A vehicle blind area early warning method based on multi-source information fusion and risk grading

Abstract

The application provides a vehicle blind area early warning method based on multi-source information fusion and risk grading, wherein the method comprises the following steps: determining the collision risk index between the target to be evaluated and each other target in the blind area range of the target to be evaluated based on the state data of the target to be evaluated and other targets in the blind area scene data model; determining the dynamic risk threshold corresponding to the current moment based on the environment data in the blind area scene data model; and determining the risk level of the target to be evaluated according to the size relationship between the collision risk index and the dynamic risk threshold of the current moment, and executing the corresponding early warning measures.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, and more specifically, to a vehicle blind spot warning method based on multi-source information fusion and risk classification. Background Technology

[0002] Current solutions mainly rely on vehicle-mounted intelligent sensing or basic roadside unit broadcasting, but significant technical bottlenecks remain in complex weather conditions and the depth of multi-source information fusion. Specifically, vehicle-mounted intelligent sensing systems are limited by physical line-of-sight obstruction and cannot effectively detect potential targets behind curves or in blind spots at intersections; while traditional roadside facilities (such as convex mirrors and warning signs) can only provide static prompts and cannot reflect real-time traffic flow dynamics at the millisecond level.

[0003] Meanwhile, existing vehicle-road cooperative systems mostly adopt a simple data pass-through mode, failing to fully realize the deep integration of roadside perception, vehicle-mounted perception, and environmental information. Multi-sensor data fusion often remains at the data level, resulting in a lack of fault tolerance and low information reliability when a single sensor fails (such as visual failure in rainy or foggy weather). Summary of the Invention

[0004] The purpose of this application is to provide a vehicle blind spot warning method based on multi-source information fusion and risk classification, so as to solve the technical problems of poor perception performance and inaccurate prediction results in existing vehicle-road cooperative systems.

[0005] In a first aspect, the present invention provides a vehicle blind spot warning method based on multi-source information fusion and risk classification. The method includes determining the collision risk index between the target to be evaluated and each other target at the current moment based on the state data of the target to be evaluated and other targets located within the blind spot range of the target to be evaluated in the blind spot scene data model; determining the dynamic risk threshold at the current moment based on the environmental data in the blind spot scene data model; determining the risk level of the target to be evaluated according to the magnitude relationship between the collision risk index at the current moment and the dynamic risk threshold, and executing corresponding warning measures.

[0006] In an optional implementation, the collision risk index between the target to be evaluated and other targets is determined by the following method. : ; in, , These are the weighting coefficients. The baseline time for collision between the target to be evaluated and other targets. The collision margin time between the target to be evaluated and other targets. This is the error coefficient.

[0007] In an optional implementation, the state data includes at least the position coordinates in the physical coordinate system, the movement speed, and the collision reference time between the target to be evaluated and other targets, based on the angle between the movement directions of the target and other targets, determined in the following manner: Based on the position coordinates of the target to be evaluated and other targets, the relative distance between the target to be evaluated and other targets is calculated; Based on the moving speed and the angle between the target to be evaluated and other targets, the relative approach speed between the target to be evaluated and other targets is calculated; Calculate the ratio between the relative distance and the relative approach speed, and use it as the collision reference time.

[0008] In an optional implementation, the step of calculating the relative approach speed between the target to be evaluated and other targets based on their moving speeds and the angle between their directions of movement specifically includes: Calculate the product between the moving speed of other targets and the angle between the moving directions of the target to be evaluated and other targets; Calculate the sum of the moving speed of the target to be evaluated and its product, as the relative approach speed.

[0009] In an optional implementation, the environmental data includes at least the real-time road surface adhesion coefficient, and the dynamic risk threshold corresponding to the current moment is determined by the following method: Calculate the ratio between the dry road surface adhesion coefficient and the real-time road surface adhesion coefficient; The dynamic risk threshold at the current moment is determined by multiplying the ratio by the baseline risk threshold.

