limit 80"), and traffic light countdown numbers.

[0103] Understandably, navigation devices use a preset semantic understanding model to process multimodal information such as images and video streams in traffic sign data in parallel or collaboratively, and extract structured features from them to obtain high-precision visual features and text features with strong semantic constraints, thereby achieving a comprehensive understanding of the environment from physical form to textual semantics.

[0104] The text features and the visual features are linked together to output structured data.

[0105] It should be noted that by deeply linking textual and visual features, the navigation device outputs structured data that not only includes the physical attributes of the environment but also incorporates textual evidence with strong semantic constraints, thus forming a dual basis for judgment.

[0106] When navigation devices rely solely on visual features for navigation, the homogenization of visual features can lead to positioning ambiguities. For example, both elevated roads have physical guardrails and solid white lines, making them difficult to distinguish visually. However, textual information (such as the upper road sign "G15 Entrance" and the lower road sign "Urban Ring Road") can directly pinpoint the road level. This also compensates for the lack of context and provides more accurate semantic support for positioning.

[0107] In one feasible implementation, the steps of inputting the traffic sign data into a preset semantic understanding model, performing semantic understanding on the traffic sign data through the preset semantic understanding model, and outputting structured data include:

[0108] The traffic sign data is input into a preset semantic understanding model. Through the preset semantic understanding model, multimodal structured feature extraction is performed on the traffic sign data to obtain visual features and traffic element features. The traffic element features include features corresponding to traffic-related detection objects detected from the traffic sign data.

[0109] It is understood that traffic element features specifically refer to the feature descriptions obtained after locating and classifying key traffic-related objects in the traffic sign data image through the target detection network in the preset semantic understanding model. Detection objects include traffic lights, static traffic signs, and dynamic or temporary traffic signs. The navigation device identifies the visual features in the traffic sign data through the preset semantic understanding model. and characteristics of transportation elements.

[0110] Specifically, the navigation device uses a highly efficient target detection network with a pre-set semantic understanding model to locate and classify key traffic elements in the image, and outputs a set of detected objects. Each object The information is ,in, class_id It includes traffic lights, static traffic signs, and dynamic or temporary traffic signs. `state` is used to describe dynamic attributes of traffic lights, such as color (red, green, yellow). For each class_id The confidence level.

[0111] Furthermore, when the detected object contains text, the navigation device can also extract the image region within the bounding box (bbox) of the sign (especially temporary signs) identified by the target detection network, and then recognize and output the text information on the sign through an optical character recognition (OCR) network with a preset semantic understanding model. For example, "Construction ahead", "Speed ​​limit 80" or "Traffic light countdown".

[0112] Furthermore, the navigation device integrates all the outputs of the pre-set semantic understanding model into a unified, structured visual scene information. This provides a comprehensive and detailed real-time environmental description for subsequent decision-making steps.

[0113] For example, refer to Figure 3 , Figure 3 The document provides schematic diagrams illustrating the automatic correction of scenes under and on the viaduct. (Automatic correction of scenes under and on the viaduct is shown below.)

[0114] S1011: The user drives the vehicle and starts the in-vehicle navigation system, planning a route that passes through a large overpass. The navigation system initializes and runs, and the GPS real-time positioning function is active.

[0115] S1012: When the navigation system detects that the vehicle is about to enter the interchange area (determined by the POI geofence trigger mechanism), the cockpit domain (DHU) system sends a call command to the intelligent driving domain to activate the forward-looking camera and the Global Navigation Satellite System (GNSS) time synchronization module.

[0116] S1013: The system triggers the forward-facing camera to capture a video stream of the road scene ahead; simultaneously, the CAN (Controller Area Network) bus analysis result indicates that the current Vehicle Speed ​​is 60 km / h. The video stream is input to the AI ​​visual recognition model for semantic analysis, yielding the following results:

[0117] Navigation system (based on GPS signal): Determines that the vehicle is located on the road "under the bridge".

[0118] AI vision model: It can identify typical concrete guardrails and continuous white solid line boundaries of an overpass in the image, and detect the road sign "G15 Expressway Entrance" in the distance.

[0119] The model outputs structured data:

[0120] {

[0121] "road_type": "elevated_highway", / / Road type determined as "elevated highway"

[0122] "on_bridge": true, / / The current vehicle is located on a bridge / elevated section (not on a ground-level road).

[0123] "detected_signs": ["G15_entrance"] / / The "G15 Expressway Entrance" sign was detected.

[0124] }

[0125] S1014: The decision module performs evidence fusion between GPS output and visual recognition results.

[0126] Among them, visual evidence supports the view that it is "on the bridge". The score is significantly higher than "under the bridge," even though the GPS location points to "under the bridge." The weighted fusion score calculation result is:

[0127]

[0128] Therefore, the final decision result It means "on the bridge".

[0129] S1015: The decision results are fed back to the navigation APP through the transparent API interface.

[0130] Interface display: When the interface button switches to the bridge, the route is refreshed, and the vehicle icons on the map are updated to the bridge road level in real time to avoid incorrect display on parallel roads below the bridge. A Toast (user prompt mechanism) is displayed: "You have been detected driving on the bridge and your route has been replanned."

[0131] Voice announcement: The system proactively prompts, "We have detected that you are driving on the bridge and have replanned your route."

[0132] HUD (Head-Up Display) projection: The elevated lane boundaries and the upcoming highway entrance are highlighted in the HUD in front of the driver, ensuring that the driver does not need to look down to confirm.

[0133] Through the above methods, the navigation device achieves intelligent correction of "on / under the bridge" ambiguity caused by GPS errors, ensuring the accuracy and real-time nature of navigation guidance.

[0134] The visual features and traffic element features are linked together to output structured data.

[0135] It should be noted that the navigation device deeply links the visual features of the road structure with the features of dynamic traffic and outputs structured data, so that the output structured data can simultaneously cover the static structure and dynamic changes of the environment.

[0136] In this embodiment, since traditional navigation relies on static maps and POI data and cannot obtain dynamic information such as traffic light countdowns and temporary construction notices, this application introduces a deep collaborative mechanism of visual perception and OCR text recognition, which can provide more accurate semantic support for positioning.

