A method, system, device and storage medium for blind spot warning of passenger vehicles

Abstract

This invention belongs to the field of blind spot warning for buses, specifically disclosing a method, system, device, and storage medium for blind spot warning of buses. By comparing the field of vision coverage of the vehicle under straight-ahead and turning conditions, it identifies the blind spot loss area caused by changes in vehicle posture, accurately identifying dynamic blind spots and avoiding the blind spot omission problem caused by static modeling, ensuring the comprehensiveness and accuracy of blind spot identification. At the same time, when a blind spot loss area is identified, it focuses on the nearby visible area around the blind spot loss area to perform real-time identification and motion trajectory capture of dynamic targets. Then, based on the potential conflict analysis between the captured target motion trajectory and the vehicle driving trajectory, it promptly generates warning information, improving the early detection capability of potential conflicts, giving drivers more time to take evasive measures, effectively preventing traffic accidents caused by blind spots, and ensuring the safety of pedestrians, non-motorized vehicles, and other vehicles.

Description

Technical Field

[0001] This invention belongs to the field of blind spot warning for buses, and specifically discloses a method, system, device and storage medium for blind spot warning for buses. Background Technology

[0002] With the advancement of urbanization and the continuous development of public transportation, buses, as an important carrier of urban travel, undertake a large number of passenger transport tasks. However, due to their large body structure and limited driver visibility, especially when turning, there are significant blind spots around the vehicle, which can easily lead to a lack of perception of pedestrians, non-motorized vehicles, and other road users, thus causing safety hazards. Therefore, monitoring and early warning of blind spots for buses is essential.

[0003] Existing technologies also include blind spot warning solutions for buses. For example, Chinese invention patent CN113744532A proposes a method and device for blind spot warning of urban traffic buses based on vehicle-road cooperation. This method acquires the vehicle's location information and inherent visual blind spot parameters to construct the vehicle's coordinate system and establishes a region representation function for each sub-blind spot based on the geometric characteristics of the blind spot. Furthermore, it combines onboard and roadside sensing devices to acquire the location information of surrounding traffic participants and transforms it into the vehicle's coordinate system. Finally, based on the vehicle's current driving behavior, the target location is substituted into the corresponding blind spot function model to achieve blind spot risk identification and warning.

[0004] The above-mentioned scheme effectively utilizes vehicle-road cooperative technology to improve the perception of targets in blind spots in complex urban traffic environments. However, the scheme still has certain limitations in practical applications: First, its blind spot modeling is mainly based on static geometric parameters, and does not fully consider the dynamic migration and deformation of the blind spot range caused by changes in vehicle body posture when the vehicle is turning. This static modeling method is prone to deviation between the detection of the blind spot area and the actual visible range when the vehicle is not traveling in a straight line, thus affecting the accuracy of risk identification. Second, the scheme lacks effective prediction of the dynamic target movement trend in terms of target behavior analysis, and fails to establish a dynamic assessment of the potential conflict between the blind spot target entry path and the vehicle's driving trajectory. Therefore, it is difficult to identify potential threats in a timely manner when facing complex traffic behaviors such as sudden approach and crossing, resulting in a lag in early warning generation and affecting the timeliness and reliability of the early warning. Summary of the Invention

[0005] Therefore, one objective of this application is to provide a method, system, device, and storage medium for warning of blind spots in buses. By focusing on the blind spots generated by the bus when turning, the method can identify the movement trend and predict the trajectory of dynamic targets within the blind spots in real time, thereby achieving accurate warning of potential conflict risks and effectively solving the problems existing in the prior art.

[0006] The objective of this invention can be achieved through the following technical solution: A blind spot warning method for a bus, comprising the following steps: S1. Generating a first panoramic field of view coverage area under straight-going conditions and generating a second panoramic field of view coverage area under turning conditions by using a wide-angle camera group deployed around the vehicle body.

[0007] S2. Spatially overlay the first panoramic field of view coverage area and the second panoramic field of view coverage area, and mark the field of view missing area of ​​the second panoramic field of view coverage area relative to the first panoramic field of view coverage area.

[0008] S3. Based on the spatial location and contour boundary information of the missing area, perform an adjacent region search operation within the second panoramic field of view coverage to determine the visible area adjacent to the missing area.