[0010] In an optional implementation, the target's first moving speed at the first acquisition time is collected based on GPS, and the target's second moving speed at the second acquisition time is collected based on BeiDou satellite data. Both the first and second acquisition times are adjacent to the current time. The target's current speed is determined as follows: Calculate the speed difference between the first and second movement speeds; Calculate the first time difference between the current time and the first acquisition time; Calculate the second time difference between the second acquisition time and the first acquisition time; Calculate the time difference ratio between the first time difference and the second time difference; Calculate the product of the time difference ratio and the difference in movement speed; Calculate the sum of the product and the first moving speed, and use it as the moving speed at the current moment.

[0011] In an optional implementation, the position coordinates of the target to be evaluated in the physical coordinate system are determined by the following method. :

[0012] in, This refers to the position coordinates of the target in the sensing coordinate system, based on data collected by the positioning sensor. To locate the latitude and longitude coordinates of the sensor, Based on device orientation angle The rotation matrix.

[0013] Secondly, the present invention provides a vehicle blind spot warning device based on multi-source information fusion and risk classification, the device comprising: The first processing module is used to determine the collision risk index between the target to be evaluated and each other target at the current moment based on the state data of the target to be evaluated and other targets within the blind zone of the target to be evaluated in the blind zone scene data model. The second processing module determines the dynamic risk threshold corresponding to the current moment based on the environmental data in the blind spot scene data model. The assessment module is used to determine the risk level of the target to be assessed based on the relationship between the current collision risk index and the dynamic risk threshold, and to execute corresponding early warning measures.

[0014] Thirdly, the present invention provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus, and the processor executes the machine-readable instructions to perform the steps of any of the vehicle blind spot warning methods based on multi-source information fusion and risk classification as described in the foregoing embodiments.

[0015] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of any of the vehicle blind spot warning methods based on multi-source information fusion and risk classification as described in the foregoing embodiments.

[0016] This application provides a vehicle blind spot warning method based on multi-source information fusion and risk classification. The method includes determining the collision risk index between the target to be evaluated and each other target within the blind spot range of the target to be evaluated based on the state data of the target to be evaluated and other targets within the blind spot range of the target to be evaluated in the blind spot scene data model at the current moment; determining the dynamic risk threshold at the current moment based on the environmental data in the blind spot scene data model; determining the risk level of the target to be evaluated based on the relationship between the collision risk index at the current moment and the dynamic risk threshold, and executing corresponding warning measures. This application achieves all-weather, high-precision blind spot safety warning service through deep fusion of roadside perception, structured reconstruction of blind spot situation, and environmentally adaptive dynamic risk assessment. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a vehicle blind spot warning method based on multi-source information fusion and risk classification, provided for embodiments of this application; Figure 2 A schematic diagram of the structure of a vehicle blind spot warning device based on multi-source information fusion and risk classification provided in this application embodiment; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0019] Current solutions primarily rely on vehicle-to-vehicle intelligent sensing or basic roadside unit broadcasting, but significant technical bottlenecks remain in complex weather conditions and the depth of multi-source information fusion. Specifically, existing technologies suffer from the following main shortcomings: 1. Perceptual abilities have physical limitations and blind spots. Intelligent sensing systems for single vehicles are limited by physical line-of-sight obstruction and cannot effectively detect potential targets behind curves or in blind spots at intersections; while traditional roadside facilities (such as convex mirrors and warning signs) can only provide static prompts and cannot reflect real-time traffic flow dynamics at the millisecond level.

[0020] 2. Insufficient depth of multi-source heterogeneous information fusion Existing vehicle-road cooperative systems mostly adopt a simple data pass-through mode, failing to fully realize the deep integration of roadside perception, vehicle perception, and environmental information. Multi-sensor data fusion often remains at the data level, lacking decision-level deep fusion based on high-precision spatiotemporal references. This results in the system lacking fault tolerance and low information reliability when a single sensor fails (such as visual failure in rainy or foggy weather).

[0021] 3. The early warning strategy is rigid and lacks environmental adaptability. Existing warning algorithms typically use fixed distance or time thresholds, failing to consider the decisive impact of road conditions on vehicle braking performance. Especially in icy and snowy road conditions in northern seasonally frozen regions, the road adhesion coefficient is significantly reduced. If warning thresholds for conventional roads are still used, it is highly likely to cause warning delays, leading to rear-end collisions or other accidents.