[0137] Based on the first embodiment of this application, in the third embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 4 Step S10, the navigation method further includes steps A01~A02:

[0138] Step A01: When the vehicle is detected to be in a preset navigation inaccuracy scenario, a resource call request is actively sent to the target sensor so that the target sensor can collect traffic sign data based on the resource call request and transmit the traffic sign data to the local device through a standardized interface;

[0139] It should be noted that the preset navigation inaccuracy scenarios refer to road scenarios that are likely to cause navigation positioning deviations, as determined by triggering mechanisms (including timed patrols, vehicle status triggers, or POI geofence triggers). These include, but are not limited to, parallel main and auxiliary roads, overlapping bridges, and elevated interchange areas. Navigation data includes GPS positioning information, map matching results, and route planning data. The target sensor is a high-precision sensor from the intelligent driving domain. When the navigation device detects that the vehicle is in a preset navigation inaccuracy scenario, it can proactively request the use of high-precision sensors from the intelligent driving domain, rather than passively receiving the final perception results from the intelligent driving domain. It can also directly obtain traffic sign data transmitted to the local system through a standardized interface. This proactive and precise sensor retrieval mechanism breaks through the limitations of traditional navigation systems that are resource-constrained or passively receive data.

[0140] Step A02: Receive traffic sign data sent by the target sensor.

[0141] Understandably, navigation devices acquire traffic sign data from intelligent driving domain target sensors through standardized interfaces. By establishing a data path for proactive requests and reliable reception, the system ensures that it can obtain the required high-quality traffic sign data in a timely manner when a preset navigation inaccuracy scenario is triggered. This solves the problem of untimely correction caused by insufficient data sources or data lag in traditional navigation systems.

[0142] In one feasible implementation, step A01 includes:

[0143] When a vehicle is detected to be in a preset navigation failure scenario, a target sensor is determined from preset visual sensor resources based on the target scene in the preset navigation failure scenario in which the vehicle is located, and a resource call request is actively sent to the target sensor.

[0144] It should be noted that when the navigation device detects that the vehicle is in a preset navigation inaccuracy scenario, it will not call up all the sensors. Instead, it will precisely request the required sensor resources on demand according to the target scenario in which the vehicle is located, effectively reducing the overhead of redundant data processing and ensuring real-time response in high-speed scenarios.

[0145] In one feasible implementation, the steps of determining a target sensor from preset visual sensor resources based on the target scene in the preset navigation inaccuracy scenario where the vehicle is located, and actively sending a resource request to the target sensor include:

[0146] When the target scenario of the preset navigation misalignment scenario is a traffic light scenario, the first sensor is determined from the preset visual sensor resources, and a resource call request is actively sent to the first sensor, the first sensor including a forward-looking camera;

[0147] As is understandable, a traffic light scenario refers to a situation where a vehicle is at or near an intersection or road segment controlled by traffic lights, and the navigation system needs to obtain the traffic light status (including color and countdown information) to provide accurate guidance. A forward-facing camera is installed at the front of the vehicle, facing the direction of travel, and has sufficient resolution and frame rate to clearly capture the status and countdown numbers of the traffic lights ahead. When the navigation device determines that the preset navigation inaccuracy scenario is a specific navigation deviation scenario like a traffic light scenario, it accurately selects the most suitable forward-facing camera as the first sensor and initiates a resource request to the sensor management middleware to call the forward-facing camera. This avoids the resource waste caused by calling unnecessary sensors (such as surround-view and side-view cameras), ensuring that the system can obtain the most critical traffic sign data with the lowest power consumption and computational load.

[0148] Alternatively, when the preset navigation inaccuracy scenario is a main-auxiliary road scenario, the second sensor is determined from the preset visual sensor resources, and a resource call request is actively sent to the second sensor, which includes a forward-looking camera and a fisheye camera.

[0149] It should be noted that the "main and auxiliary road scenario" refers to a situation where a vehicle is traveling on a road with parallel main and auxiliary roads, and GPS positioning may be unable to accurately distinguish whether the vehicle is on the main road or the auxiliary road due to signal drift or overlapping road projections. The fisheye camera provides a wide-angle view of the vehicle's surroundings, making it particularly suitable for monitoring adjacent lane conditions, the relative position of the vehicle to lane lines, and blind spots, providing close-range spatial context for determining the main and auxiliary road. In the main and auxiliary road scenario, the navigation device simultaneously utilizes both the forward-facing camera and the fisheye camera, two sensors with complementary characteristics.

[0150] For example, refer to Figure 5 , Figure 5 An automatic correction diagram for the main-auxiliary road scenario is provided. The automatic correction steps for the main-auxiliary road scenario are as follows:

[0151] S101: When a user drives a vehicle and starts the in-vehicle navigation system, the planned route passes through urban expressways, including multiple scenarios where main roads and auxiliary roads run parallel.

[0152] S102: When the vehicle enters the POI electronic fence area at the intersection of the main and auxiliary roads (the triggering condition is met), the cockpit domain DHU calls the forward-facing camera and the fisheye camera, and requests CAN signals such as the vehicle's turn signals and steering wheel angle from the intelligent driving domain.

[0153] S103: The system triggers the forward-facing camera to capture a video stream of the road scene ahead; simultaneously, the CAN bus analysis result indicates that the current VehicleSpeed ​​= 30 km / h. The video stream is input to the AI ​​visual recognition model for semantic analysis, yielding the following results:

[0154] The forward-facing camera detected a fork in the road ahead, and the text "Auxiliary Road Entrance 2KM" and the dashed and solid lane lines appeared in the image. The fisheye camera provided a wide-angle view of the adjacent lanes.

[0155] AI visual models output structured data:

[0156] {

[0157] "road_type": "main_road", / / Meaning: The road the vehicle is currently on is identified as a "main road" (i.e., the main line of an urban expressway or highway).

[0158] "detected_signs": ["Auxiliary road entrance"], / / Meaning: A traffic sign was detected in the field of view, and OCR identified its text as "Auxiliary road entrance".

[0159] "lane_boundary": "solid_line_right", / / Meaning: The right boundary line of the current lane is a solid white (or yellow) line.

[0160] "adjacent_lane_status": "safe" / / Meaning: The adjacent lane (usually referring to the right-hand auxiliary lane or the lane adjacent to the main road) is currently free of conflicting vehicles or obstacles, and its status is safe.

[0161] }

[0162] GPS navigation information: The vehicle is in the auxiliary lane, but the model determines that the current lane boundary is a solid line, and the vehicle has not yet entered the auxiliary lane.

[0163] S104: Decision module compares navigation system and visual recognition results

[0164] Visual evidence indicates the vehicle is still on the main road, and the vehicle has not yet entered the right-hand auxiliary road entrance. The GPS system may have "drifted," misjudging the vehicle as being on the auxiliary road. After weighted fusion of evidence, the main road scores higher than the auxiliary road.