[0009] S4. Perform dynamic target detection on the visible area adjacent to the missing area of ​​vision, and record the last spatial coordinates and motion vector of the target when it disappears.

[0010] S5. Starting from the last spatial coordinates, generate the predicted trajectory line according to the motion vector, and calculate the conflict time window between the predicted trajectory and the vehicle trajectory by combining the real-time motion trajectory of the bus.

[0011] S6. Based on the conflict time window, blind spot zoning and warning are issued for areas with missing vision.

[0012] Combining all the above technical solutions, the positive effects of this invention are as follows: 1. By comparing the field of vision coverage of a vehicle under straight-line and turning conditions, this invention identifies the field of vision loss area caused by changes in vehicle posture, accurately identifies dynamic blind spots, avoids the problem of blind spot omission caused by static modeling, and ensures the comprehensiveness and accuracy of blind spot identification.

[0013] 2. When a blind spot is detected, this invention focuses on the nearby visible area around the blind spot to perform real-time identification and motion trajectory capture of dynamic targets. Then, based on the potential conflict analysis between the captured target motion trajectory and the vehicle driving trajectory, it generates early warning information in a timely manner, which improves the ability to detect potential conflicts early, gives drivers more time to take evasive measures, effectively prevents traffic accidents caused by blind spots, and protects the safety of pedestrians, non-motorized vehicles and other vehicles. Attached Figure Description

[0014] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0015] Figure 1 This is a diagram illustrating the implementation steps of a blind spot warning method for buses according to the present invention.

[0016] Figure 2 This is a flowchart illustrating the operation of the time window for analyzing the conflict between the predicted trajectory of the target and the real-time trajectory of the bus in this invention.

[0017] Figure 3 This is a schematic diagram illustrating the blind spot zoning and early warning operation for areas with missing vision in this invention.

[0018] Figure 4 This is a module connection diagram of a bus blind spot warning system according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1

[0021] See Figure 1 As shown, the present invention proposes a blind spot warning method for buses, including the following steps: S1. A first panoramic field of view coverage area is generated in the straight-going condition by a wide-angle camera group deployed around the vehicle body, and a second panoramic field of view coverage area is generated in the turning condition.

[0022] As one possible implementation of the above steps, the specific generation process of the first panoramic field of view coverage area is as follows: when the vehicle is in a straight-line driving condition, the wide-angle camera group deployed around the vehicle body is activated, and each camera simultaneously acquires images.

[0023] The aforementioned method of capturing images around the vehicle utilizes wide-angle cameras installed around the vehicle to capture 360-degree environmental information. Each camera is responsible for monitoring a specific direction, ensuring there are no blind spots.

[0024] Construct a rectangular projection coordinate system parallel to the ground with the longitudinal axis of the vehicle body as the axis of symmetry.

[0025] In the preferred embodiment, the boundary of the rectangular plane is determined by the length and width of the vehicle body, and it is symmetrically extended with the longitudinal axis of the vehicle body as the axis of symmetry to form a rectangular area that is centrally symmetrical about the longitudinal axis of the vehicle body. Its range covers the visible area within a certain distance around the vehicle. Specifically, the rectangular projection coordinate system is established as follows: the geometric center of the vehicle is used as the origin of the coordinate system. This point is usually the reference point of the vehicle coordinate system to ensure consistency with the vehicle kinematic model.

[0026] Set the X-axis direction: Extend forward along the longitudinal axis of the vehicle body to indicate the direction of vehicle movement.

[0027] Y-axis: Perpendicular to the vehicle's longitudinal axis, extending horizontally to the right, representing the vehicle's lateral direction. Z-axis: Perpendicular to the ground, extending upwards, representing the vertical direction.

[0028] It's important to understand that the rectangular projection coordinate system constructed using the vehicle's longitudinal axis is necessary because this axis represents the primary direction of vehicle movement and serves as the driver's reference for environmental perception, steering, and obstacle avoidance. Establishing the projection coordinate system around this axis ensures that the constructed panoramic image aligns with the vehicle's direction of motion, providing a unified geometric reference framework for subsequent field-of-view coverage analysis, target trajectory prediction, and conflict assessment. Furthermore, if the projection coordinate system is not aligned with the vehicle's structure during multi-camera image stitching, geometric distortion, viewpoint misalignment, or ghosting can easily occur, affecting the integrity and accuracy of the panoramic view. Establishing a rectangular projection coordinate system using the vehicle's longitudinal axis as the axis of symmetry ensures that images from each camera are spatially aligned and transformed within a unified coordinate system, thereby improving image stitching accuracy and visual consistency and reducing stitching errors.