[0022] 4. The interaction methods are fragmented and highly disruptive. The lack of comprehensive situational awareness and structured representation of traffic conditions within blind spots results in fragmented information being delivered to drivers, failing to provide them with an intuitive and complete "God's-eye view" of the blind spots. Furthermore, the absence of intelligent suppression mechanisms for repeated alarms leads to frequent invalid alarms causing driver desensitization and reducing the system's usability.

[0023] Based on this, this application provides a vehicle blind spot warning method based on multi-source information fusion and risk classification.

[0024] The technical solutions in the embodiments of this application will now be described with reference to the accompanying drawings.

[0025] Example 1 This paper presents a vehicle blind spot warning method based on multi-source information fusion and risk classification. Through deep fusion of roadside perception, structured reconstruction of blind spot situation, and dynamic risk assessment adapted to the environment, it achieves all-weather, high-precision blind spot safety services. Specific technical solutions include: Step 1: Decision-level fusion of multi-source heterogeneous information To address the fusion error caused by asynchronous sampling from different sensors, in one embodiment of this application, a spatiotemporal synchronization and state compensation mechanism with millisecond-level accuracy can be constructed.

[0026] Specifically, timestamp interpolation is used for compensation in the time dimension. The system uses GPS / BeiDou dual-mode time synchronization as a global unified time source, controlling the synchronization accuracy of all sensing units to within 10ms. To address data frame misalignment caused by network transmission latency and inconsistent sampling rates, a linear timestamp interpolation algorithm is used for state compensation. Let the target alignment time be... The times of the two adjacent data frames are respectively and The corresponding state value (position or velocity) is and Then at time compensation state The calculation formula is: .

[0027] Using this formula, the system can accurately reconstruct the target state of each sensor at a unified moment, eliminating time-slip error.

[0028] Spatial dimension utilizes coordinate system mapping and transformation. To achieve multi-view spatial registration, the system establishes a unified geographic coordinate transformation model. The local Cartesian coordinate system of the roadside sensing devices is accurately converted to a unified geographic coordinate system, achieving precise physical overlap between radar point clouds and video pixels. The coordinate transformation employs a rotation and translation matrix model: ; in, For sensor local coordinates, For the converted unified geographic coordinates, The latitude and longitude coordinates of the equipment installation point. Based on device orientation angle The rotation matrix.

[0029] The system first extracts the visual features (target classification, attributes) from the video sensor and the motion features (distance, speed) from the millimeter-wave radar. It then uses the radar's strong penetration in adverse weather conditions (rain, fog, snow) to compensate for the shortcomings of visual perception, thus achieving robust detection in all weather conditions.

[0030] To address the challenge of matching data with different dimensions (video pixels and radar distance), a feasible implementation can employ Mahalanobis distance as the association cost function. The video target state vector is then calculated. With radar target state vector Distance between : ; in, Let be the state error covariance matrix. When... If the value is less than a preset threshold, the video and the radar detect the same target, and the association is successful.

[0031] The Hungarian Algorithm can also be used to replace the simple Mahalanobis distance threshold determination. A global cost matrix is ​​constructed, and bipartite graph matching is used to find the globally optimal sensor target pairing, resulting in higher association accuracy in multi-target dense scenes.

[0032] For successfully associated targets, a Kalman filter framework is used for deep attribute fusion to output the optimal state estimate. Prediction steps: ; ; Update steps: ; ; in, Here is the state transition matrix. For the observation matrix, and These are the covariances of process noise and observation noise, respectively. This is the Kalman gain. Through this filtering process, the system can smooth sensor noise and output a high-precision target trajectory.

[0033] Unscented Kalman filtering (UKF) can also be used instead of standard Kalman filtering. For the nonlinear motion characteristics of vehicles in curves or complex road conditions, UKF is used for state estimation and can track nonlinear trajectories more accurately than linear KF.

[0034] By establishing a unified spatiotemporal benchmark, the heterogeneity of roadside video, millimeter-wave radar, and vehicle-mounted perception data in terms of sampling frequency and spatial location is eliminated. Combined with state estimation strategies, high-precision blind spot target detection can be achieved in all weather conditions.