[0165]

[0166] Therefore, the final navigation correction result The vehicle was determined to be on the main road.

[0167] S105: Final results are transmitted to the navigation system

[0168] The interface displays: The vehicle icon on the map has been corrected from an incorrect trajectory on the auxiliary road to the correct lane on the main road.

[0169] Voice prompt: "Visual perception has detected that you are driving on the main road. Your route has been replanned. The exit on the auxiliary road is 2 kilometers ahead."

[0170] HUD Display: Highlights the main road straight-ahead guide arrows and plans the auxiliary road directions in the head-up display to avoid driver confusion.

[0171] Through this embodiment, the navigation device achieves automatic identification and error correction in main and auxiliary road scenarios, solving the limitation of traditional navigation systems that cannot distinguish between main and auxiliary roads, thereby improving the accuracy of route guidance and driving safety.

[0172] In this embodiment, the on-demand allocation strategy avoids the resource waste caused by blindly calling all sensors in traditional systems. It ensures that when a specific navigation problem occurs, the most effective perception data can be obtained with the optimal sensor configuration, thereby significantly improving the system's resource utilization efficiency and real-time performance while ensuring the accuracy of navigation correction.

[0173] Based on the first embodiment of this application, in the fourth embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description and will not be repeated hereafter. Based on this, before step S10, the navigation method further includes step C01:

[0174] Step C01: When the vehicle meets any preset triggering mechanism, it is determined that the vehicle is in a preset navigation inaccuracy scenario. The preset triggering mechanisms include timed polling triggering, vehicle status triggering, and geofence triggering.

[0175] Understandably, the preset triggering mechanism is a set of judgment logics pre-set by the system to actively identify conditions that may lead to navigation deviation. The navigation device actively identifies preset navigation inaccuracy scenarios through a triple triggering mechanism; that is, when any one of the three conditions is met, the visual compensation module will be activated.

[0176] In one possible implementation, the following is included before step C01:

[0177] When the vehicle meets the polling frequency of the current polling mode, it is determined that the vehicle responds to the timed polling trigger. The polling mode includes cruise mode, city mode and standby mode. The polling frequency of the timed polling trigger is determined based on the road type on which the vehicle is driving.

[0178] It should be noted that the polling mode is a perception strategy state defined based on the vehicle's driving environment. When the navigation device confirms that the vehicle meets the periodic check frequency dynamically set in the polling mode, it determines that the vehicle will respond to the timed polling trigger.

[0179] In one feasible implementation, when the vehicle meets the polling frequency of the current polling mode, the step of determining the vehicle's response to the timed polling trigger includes:

[0180] Determine the type of road the vehicle is traveling on;

[0181] It is understandable that road types include at least highways, urban expressways, arterial roads, auxiliary roads, ramps, bridges (both above and below bridges), tunnels, and other categories with different navigation characteristics. The navigation device determines the current road type the vehicle is traveling on based on navigation data, i.e., the vehicle's driving environment state machine. and according to Dynamically switch polling mode:

[0182]

[0183] If the road type is a stable driving type, then the polling mode triggered by the timed polling is set to cruise mode;

[0184] It should be noted that the stable driving type This refers to road environments where the vehicle is in a relatively simple road structure, such as highways or closed expressways, with a low frequency of changes in traffic elements. When the navigation device detects that the road type the vehicle is traveling on is a stable driving type, it determines that the polling mode triggered by the timed polling is cruise mode. Cruise mode is suitable for stable driving scenarios on highways or expressways, where the visual compensation requirement is low and the polling frequency is low. Set to normal frequency, focusing on monitoring large traffic signs and road structures ahead.

[0185] If the road type is a complex driving type, then the polling mode triggered by the timed polling is set to city mode;

[0186] Understandably, complex driving types This refers to the vehicle being in an area with numerous traffic lights in the city. When the navigation device detects that the vehicle is traveling on a complex road type, it determines that the polling mode triggered by the timed polling is the city mode. When the city mode is automatically activated, it switches to high-frequency polling. It prioritizes using the forward-facing camera, focusing on traffic lights and pedestrian detection.

[0187] If the road type is a stationary driving type, then the polling mode triggered by the timed polling is set to standby mode.

[0188] It should be noted that the stationary driving type This refers to scenarios involving parking lots or vehicles that have been stationary for extended periods. When the navigation device detects that the road type the vehicle is traveling on is a stationary driving type, it determines that the polling mode triggered by the timed polling is in standby mode. When standby mode is automatically enabled, polling is almost completely disabled, maintaining only the basic listening frequency. .

[0189] Furthermore, the polling frequency function can be expressed as:

[0190]

[0191] In one possible implementation, the following is included before step C01:

[0192] When the navigation application is running, it acquires vehicle status information and navigation data;

[0193] It should be noted that vehicle status information includes VehicleSpeed (vehicle speed) SteeringAngle (Steering wheel angle) TurnIndicators (Turn signal) GearPosition (Gear position) and other specific messages. When the in-vehicle navigation system is running, the navigation device acquires navigation data and vehicle status information obtained through the vehicle's CAN bus interface.

[0194] Based on the vehicle status information and the navigation data, the vehicle response is triggered.

[0195] Understandably, the navigation device combines vehicle status information and navigation data to achieve predictive triggering. The triggering state vector is as follows:

[0196]

[0197] In one feasible implementation, the step of determining the vehicle's response to the vehicle state trigger based on the vehicle state information and the navigation data includes at least one of the following:

[0198] If the vehicle speed in the vehicle status information is within a preset threshold range, and based on the navigation data it is determined that the vehicle is within the area of ​​the traffic light geofence, then it is determined that the vehicle responds to the vehicle status trigger.

[0199] It should be noted that the preset threshold range refers to the speed range set to determine whether a vehicle is waiting at a traffic light, usually defined as a range close to zero speed. The traffic light geofence area refers to an electronic fence area containing traffic light location information, constructed through multi-source data fusion, and its range usually covers a certain distance before and after the stop line at the intersection. When the navigation device detects that the vehicle is nearly stationary and is currently at a traffic light intersection, it determines that the vehicle response status has been triggered.

[0200] Specifically, when the navigation device determines Furthermore, when the navigation information confirms that the vehicle is near the stop line at a traffic light intersection, the vision system is triggered to focus on recognizing the traffic light status and the starting situation of the vehicle in front.