[0029] Distortion correction and coordinate transformation are performed on the original wide-angle images output by each camera, and each image is projected from its original viewpoint onto the aforementioned rectangular projection coordinate system.

[0030] The distortion correction mentioned above is necessary because wide-angle lenses have optical distortion problems such as barrel distortion. The original images acquired by each camera need to be distorted first.

[0031] The corrected images are stitched together to form a continuous rectangular field of view.

[0032] The above-mentioned pixel-level fusion of the corrected image through image stitching eliminates visual differences and geometric misalignments in overlapping areas of the images, generating a continuous and seamless rectangular panoramic image, which is used to characterize the range of the surrounding environment that the driver can effectively perceive through the visual system when the vehicle is traveling straight.

[0033] As another possible implementation of the above steps, when the turn signal is detected to be activated or the steering wheel angle exceeds a preset threshold, the wide-angle camera group located around the vehicle body is simultaneously activated to start real-time image acquisition.

[0034] In the above operation, triggering image acquisition when the vehicle is about to turn or is in the process of turning helps to promptly acquire newly emerging blind spot areas due to changes in vehicle posture, thereby improving the real-time performance and dynamic adaptability of blind spot recognition. A preset threshold for the steering wheel angle is used to identify whether the vehicle has entered a turning condition with a significant risk of blind spot changes. This threshold reflects the critical point at which the vehicle's steering behavior significantly affects the blind spot range. Specifically, it can be achieved by collecting data on common steering behaviors of vehicles at intersections and on mountain roads, statistically analyzing the distribution of common steering angles, and identifying the minimum steering angle threshold that effectively characterizes turning behavior without affecting normal driving operations.

[0035] The projected coordinate system, with the vehicle's longitudinal axis as the axis of symmetry, is dynamically adjusted based on the steering angle of the vehicle's front wheels.

[0036] Applying this to the above operations, when the vehicle is traveling straight, the projection coordinate system is symmetrically distributed around the vehicle's longitudinal axis, forming a centrally symmetrical rectangular projection area at the front, rear, left, and right sides of the vehicle. However, when the vehicle is turning, due to the change in the vehicle's posture, the originally symmetrical blind spot area will shift, with the blind spot mainly concentrated on the inside of the turn.

[0037] To more accurately map the visual coverage area at this moment, the front wheel steering angle is obtained during vehicle turning. This angle reflects the current steering magnitude of the vehicle, and the projection coordinate system is dynamically offset and modeled based on the front wheel steering angle. Specifically, this involves shifting the center point of the projection area a certain distance inwards from the turning direction based on the steering direction and angle magnitude. This makes the projection area closer to the vehicle's actual visible range, ensuring that the X-axis direction of the projection coordinate system is consistent with the vehicle's current driving direction, thereby guaranteeing the geometric consistency of the image projection.

[0038] The above operation takes into account the vehicle's attitude changes during turning, ensuring that the coordinate system can accurately reflect the vehicle's actual position and orientation relative to the ground.

[0039] The original images output by each wide-angle camera are subjected to distortion correction and coordinate transformation, and then mapped to a projection coordinate system that is dynamically adjusted according to the steering angle of the vehicle's front wheels.

[0040] The corrected images are stitched together to form a continuous curved edge field of view.

[0041] S2. Spatially overlay the first panoramic field of view coverage area and the second panoramic field of view coverage area, and mark the field of view missing area of ​​the second panoramic field of view coverage area relative to the first panoramic field of view coverage area.

[0042] The specific implementation process of the above steps is as follows: extract the closed envelope boundary of the outer contour of the first panoramic field of view coverage area and the second panoramic field of view coverage area respectively.

[0043] It should be noted that the closed envelope boundaries of the two aforementioned field-of-view coverage areas serve as the spatial geometric boundaries of their respective visible regions.