[0035] In one feasible embodiment, before the target detection box is generated, the radar point cloud features and the camera image features can be mapped to the same space and stitched together, and then input into a unified deep neural network for detection, which may improve the detection effect of small targets.

[0036] Step 2: Structured Construction and Dynamic Maintenance of the Blind Zone Global Situation Map Full-element structured scene modeling (S={G, O, E}).

[0037] To address the issue of fragmented information in traditional early warning systems, this embodiment also establishes a standardized blind zone scenario data model. .in, The static linear characteristics of the road segment where the blind spot is located are described, including road curvature, slope, number of lanes, and lane width. These geometric parameters provide a spatial reference for determining whether a vehicle is in a danger zone.

[0038] Define the set of real-time states of all traffic participants within the blind spot. Each goal The attribute vector includes: target unique identifier (ID), target type (Type, such as large vehicle / pedestrian), position coordinates (Pos), velocity vector (Vel), size information (Size), detection confidence (Conf), and timestamp (t).

[0039] Describe the current meteorological conditions (sunlight, visibility) and road surface physical conditions (dry / wet, icy, snow) to provide environmental context parameters for subsequent risk assessment models.

[0040] In one feasible implementation, given the highly dynamic nature of traffic flow in blind spots, an incremental update strategy can be adopted to maintain the real-time nature of the situation map.

[0041] When the sensor detects a new target in consecutive frames, the system assigns it a globally unique ID and adds it to the dynamic target set based on the initial detection confidence level. .

[0042] For continuously tracked targets, Kalman filtering can be used to smoothly update position and velocity attributes, eliminate observation noise jitter, and ensure the continuity and smoothness of the target trajectory in the map.

[0043] To prevent "ghost targets" from consuming system resources for extended periods, a dual-target removal logic can be designed. The target will be removed from the map when any of the following conditions are met: Timeout determination: The duration during which the target is not detected continuously. Exceeding the preset timeout threshold ; Boundary crossing determination: Target's position coordinates Exceeding the effective monitoring range of the blind zone defined by the system. The deletion determination formula is as follows: .

[0044] Based on the structured data described above, the system generates a bird's-eye view centered on the vehicle on the in-vehicle terminal. Targets with different risk levels are rendered using different colored legends, intuitively presenting the dynamics of vehicles that cannot be directly observed in blind spots.

[0045] Furthermore, occupancy grid map modeling can be used instead of object list-based structured modeling. Blind spots are divided into grids, with probability values ​​representing the likelihood of each grid being occupied, used to describe irregular obstacles or non-standard traffic participants.

[0046] Furthermore, based on graph database technology, knowledge graphs containing entity relationships (such as "vehicle A - following - vehicle B", "pedestrian - located - sidewalk") can be constructed to facilitate more advanced logical reasoning.

[0047] Long Short-Term Memory (LSTM) trajectory prediction can also replace Kalman filter smoothing. By using recurrent neural networks such as LSTM or GRU to process historical trajectory sequences, the future position of the target can be predicted, capturing complex behavioral patterns of human drivers.

[0048] The adaptive fusion module, based on pyramid pooling and weight prediction networks, can dynamically adjust the fusion weights of the three streams according to the input content. This adaptive mechanism enables the system to automatically select the optimal fusion strategy based on scene characteristics (such as glare severity, motion intensity, etc.), significantly improving robustness and generalization ability.

[0049] Step 3: Construct a dynamic risk assessment and adaptive threshold evolution model based on multidimensional spatiotemporal constraints. To address both emergency avoidance (imminent collision) and safe following (potential risk) driving scenarios, one embodiment of this application establishes a comprehensive risk assessment model based on reciprocal weighting. This model includes: Time to Collision (TTC): It reflects the urgency of a collision occurring under the current state of motion. The calculation formula is: ; in The distance is relative. The relative approximation speed.

[0050] When two vehicles are traveling towards each other: ; in, For the speed of this vehicle, For the target speed, The angle between the directions of motion of the two vehicles.

[0051] Collision Margin Time (THW): This reflects the driver's reaction time and operational margin at the current vehicle speed. The calculation formula is: ; in This is the current speed of the vehicle.