[0201]

[0202] in, This indicates that the triggering condition has been met. Indicates a traffic light geofence. Indicates the vehicle's location.

[0203] If the steering wheel angle in the vehicle status information is greater than a preset angle threshold, and based on the navigation data it is determined that the vehicle is within the area of ​​the ramp geofence, then it is determined that the vehicle responds to the vehicle status trigger.

[0204] Understandably, the preset turning angle threshold refers to the angle limit value used to determine whether the driver is performing a large steering operation. It is typically set based on the vehicle model and steering characteristics, for example, 45 degrees. The area within the ramp geofence refers to the electronic fence range that includes the entrance / exit areas of the highway ramp. When the navigation device detects an abnormally large steering wheel angle and the vehicle is in a critical area of ​​the ramp, the turn signal is activated.

[0205] Specifically, when the navigation device determines When making a large steering wheel turn and being at a ramp or intersection, the unusually large steering wheel angle may indicate that the driver is hesitant or has taken the wrong turn. The vision system is immediately triggered to prioritize recognizing the road sign text and lane directional arrows and compare them with the expected route in the navigation.

[0206] in, This indicates a ramp geofence.

[0207] If the turn signal in the vehicle status information is active, then it is determined that the vehicle is responding to the vehicle status trigger.

[0208] It should be noted that the turn signal being active means that the left or right turn signal is confirmed to be turned on (status value ON) by parsing the TurnIndicators signal on the CAN bus. When the navigation device detects that the turn signal is active, the turn signal is considered active.

[0209] Specifically, when the navigation device determines When the turn signal is activated, it is treated as a clear signal of the driver's intention to change lanes or turn, triggering the vision system to perform blind spot monitoring and lane line type (real and virtual) recognition of the target lane.

[0210] In one possible implementation, the method further includes the following step prior to step C01:

[0211] Obtain navigation data, and based on the vehicle location coordinates in the navigation data, determine whether the vehicle is within an area of ​​a preset number of preset geofences or greater.

[0212] Understandably, preset geofences include intersections of main and auxiliary roads, areas on and under bridges, areas with dense traffic lights, and areas with temporary traffic signs. The number of preset geofences can be set to two, meaning that entering two or more geofences simultaneously will trigger the navigation system. The navigation device determines whether the vehicle is within the spatial range of multiple preset geofences based on its location.

[0213] Specifically, when a POI geofence is triggered, the navigation POI system constructs and dynamically updates an electronic fence for high-confidence demand areas through multi-source data fusion. The data sources include high-precision maps with centimeter-level accuracy (including RoadGraph road topology), traffic light data from traffic management departments, temporary traffic control information, and crowdsourced data that has been cleaned and verified through machine learning.

[0214] For different scenarios, the POI system and the navigation system interact using an event-driven architecture. The navigation system subscribes to POI data updates via the Internet of Things communication protocol (MQTT) and calculates the vehicle's position in real time based on the geographic coordinate system standard (WGS84) coordinates. With electronic fence Relationship:

[0215] Therefore, the system is triggered when it detects that a vehicle has entered a specific geofence, such as at the intersection of main and auxiliary roads, on or under bridges, in areas with dense traffic lights, or in areas with temporary traffic signs.

[0216] If it is in the preset geofence, the weights corresponding to the preset geofence are summed to obtain the comprehensive weight;

[0217] It should be noted that when the navigation device determines that the current vehicle is within the spatial range of two or more geofences, it will use a weighted summation mechanism to quantify and superimpose the impact of multiple navigation offset scenarios that the vehicle may be in at the same time, and obtain a comprehensive weight.

[0218] If the overall weight is greater than the weight threshold, then the vehicle response geofence is determined to be triggered.

[0219] Understandably, the weight threshold is a pre-set trigger limit value used to determine whether the overall weight has reached a level that requires initiating visual perception; for example, it might be set to 0.7. When the navigation device determines that the overall weight is greater than the weight threshold, it determines that the vehicle response should be triggered by the geofence.

[0220] For example, the navigation device assigns trigger weights to each POI area. For example, main and auxiliary roads 0.4, bridges and underpasses 0.3, traffic lights 0.2, temporary signs 0.1;

[0221]

[0222] When the navigation device determines that multiple areas are superimposed, the comprehensive trigger value > 0.7 triggers visual perception, ultimately achieving efficient and accurate visual perception triggering.

[0223] Furthermore, the navigation device employs a triple-trigger integrated decision-making process, with the final triggering conditions being:

[0224]

[0225] That is, the visual compensation module will be activated when any one of the three conditions is met. Timed polling trigger, Vehicle status trigger, Geofencing triggered.

[0226] In this embodiment, the navigation device triggers the sensor to input data through a triple triggering mechanism. The device will work as long as one of the three mechanisms is satisfied, which avoids unnecessary consumption of system resources in secondary scenarios and improves the timeliness of data acquisition.

[0227] Based on the first embodiment of this application, in the fourth embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description, and will not be repeated hereafter. On this basis, refer to Figure 6 The navigation method further includes steps D01-D04: adjusting the vehicle's navigation decision based on the adjusted weighting and the structured data.

[0228] Step D01: Obtain a predefined set of scene states;

[0229] It should be noted that the scene state set refers to a complete enumeration of all possible vehicle navigation scenarios defined in advance. By establishing a complete scene state set in advance, the navigation device provides a clear target judgment space for subsequent multi-source data fusion decision-making.

[0230] Step D02: Based on the adjusted weight ratio and the structured data, determine the confidence level of the vehicle in different scenario states in the scenario state set;

[0231] Understandably, navigation devices calculate confidence levels for different scenario states based on the weighting of multi-source data in a multi-source data fusion decision model and the features in structured data.

[0232] Step D03: Based on the confidence level, determine the target scene state in which the vehicle is located;

[0233] It should be noted that the navigation device selects the scenario state with the highest confidence level as the target scenario state in which the vehicle is located.

[0234] Step D04: Based on the target scene state, compensate the vehicle's navigation decision.

[0235] It is understandable that the navigation device compensates for the original navigation results based on a single GPS data according to the final determined target scene state.

[0236] In one feasible implementation, step D03 includes:

[0237] The visual score is determined based on the number of overlaps between each feature in the structured data and the visual features corresponding to different scene states in the scene state set.

[0238] It should be noted that visual features are a predefined set of key visual characteristics that serve as typical visual evidence for determining whether a vehicle is in a certain state. The navigation device determines a visual score based on the number of predefined key visual features that match the currently identified structured data features.