[0052] Overall Risk Index : ; By introducing weighting coefficients and The system can focus on the TTC indicator in close-range high-risk scenarios and on the THW indicator in long-distance following scenarios, thus constructing a risk quantification standard that covers all scenarios.

[0053] In response to the physical characteristics of icy and slippery road surfaces that significantly extend vehicle braking distance, one embodiment of this application proposes a dynamic evolution algorithm for warning thresholds, which solves the safety hazard of delayed warnings under severe weather conditions caused by traditional fixed thresholds.

[0054] The system can receive the current road surface adhesion coefficient uploaded by the road surface condition sensor in real time. And based on the dry pavement reference coefficient Risk classification threshold Perform dynamic correction: ; When the road surface is icy, it causes When the threshold is lowered, the correction coefficient increases, thus lowering the risk index threshold. A deviation occurs. In this way, the system can trigger a warning at a greater distance or at an earlier time, proactively allowing the driver additional physical braking distance due to slippery road conditions, significantly improving the system's active safety protection capabilities in frozen areas and severe weather conditions.

[0055] By constructing a comprehensive risk assessment system that integrates collision time and vehicle headway, a dynamic evolution mechanism for risk thresholds based on road surface adhesion coefficient is proposed, enabling precise quantification of driving risks under different weather and road conditions.

[0056] Step 4: Multimodal hierarchical early warning and intelligent suppression strategy based on cognitive load management To ensure safety notification while minimizing driver disruption, the system uses a comprehensive risk index. The early warning response is divided into three levels, and different channels such as vision, hearing and touch are used to provide coordinated prompts.

[0057] ; Among them, the prompt level ( (For vehicles with low risk, visual interaction is used only.) Blind spot situation maps are displayed on the in-vehicle screen using green or blue icons, providing basic information support without interrupting driving.

[0058] Warning level ( The risk is moderate, and a dual "visual + auditory" interaction is used. The screen icon turns yellow and flashes, while triggering a concise voice announcement (such as "Traffic approaching on the curve ahead") to guide the driver's attention.

[0059] Emergency level ( The risk is extremely high, so a multimodal enhanced interaction of "visual-auditory-touch" is adopted. The screen flashes red at a high frequency, accompanied by a rapid warning sound, and triggers the steering wheel or seat to vibrate in conjunction with the vehicle's CAN bus, ensuring that the driver can perceive the danger and take braking measures within milliseconds.

[0060] To address the information overload problem caused by repeated alarms triggered on the same target within a short period, this invention establishes a dynamic suppression model based on the number of alarms. Suppression factor calculation model: ; in, This represents the cumulative number of warnings issued for the current target. The warning intensity attenuation coefficient, The function sets a lower limit for the inhibition factor.

[0061] With the number of alarms The increase of inhibitory factors The alarm frequency or volume decreases exponentially, automatically reducing the frequency or volume of subsequent alarms. This mechanism avoids driver frustration caused by the "boy who cried wolf" effect, and by setting a lower limit (0.3), it ensures that the system retains a minimum level of safety alert capability even after multiple suppressions, achieving a balance between user experience and safety assurance.

[0062] The vehicle-road-cloud collaborative system provided in this application uses the C-V2X protocol in its vehicle-road communication subsystem to push the aforementioned situational and early warning information to vehicles in milliseconds. Upon receiving the information, the onboard unit (OBU) performs differentiated interactions based on the warning level. At the prompt level, only a green situational map is displayed on the screen; at the warning level, a yellow flashing indicator and voice broadcast are added; and at the emergency level, a red high-frequency flashing indicator, alarm sound, and haptic feedback on the steering wheel are triggered. This system exhibits high detection accuracy in adverse weather conditions, increases the advance warning time on icy and snowy roads, and significantly improves the safety and reliability of driving in blind spots. It can be directly applied to practical road traffic information collection systems, providing traffic management departments with an effective participant selection strategy. Through reasonable participant selection, the quality and scope of data collection can be maximized within a limited budget, providing strong support for the construction of intelligent transportation systems.

[0063] By employing a tiered interaction model tailored to drivers' cognitive characteristics, differentiated multimodal feedback enables the accurate transmission of risk information. Furthermore, an exponentially decaying warning suppression mechanism is introduced to effectively address driver desensitization and fatigue issues caused by frequent, repetitive alarms.

[0064] Example 2 Figure 1 A flowchart illustrating a vehicle blind spot warning method based on multi-source information fusion and risk classification, provided as an embodiment of this application. Figure 1As shown in the figure, an embodiment of this application provides a vehicle blind spot warning method based on multi-source information fusion and risk classification, including: S1, Data Model Based on Blind Spot Scene By analyzing the status data of the target to be evaluated and other targets within the blind zone of the target to be evaluated, the collision risk index between the target to be evaluated and each other target at the current moment is determined.

[0065] The target here can be vehicles or pedestrians, etc.

[0066] Here, the collision risk index between the target to be evaluated and other targets can be determined in the following way. : ; in, , These are the weighting coefficients. The baseline time for collision between the target to be evaluated and other targets. The collision margin time between the target to be evaluated and other targets. This is the error coefficient.

[0067] The state data includes at least the position coordinates in the physical coordinate system, the movement speed, and the collision reference time between the target and other targets based on the angle between their movement directions. Based on the position coordinates of the target to be evaluated and other targets, the relative distance between the target to be evaluated and other targets is calculated; based on the moving speed and the angle between the moving direction of the target to be evaluated and other targets, the relative approach speed between the target to be evaluated and other targets is calculated; the ratio between the relative distance and the relative approach speed is calculated as the collision reference time.

[0068] The step of calculating the relative approach speed between the target and other targets based on their moving speeds and the angle between their directions of motion specifically includes: Calculate the product between the moving speeds of other targets and the angle between the moving directions of the target being evaluated and the other targets. Calculate the sum of the moving speeds of the target being evaluated and the product, as the relative approach speed.

[0069] In one feasible embodiment, the target's first moving speed at a first acquisition time is collected based on GPS, and the target's second moving speed at a second acquisition time is collected based on BeiDou satellite data. Both the first and second acquisition times are adjacent to the current time. The target's current movement speed can be determined in the following ways: Calculate the speed difference between the first and second moving speeds; calculate the first time difference between the current time and the first acquisition time; calculate the second time difference between the second acquisition time and the first acquisition time; calculate the time difference ratio between the first and second time differences; calculate the product of the time difference ratio and the speed difference; calculate the sum of the product and the first moving speed as the current speed.

[0070] In another feasible embodiment, the position coordinates of the target to be evaluated in the physical coordinate system can be determined in the following manner. :

[0071] in, This refers to the position coordinates of the target in the sensing coordinate system, based on data collected by the positioning sensor. To locate the latitude and longitude coordinates of the sensor, Based on device orientation angle The rotation matrix.

[0072] S2. Based on the environmental data in the blind spot scenario data model, determine the dynamic risk threshold corresponding to the current moment.

[0073] In step S2, the environmental data includes at least the real-time road surface adhesion coefficient, and the dynamic risk threshold corresponding to the current moment can be determined in the following way: Calculate the ratio between the dry pavement adhesion coefficient and the real-time pavement adhesion coefficient. Based on the product of this ratio and the baseline risk threshold, determine the dynamic risk threshold for the current moment.

[0074] Here, you can use the comprehensive risk index. The early warning response is divided into three levels, and different channels such as vision, hearing and touch are used to provide coordinated prompts.

[0075] .

[0076] S3. Based on the relationship between the current collision risk index and the dynamic risk threshold, determine the risk level of the target to be assessed and implement the corresponding early warning measures.

[0077] In step S3, the prompt level can employ visual interaction. A blind spot situation map is displayed on the vehicle's screen using green or blue icons, providing basic information support without interrupting driving concentration.

[0078] Warning levels can employ a dual "visual + auditory" interaction. The screen icon turns yellow and flashes, while simultaneously triggering a concise voice announcement (such as "Oncoming traffic around the bend") to guide the driver's attention.

[0079] In emergency situations, a multimodal enhanced interaction combining visual, auditory, and tactile feedback can be employed. The screen flashes red at a high frequency, accompanied by a rapid warning sound, and the vehicle's CAN bus is activated to trigger vibrations in the steering wheel or seat, ensuring that the driver perceives the danger and takes braking action within milliseconds.