[0239] Specifically, It is the evidence score, or visual score, based on AI visual recognition. It is a linear combination of multiple visual feature functions: in, It is the first An indicator function for each visual feature (1 if detected, 0 otherwise). Is this feature useful for determining the state? Importance weights.

[0240] For example, in the decision-making process between main and auxiliary roads / elevated bridges, when the state space... For the state "on the bridge", the visual evidence score is... The main characteristics determining this are: f1 detects the viaduct guardrail, f2 detects the gantry-type highway sign, and f3 provides a wide field of view with no upper obstructions. The weights of these characteristics are... If both are positive, and the GPS signal from the navigation system is displayed under the bridge, but the vision system detects the guardrail and highway sign, then... This will result in a high score, making the final Exceed The system then determines that the vehicle is on the bridge.

[0241] For traffic light countdown decisions, a visual model identifies the color status of the traffic light at that moment, and the model's OCR capability identifies the numbers on the countdown sign. Alternatively, in the absence of a number plate, the remaining time can be estimated by analyzing a sequence of consecutive image frames showing the switching of traffic light states.

[0242] For temporary road condition decisions, the visual model recognizes... The decision-making system then matches this information with the navigation API, triggering a route replanning instruction.

[0243] Based on the navigation data, the degree of matching between the vehicle's position coordinates and the map network corresponding to the different scene states is calculated to obtain a positioning score;

[0244] Understandably, a map network refers to the road network topology contained in a high-precision map, where each road has precise geographic coordinates and attribute information (such as road type, number of lanes, etc.). The matching degree is measured by calculating the lateral distance from the vehicle's current GPS coordinates to the road corresponding to the candidate state; the closer the distance, the higher the matching degree. The navigation device determines the positioning score based on the lateral distance from the vehicle's current GPS coordinates to the road corresponding to the candidate state.

[0245] Specifically, It is an evidence score based on GPS positioning, namely the positioning score, which measures the current GPS coordinates and status. The degree of matching with the corresponding map road network. For example, it can be based on the lateral distance from the GPS location point to the corresponding lane. The closer the distance, the higher the score.

[0246] Based on the adjusted weight ratio, the positioning score and the visual score are weighted and summed to obtain the confidence level under different scene states.

[0247] It should be noted that the navigation device uses dynamically adjusted weight ratios to perform weighted fusion of positioning scores and visual scores to obtain the final confidence level of different candidate scene states.

[0248] Specifically, the confidence score function is defined as:

[0249]

[0250] In the formula, and These are the weighting coefficients for GPS information and visual information, respectively, which can be dynamically adjusted based on GPS signal strength (such as the number of satellites and HDOP value). For example, in areas with weak GPS signals, the weighting coefficients can be dynamically reduced. ,improve .

[0251] Furthermore, the navigation device defines a scene state space. Each state This represents a possible vehicle navigation scenario, for example. "Driving on the elevated highway" "Driving under the overpass" "Driving on the main road." At any time The system deals with each possible state. Calculate a confidence score The highest score is the system's final decision. ,in, .

[0252] In this embodiment, a confidence assessment model that can adapt to scene changes is constructed by using dynamically adjusted weights to weight and fuse the positioning score and the visual score.

[0253] In one feasible implementation, the preset navigation inaccuracy scenario includes at least one of the following: starting positioning scenario, main and auxiliary road scenario, bridge and under-bridge scenario, and traffic light scenario.

[0254] Specifically, the starting positioning scenario refers to the situation where, when a vehicle begins to move from a stationary state, the positioning is ambiguous due to the unstable GPS signal or inaccurate initial direction determination. The main-auxiliary road scenario refers to the situation where, when a vehicle is traveling on a road with parallel main and auxiliary roads, GPS horizontal errors prevent the system from accurately determining whether the vehicle is on the main or auxiliary road. The bridge-on / under-bridge scenario refers to the situation where, when a vehicle is traveling on a grade-separated road (such as an overpass or interchange), insufficient GPS vertical accuracy prevents the system from distinguishing whether the vehicle is on or under the bridge. The traffic light scenario refers to the situation where, when a vehicle approaches or is at a traffic light-controlled intersection, dynamic information such as the traffic light status and countdown timer is needed to optimize navigation.

[0255] Based on the first embodiment of this application, in the fifth embodiment of this application, the content that is the same as or similar to that in the first embodiment described above can be referred to the above description and will not be repeated hereafter. Based on this, after step S30, the navigation method further includes step E01:

[0256] Step E01: Return the adjusted navigation decision to the visual interface and display it via voice and text.

[0257] It should be noted that the navigation device will return the adjusted navigation decision to the vehicle's visual interface through the navigation APP's transparent API interface, and combine voice and text for proactive human-computer interaction.

[0258] In one feasible implementation, step E01 includes:

[0259] The adjusted navigation is returned to the visualization interface, and the vehicle icons in the visualization interface are automatically updated based on the location information. The dynamic traffic information is then overlaid and displayed on the visualization interface.

[0260] It is understood that the visual interface refers to the central control screen or integrated navigation display interface of the in-vehicle infotainment system. Dynamic traffic information refers to real-time traffic elements obtained through AI visual recognition, including but not limited to traffic light countdowns, temporary traffic signs (such as construction warnings and temporary speed limits), and lane-level guidance information. The navigation device returns the adjusted navigation to the visual interface and, based on the precise location information corresponding to the final determined target scene state, corrects the display position of vehicle icons and road levels on the map in real time, and overlays them on the base map as layers.

[0261] This implementation integrates AI visual recognition results with navigation data, enabling the vehicle icon to be accurately marked on the main road, auxiliary road, or on / under a bridge in the visual interface. At the same time, the countdown information of the traffic lights ahead and the identified temporary traffic signs (construction signs, temporary speed limit signs, road detour prompts, etc.) are also displayed in the navigation interface as overlay layers, thereby achieving intuitive feedback on environmental information. Furthermore, the "manual confirmation" button that requires manual confirmation of the current navigation in related technologies is replaced by the autonomous decision-making of the navigation device.

[0262] In one feasible implementation, step E01 includes:

[0263] The navigation is sent to the enhanced display module, which then projects information corresponding to the display mode onto the driver's field of vision based on the navigation, wherein the display mode is determined based on the driving scenario.