[0080] Example 3 Figure 2 This is a schematic diagram of a vehicle blind spot warning device based on multi-source information fusion and risk classification, provided as an embodiment of this application. Figure 2 As shown, based on the same inventive concept, this application also provides a vehicle blind spot warning device 20 based on multi-source information fusion and risk classification. The device includes: The first processing module 210 is used to determine the collision risk index between the target to be evaluated and each other target at the current moment based on the state data of the target to be evaluated and other targets located within the blind zone of the target to be evaluated in the blind zone scene data model. The second processing module 220 determines the dynamic risk threshold corresponding to the current moment based on the environmental data in the blind spot scene data model. The assessment module 230 is used to determine the risk level of the target to be assessed based on the relationship between the current collision risk index and the dynamic risk threshold, and to execute corresponding early warning measures.

[0081] In a preferred embodiment, the first processing module 210 determines the collision risk index between the target to be evaluated and other targets in the following manner. : ; in, , These are the weighting coefficients. The baseline time for collision between the target to be evaluated and other targets. The collision margin time between the target to be evaluated and other targets. This is the error coefficient.

[0082] In a preferred embodiment, the state data includes at least the position coordinates in the physical coordinate system, the movement speed, and the angle between the target to be evaluated and other targets based on their movement directions. The first processing module 210 determines the collision reference time between the target to be evaluated and other targets in the following manner: Based on the position coordinates of the target to be evaluated and other targets, the relative distance between the target to be evaluated and other targets is calculated; Based on the moving speed and the angle between the target to be evaluated and other targets, the relative approach speed between the target to be evaluated and other targets is calculated; Calculate the ratio between relative distance and relative approach speed, and use it as the collision reference time.

[0083] In a preferred embodiment, the first processing module 210 calculates the relative approach speed between the target to be evaluated and other targets based on the moving speed and the angle between their directions of movement, specifically including: Calculate the product between the moving speed of other targets and the angle between the moving directions of the target to be evaluated and other targets; Calculate the sum of the moving speed of the target to be evaluated and its product, as the relative approach speed.

[0084] In a preferred embodiment, the environmental data includes at least the real-time road surface adhesion coefficient, and the second processing module 220 determines the dynamic risk threshold corresponding to the current moment in the following manner: Calculate the ratio between the dry road surface adhesion coefficient and the real-time road surface adhesion coefficient; The dynamic risk threshold at the current moment is determined by multiplying the ratio by the baseline risk threshold.

[0085] In a preferred embodiment, the target's first moving speed at a first acquisition time is collected based on GPS, and the target's second moving speed at a second acquisition time is collected based on BeiDou satellite data. Both the first and second acquisition times are adjacent to the current time. The first processing module 210 determines the target's current moving speed in the following way: Calculate the speed difference between the first and second movement speeds; Calculate the first time difference between the current time and the first acquisition time; Calculate the second time difference between the second acquisition time and the first acquisition time; Calculate the time difference ratio between the first time difference and the second time difference; Calculate the product of the time difference ratio and the difference in movement speed; Calculate the sum of the product and the first moving speed, and use it as the moving speed at the current moment.

[0086] In a preferred embodiment, the first processing module 210 determines the position coordinates of the target to be evaluated in the physical coordinate system in the following manner. :

[0087] in, This refers to the position coordinates of the target in the sensing coordinate system, based on data collected by the positioning sensor. To locate the latitude and longitude coordinates of the sensor, Based on device orientation angle The rotation matrix.

[0088] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 3 As shown, the electronic device 300 includes a processor 310, a memory 320, and a bus 330.

[0089] The memory 320 stores machine-readable instructions that can be executed by the processor 310. When the electronic device 300 is running, the processor 310 and the memory 320 communicate via the bus 330. When the machine-readable instructions are executed by the processor 310, the steps of a vehicle blind spot warning method based on multi-source information fusion and risk classification as described in the above method embodiment can be executed. For specific implementation details, please refer to the method embodiment, which will not be repeated here.