[0264] It's important to note that the augmented display module refers to the vehicle's head-up display system, particularly those with augmented reality (AR-HUD) capabilities. The information displayed in the projection mode refers to the information content and presentation format dynamically selected by the HUD based on the current driving scenario. For example, in simplified mode, only lane guidance arrows and basic speed limit information are displayed, while in augmented mode, additional road layers (such as on / under bridges), traffic light countdowns, and AR icons for temporary traffic signs are overlaid. Navigation based on driving scenario means that the system automatically switches the level of detail and display style of the HUD information based on real-time perceived parameters such as road type, traffic density, and vehicle speed, intelligently projecting the information corresponding to the display mode within the driver's primary field of vision. This scenario-based adaptive display mode mechanism ensures that the most appropriate amount of information is provided under different driving needs (such as simplified information for high-speed driving and enhanced guidance for complex intersections), effectively reducing the driver's cognitive load.

[0265] For example, the navigation device can implement HUD enhanced display. In a cockpit environment that supports head-up display (HUD), the final navigation result is synchronously projected onto the forward field of view. The HUD can highlight the target lane boundary, on / off road level prompts, and traffic light countdowns in augmented reality mode, and provide multiple display modes. For example, the simplified mode only outputs route and lane guidance, while the enhanced mode overlays more temporary traffic elements. Users can switch display modes through voice control or interaction with the central control screen, thereby achieving flexible adaptation to different driving needs.

[0266] In one feasible implementation, step E01 includes:

[0267] The navigation is sent to the voice playback module so that the voice playback module can play the voice information corresponding to the navigation.

[0268] As is understandable, the voice playback module refers to the audio output unit of the in-vehicle system, which is usually integrated with the in-vehicle audio system. The navigation device actively conveys navigation information to the driver through voice broadcast, realizing the auditory transmission of key navigation information.

[0269] For example, navigation devices can implement a proactive interaction mechanism, actively outputting information to the driver by combining voice broadcasts and text prompts when triggered in key scenarios. For instance, when the vehicle's location is corrected from visual recognition to the main road, the system automatically outputs a confirmation prompt; in the case of a traffic light near an intersection, the system broadcasts the remaining countdown and provides braking advice; the system uses voice broadcasts such as "You have been corrected to the main road, please drive with confidence," "Red light ahead, we recommend smooth braking," and "Temporary construction detected ahead, a new route has been planned for you," achieving an experience upgrade from "passive query" to "proactive notification." When a temporary construction section is identified, the system automatically pushes a new planned route and provides a prompt. Through the above proactive interaction mechanism, the driver can obtain the decision result without additional operation, avoiding the traditional manual confirmation interaction steps.

[0270] In this implementation, a redundancy mechanism is established between voice, text, and HUD display, significantly optimizing the human-computer interaction experience and navigation reliability. When in-vehicle noise is high, making voice prompts difficult to perceive, the system automatically enhances the prompts on the central control screen and HUD. When external light affects the visibility of the HUD projection, the voice prompts automatically increase in volume and are supplemented with high-contrast screen prompts. Through multimodal redundancy design, the final navigation result is ensured even in complex driving environments. It can be reliably obtained by the driver.

[0271] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the navigation method of this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0272] This application also provides a navigation device, please refer to... Figure 7 The navigation device includes:

[0273] The acquisition module 10 is used to acquire multi-source data and driving scenario information corresponding to the vehicle, wherein the multi-source perception data includes at least navigation data and traffic sign data;

[0274] The weight adjustment module 20 is used to adjust the weight ratio of each data source in the multi-source data in the navigation based on the driving scenario information.

[0275] The semantic understanding module 30 is used to perform semantic understanding on the traffic sign data to obtain structured data;

[0276] The navigation adjustment module 40 is used to adjust the vehicle's navigation decisions based on the adjusted weight ratios and the structured data.

[0277] Optionally, the semantic understanding module includes:

[0278] The output submodule is used to input the traffic sign data into a preset semantic understanding model, perform semantic understanding on the traffic sign data through the preset semantic understanding model, and output structured data, wherein the structured data includes at least one of the road physical attributes, road type attributes, and road function information.

[0279] Optionally, the output submodule includes:

[0280] The first feature extraction unit is used to input the traffic sign data into a preset semantic understanding model, and to perform visual feature extraction and text recognition on the traffic sign data through the preset semantic understanding model to obtain visual features and text features. The text features include traffic-related text information parsed from the traffic sign data. The text features and the visual features are linked together to output structured data.

[0281] The second feature extraction unit is used to input the traffic sign data into a preset semantic understanding model, and to perform multimodal structured feature extraction on the traffic sign data through the preset semantic understanding model to obtain visual features and traffic element features. The traffic element features include features corresponding to traffic-related detection objects detected from the traffic sign data. The visual features and the traffic element features are linked together to output structured data.

[0282] The compensation unit is used to acquire a predefined set of scene states; determine the confidence level of the vehicle in different scene states in the set of scene states based on the adjusted weight ratio and the structured data; determine the target scene state of the vehicle based on the confidence level; and compensate the vehicle's navigation decision based on the target scene state.

[0283] Optionally, the acquisition module includes:

[0284] The active sending submodule is used to actively send a resource call request to the target sensor when the vehicle is detected to be in a preset navigation inaccuracy scenario, so that the target sensor can collect traffic sign data based on the resource call request and transmit the traffic sign data to the local device through a standardized interface; and receive the traffic sign data sent by the target sensor.

[0285] Optionally, the active sending submodule includes:

[0286] The triggering submodule is used to determine that the vehicle is in a preset navigation inaccuracy scenario when the vehicle meets any preset triggering mechanism. The preset triggering mechanisms include timed polling triggering, vehicle status triggering, and geofence triggering.

[0287] The resource determination unit is used to determine the target sensor from the preset visual sensor resources based on the target scene in the preset navigation inaccuracy scenario when the vehicle is detected to be in a preset navigation inaccuracy scenario, and actively send a resource call request to the target sensor.

[0288] Optionally, the resource determination unit includes:

[0289] The scene invocation subunit is used to determine the first sensor from the preset visual sensor resources and actively send a resource invocation request to the first sensor when the target scene of the preset navigation inaccuracy scene is a traffic light scene, wherein the first sensor includes a forward-looking camera; or, when the preset navigation inaccuracy scene is a main and auxiliary road scene, determine the second sensor from the preset visual sensor resources and actively send a resource invocation request to the second sensor, wherein the second sensor includes a forward-looking camera and a fisheye camera.