[0090] This application also provides a computer-readable storage medium storing a computer program. When the computer program is run by a processor, it can execute the steps of a vehicle blind spot warning method based on multi-source information fusion and risk classification as described in the above method embodiments. For specific implementation details, please refer to the method embodiments, which will not be repeated here.

[0091] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0092] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0093] Furthermore, the units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0094] Furthermore, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0095] It should be noted that if the function is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0096] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0097] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A vehicle blind spot warning method based on multi-source information fusion and risk classification, characterized in that, The method includes: Based on the state data of the target to be evaluated and other targets within the blind zone of the target to be evaluated in the blind zone scene data model, the collision risk index between the target to be evaluated and each other target at the current moment is determined. Based on the environmental data in the blind spot scenario data model, the dynamic risk threshold corresponding to the current moment is determined. Based on the relationship between the current collision risk index and the dynamic risk threshold, the risk level of the target to be assessed is determined, and corresponding early warning measures are implemented.

2. The method according to claim 1, characterized in that, The collision risk index between the target to be evaluated and other targets is determined using the following method. : ； in, , These are the weighting coefficients. The baseline time for collision between the target to be evaluated and other targets. The collision margin time between the target to be evaluated and other targets. This is the error coefficient.

3. The method according to claim 2, characterized in that, The state data includes at least the position coordinates in the physical coordinate system, the movement speed, and the collision reference time between the target and other targets based on the angle between their movement directions. Based on the position coordinates of the target to be evaluated and other targets, the relative distance between the target to be evaluated and other targets is calculated; Based on the moving speed and the angle between the target to be evaluated and other targets, the relative approach speed between the target to be evaluated and other targets is calculated; Calculate the ratio between the relative distance and the relative approach speed, and use it as the collision reference time.

4. The method according to claim 3, characterized in that, The steps for calculating the relative approach speed between the target and other targets, based on the moving speed and the angle between their directions of motion, specifically include: Calculate the product between the moving speed of other targets and the angle between the moving directions of the target to be evaluated and other targets; The sum of the moving speed of the target to be evaluated and the product is calculated as the relative approach speed.

5. The method according to claim 1, characterized in that, Environmental data, including at least the real-time road surface adhesion coefficient, is used to determine the dynamic risk threshold at the current moment using the following methods: Calculate the ratio between the dry road surface adhesion coefficient and the real-time road surface adhesion coefficient; The dynamic risk threshold corresponding to the current moment is determined based on the product of the ratio and the benchmark risk threshold.

6. The method according to claim 1, characterized in that, The target's first moving speed at the first acquisition time is based on GPS data, and the target's second moving speed at the second acquisition time is based on BeiDou satellite data. Both the first and second acquisition times are adjacent to the current time. The target's current speed is determined as follows: Calculate the speed difference between the first and second movement speeds; Calculate the first time difference between the current time and the first acquisition time; Calculate the second time difference between the second acquisition time and the first acquisition time; Calculate the time difference ratio between the first time difference and the second time difference; Calculate the product of the time difference ratio and the speed difference. Calculate the sum of the product and the first moving speed, and use it as the moving speed at the current moment.

7. The method according to claim 1, characterized in that, The position coordinates of the target in the physical coordinate system are determined using the following method. : in, This refers to the position coordinates of the target in the sensing coordinate system, based on data collected by the positioning sensor. To locate the latitude and longitude coordinates of the sensor, Based on device orientation angle The rotation matrix.

8. A vehicle blind spot warning device based on multi-source information fusion and risk classification, characterized in that, The device includes: The first processing module is used to determine the collision risk index between the target to be evaluated and each other target at the current moment based on the state data of the target to be evaluated and other targets within the blind zone of the target to be evaluated in the blind zone scene data model. The second processing module determines the dynamic risk threshold corresponding to the current moment based on the environmental data in the blind spot scene data model. The assessment module is used to determine the risk level of the target to be assessed based on the relationship between the current collision risk index and the dynamic risk threshold, and to execute corresponding early warning measures.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus, and the processor executes the machine-readable instructions to perform the steps of the vehicle blind spot warning method based on multi-source information fusion and risk classification as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the vehicle blind spot warning method based on multi-source information fusion and risk classification as described in any one of claims 1 to 7.

Images