[0290] Optionally, the triggering submodule includes:

[0291] A timed polling triggering unit is used to determine that the vehicle responds to a timed polling trigger when the vehicle meets the polling frequency of the current polling mode. The polling mode includes cruise mode, city mode and standby mode, and the polling frequency of the timed polling trigger is determined based on the road type on which the vehicle is driving.

[0292] The vehicle status triggering unit is used to acquire vehicle status information and navigation data when the navigation application is running; and to determine the vehicle's response to the vehicle status trigger based on the vehicle status information and the navigation data.

[0293] The geofence triggering unit is used to acquire navigation data and, based on the vehicle location coordinates in the navigation data, determine whether the vehicle is within an area of ​​a preset number of preset geofences; if it is, the weights corresponding to the preset geofences are summed to obtain a comprehensive weight; if the comprehensive weight is greater than a weight threshold, the vehicle is determined to respond to the geofence trigger.

[0294] Optionally, the timed polling triggering unit includes:

[0295] The polling mode determination subunit is used to determine the road type on which the vehicle is traveling; if the road type is a stable driving type, the polling mode triggered by the timed polling is set to cruise mode; if the road type is a complex driving type, the polling mode triggered by the timed polling is set to city mode; if the road type is a stationary driving type, the polling mode triggered by the timed polling is set to standby mode.

[0296] Optionally, the vehicle status triggering unit includes:

[0297] The vehicle status triggering subunit is used to determine that the vehicle responds to the vehicle status trigger if the vehicle speed in the vehicle status information is within a preset threshold range and the vehicle is determined to be within the area of ​​the traffic light geofence based on the navigation data.

[0298] The vehicle status triggering subunit is used to determine that the vehicle responds to the vehicle status trigger if the steering wheel angle in the vehicle status information is greater than a preset angle threshold, and the vehicle is determined to be within the area of ​​the ramp geofence based on the navigation data.

[0299] The vehicle status triggering subunit is used to determine that the vehicle responds to the vehicle status trigger if the turn signal in the vehicle status information is in an active state.

[0300] Optionally, the compensation unit includes:

[0301] The confidence determination subunit is used to determine the visual score based on the number of overlaps between each feature in the structured data and the visual features corresponding to different scene states in the scene state set; based on the navigation data, it calculates the degree of matching between the vehicle position coordinates and the map network corresponding to the different scene states to obtain the positioning score; based on the adjusted weight ratio, it performs a weighted summation of the positioning score and the visual score to obtain the confidence score under the different scene states.

[0302] The navigation device provided in this application, employing the navigation method described in the above embodiments, can solve the technical problems of navigation. Compared with the prior art, the beneficial effects of the navigation device provided in this application are the same as those of the navigation method provided in the above embodiments, and other technical features in the navigation device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0303] This application provides a navigation device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the navigation method in Embodiment 1 above.

[0304] The following is for reference. Figure 8 The diagram illustrates a structural schematic of a navigation device suitable for implementing embodiments of this application. The navigation device in these embodiments may include, but is not limited to, mobile terminals such as mobile phones, laptops, tablets, digital broadcast receivers, PDAs (Personal Digital Assistants), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 8 The navigation device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0305] like Figure 8As shown, the navigation device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the navigation device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touchscreen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows the navigation device to communicate wirelessly or wiredly with other devices to exchange data. Although navigation devices with various systems are shown in the figures, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0306] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0307] The navigation device provided in this application, employing the navigation method described in the above embodiments, can solve the technical problems of navigation. Compared with the prior art, the beneficial effects of the navigation device provided in this application are the same as those of the navigation method provided in the above embodiments, and other technical features of the navigation device are the same as those disclosed in the method of the previous embodiment, and will not be repeated here.

[0308] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0309] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0310] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to perform the navigation method in the above embodiments.

[0311] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0312] The aforementioned computer-readable storage medium may be included in the navigation device; or it may exist independently and not be assembled into the navigation device.

[0313] The aforementioned computer-readable storage medium carries one or more programs that, when executed by a navigation device, cause the navigation device to: acquire multi-source data and driving scenario information corresponding to the vehicle, wherein the multi-source perception data includes at least navigation data and traffic sign data; adjust the weight ratio of each data source in the multi-source data in the navigation based on the driving scenario information; and adjust the vehicle's navigation decision based on the adjusted weight ratio and the structured data.

[0314] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0315] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0316] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0317] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described navigation method, thereby solving the technical problem of navigation. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the navigation method provided in the above embodiments, and will not be repeated here.

[0318] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the navigation method described above.

[0319] The computer program product provided in this application can solve the technical problem of navigation. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as the beneficial effects of the navigation method provided in the above embodiments, and will not be repeated here.

[0320] The above description is only a part of the embodiments of this application and does not limit the scope of protection of this application. All equivalent structural transformations made under the technical concept of this application and using the content of this application specification and drawings, or direct / indirect applications in other related technical fields, are included in the scope of protection of this application.

Claims

1. A navigation method, characterized in that, The navigation method includes: Acquire multi-source data and driving scenario information corresponding to the vehicle, wherein the multi-source data includes at least navigation data and traffic sign data; Based on the driving scenario information, adjust the weight ratio of each data source in the multi-source data in the navigation; The traffic sign data is input into a preset semantic understanding model. The preset semantic understanding model performs semantic understanding on the traffic sign data and outputs structured data. The structured data includes at least one of the road's physical attributes, road type attributes, and road function information. The vehicle's navigation decisions are adjusted based on the adjusted weighting and the structured data. The step of inputting the traffic sign data into a preset semantic understanding model, performing semantic understanding on the traffic sign data through the preset semantic understanding model, and outputting structured data includes: The traffic sign data is input into a preset semantic understanding model. The preset semantic understanding model is used to extract visual features and recognize text in the traffic sign data to obtain visual features and text features. The text features include traffic-related text information parsed from the traffic sign data. The text information is used to lock the road level. The text features and the visual features are linked together to output structured data.

2. The navigation method as described in claim 1, characterized in that, The step of inputting the traffic sign data into a preset semantic understanding model, performing semantic understanding on the traffic sign data through the preset semantic understanding model, and outputting structured data includes: The traffic sign data is input into a preset semantic understanding model. Through the preset semantic understanding model, multimodal structured feature extraction is performed on the traffic sign data to obtain visual features and traffic element features. The traffic element features include features corresponding to traffic-related detection objects detected from the traffic sign data. The visual features and traffic element features are linked together to output structured data.

3. The navigation method as described in claim 1, characterized in that, The steps for acquiring multi-source data and driving scenario information corresponding to the vehicle include: When a vehicle is detected to be in a preset navigation inaccuracy scenario, a resource request is actively sent to the target sensor, so that the target sensor can collect traffic sign data based on the resource request and transmit the traffic sign data to the local device through a standardized interface; Receive the traffic sign data sent by the target sensor.

4. The navigation method as described in claim 3, characterized in that, The step of actively sending a resource request to the target sensor when the vehicle is detected to be in a preset navigation inaccuracy scenario includes: When a vehicle is detected to be in a preset navigation failure scenario, a target sensor is determined from preset visual sensor resources based on the target scene in the preset navigation failure scenario in which the vehicle is located, and a resource call request is actively sent to the target sensor.

5. The navigation method as described in claim 4, characterized in that, The target sensor includes a first sensor and a second sensor. The step of determining the target sensor from preset visual sensor resources based on the target scene in the preset navigation inaccuracy scenario in which the vehicle is located, and actively sending a resource access request to the target sensor includes: When the target scenario of the preset navigation misalignment scenario is a traffic light scenario, the first sensor is determined from the preset visual sensor resources, and a resource call request is actively sent to the first sensor, the first sensor including a forward-looking camera; or, When the preset navigation inaccuracy scenario is a main and auxiliary road scenario, the second sensor is determined from the preset visual sensor resources, and a resource call request is actively sent to the second sensor. The second sensor includes a forward-looking camera and a fisheye camera.

6. The navigation method as described in claim 3, characterized in that, Before the step of actively sending a resource request to the target sensor when the vehicle is detected to be in a preset navigation inaccuracy scenario, the following steps are included: When a vehicle meets any of the preset triggering mechanisms, it is determined that the vehicle is in a preset navigation inaccuracy scenario. The preset triggering mechanisms include timed polling triggering, vehicle status triggering, and geofence triggering.

7. The navigation method as described in claim 6, characterized in that, Before the step of determining that the vehicle is in a preset navigation inaccuracy scenario when the vehicle meets any preset triggering mechanism, the following steps are included: When the vehicle meets the polling frequency of the current polling mode, it is determined that the vehicle responds to the timed polling trigger. The polling mode includes at least one of cruise mode, city mode and standby mode. The polling frequency of the timed polling trigger is determined based on the road type on which the vehicle is driving.

8. The navigation method as described in claim 7, characterized in that, Before the step of determining that the vehicle responds to the timed polling trigger when the vehicle meets the polling frequency of the current polling mode, the following steps are included: Determine the type of road the vehicle is traveling on; If the road type is a stable driving type, then the polling mode triggered by the timed polling is set to cruise mode; If the road type is a complex driving type, then the polling mode triggered by the timed polling is set to city mode; If the road type is a stationary driving type, then the polling mode triggered by the timed polling is set to standby mode.

9. The navigation method as described in claim 6, characterized in that, Before the step of determining that the vehicle is in a preset navigation inaccuracy scenario when the vehicle meets any preset triggering mechanism, the method further includes: When the navigation application is running, it acquires vehicle status information and navigation data; Based on the vehicle status information and the navigation data, the vehicle response is triggered.

10. The navigation method as described in claim 9, characterized in that, The step of determining the vehicle's response to the vehicle status trigger based on the vehicle status information and the navigation data includes at least one of the following: If the vehicle speed in the vehicle status information is within a preset threshold range, and based on the navigation data it is determined that the vehicle is within the area of ​​the traffic light geofence, then it is determined that the vehicle responds to the vehicle status trigger. If the steering wheel angle in the vehicle status information is greater than a preset angle threshold, and based on the navigation data it is determined that the vehicle is within the area of ​​the ramp geofence, then it is determined that the vehicle responds to the vehicle status trigger. If the turn signal in the vehicle status information is active, then it is determined that the vehicle is responding to the vehicle status trigger.

11. The navigation method as described in claim 6, characterized in that, Before the step of determining that the vehicle is in a preset navigation inaccuracy scenario when the vehicle meets any preset triggering mechanism, the method further includes: Obtain navigation data, and based on the vehicle location coordinates in the navigation data, determine whether the vehicle is within an area of ​​a preset number of preset geofences or greater. If it is in the preset geofence, the weights corresponding to the preset geofence are summed to obtain the comprehensive weight; If the overall weight is greater than the weight threshold, then the vehicle response geofence is determined to be triggered.

12. The navigation method as described in claim 1, characterized in that, The step of adjusting the vehicle's navigation decision based on the adjusted weighting and the structured data includes: Get a predefined set of scene states; Based on the adjusted weighting and the structured data, the confidence level of the vehicle in different scenario states within the scenario state set is determined. Based on the confidence level, the target scene state of the vehicle is determined; The vehicle's navigation decisions are compensated based on the target scene state.

13. The navigation method as described in claim 12, characterized in that, The step of determining the confidence level of the vehicle in different scenario states within the scenario state set based on the adjusted weighting and the structured data includes: The visual score is determined based on the number of overlaps between each feature in the structured data and the visual features corresponding to different scene states in the scene state set. Based on the navigation data, the degree of matching between the vehicle's position coordinates and the map network corresponding to the different scene states is calculated to obtain a positioning score; Based on the adjusted weight ratio, the positioning score and the visual score are weighted and summed to obtain the confidence level under different scene states.

14. A navigation device, characterized in that, The device includes: The acquisition module is used to acquire multi-source data and driving scenario information corresponding to the vehicle, wherein the multi-source data includes at least navigation data and traffic sign data; The weight adjustment module is used to adjust the weight ratio of each data source in the multi-source data in the navigation based on the driving scenario information. The semantic understanding module includes: The output submodule is used to input the traffic sign data into a preset semantic understanding model, perform semantic understanding on the traffic sign data through the preset semantic understanding model, and output structured data, wherein the structured data includes at least one of the road physical attributes, road type attributes, and road function information; The navigation adjustment module is used to adjust the vehicle's navigation decisions based on the adjusted weight ratios and the structured data. The output submodule includes: The first feature extraction unit is used to input the traffic sign data into a preset semantic understanding model, and to extract visual features and recognize text through the preset semantic understanding model to obtain visual features and text features. The text features include traffic-related text information parsed from the traffic sign data, and the text information is used to lock the road level. The text features and the visual features are linked together to output structured data.

15. A navigation device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the navigation method as described in any one of claims 1 to 13.

16. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the navigation method as described in any one of claims 1 to 13.

17. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the navigation method as described in any one of claims 1 to 13.

Images