[0044] Using the closed envelope boundary of the first panoramic field of view coverage area as a reference, the closed envelope boundary of the second panoramic field of view coverage area is geometrically superimposed on it, and the first panoramic field of view coverage area that is not covered by the second panoramic field of view coverage area is extracted as the field of view missing area.

[0045] It should be added that before geometrically superimposing the closed envelope boundaries of the two covered domains, the two closed envelope boundaries need to be uniformly mapped to the same reference coordinate system and the coordinates should be normalized or aligned to eliminate geometric errors caused by coordinate offset or rotation.

[0046] It is important to understand that the aforementioned areas of visual loss represent the range of visibility lost when a vehicle transitions from a straight-ahead to a turning state.

[0047] S3. Based on the spatial location and contour boundary information of the missing area, perform an adjacent region search operation within the second panoramic field of view coverage to determine the visible area adjacent to the missing area.

[0048] As an optional implementation of the above steps, the outline boundary of the missing field of view is marked within the second panoramic field of view coverage area based on the spatial location and outline boundary information of the missing field of view area.

[0049] Based on the contour boundaries of the region with missing visual field, boundary adjacency analysis is performed to identify image regions that share a common boundary with the region with missing visual field. The boundary features of these image regions are then traced until continuous image regions without image stitching are identified. These continuous regions that are not affected by stitching and share a common boundary with the region with missing visual field are defined as the adjacent visible range.

[0050] It is important to understand that image regions sharing a common boundary with the blind spot area are considered adjacent visible regions because they are spatially directly adjacent to the blind spot area and belong to the spatial neighborhood of the blind spot change area. Targets within this region have a high probability of being visible before the blind spot forms or re-entering the field of view after the blind spot disappears. Therefore, this is a key area for dynamic target recognition and trajectory prediction. Targets usually appear in the adjacent visible region before entering or leaving the blind spot. Through continuous monitoring of this region, the system can capture the target's entry trajectory, movement direction, and speed information, providing a data foundation for subsequent trajectory prediction and conflict warning for blind spot targets. Furthermore, focusing on monitoring only the area adjacent to the blind spot, rather than performing indiscriminate analysis on the entire panoramic view, helps reduce unnecessary image processing computation and improves system response efficiency.

[0051] In particular, during image stitching, factors such as differences in field of view between cameras, projection transformation errors, or uneven lighting may introduce visual breaks, ghosting, or mismatches into the stitching area. Therefore, when identifying a region that shares a common boundary with the region with a missing field of view, the adjacent visible range is selected as the region without stitching around the region with a missing field of view to ensure that the extracted adjacent visible range has spatial geometric consistency and image integrity.

[0052] The adjacent visible areas are divided according to their respective image acquisition sources to generate visible sub-regions.

[0053] The image acquisition sources mentioned above correspond to wide-angle cameras.

[0054] S4. Perform dynamic target detection on the visible area adjacent to the missing area of ​​vision, and record the last spatial coordinates and motion vector of the target when it disappears.

[0055] As an innovative implementation of the above scheme, moving objects entering the area are identified in real time within the adjacent visible sub-regions surrounding the area with missing field of view. Once a dynamic target is detected, its trajectory is continuously tracked.

[0056] If a dynamic target disappears from its visible sub-region in the current image frame during tracking, a target matching search operation is performed on the spatially adjacent region of the visible sub-region within a set time window to determine whether the target has migrated to an adjacent field of view. If the target is not detected again in the adjacent region, the time when the target disappeared is recorded, and the adjacent visible sub-region where it disappeared is located.

[0057] Starting from the moment of disappearance, a set number of historical image frames are traced backward to extract the spatial location information of the target in consecutive frames.

[0058] For the target's position coordinates in consecutive frames, calculate the displacement vector of its centroid between adjacent frames. This displacement vector reflects the target's motion trend and directional characteristics before it disappears.

[0059] Applying the above scheme, the expression for the displacement vector of the centroid between adjacent frames is: ,in and They represent the target at the 1st and 2nd positions, respectively. Frame and The centroid coordinates of the frame.

[0060] See Figure 2 As shown, S5. Starting from the last spatial coordinates, generate the predicted trajectory line according to the motion vector, and calculate the conflict time window between the predicted trajectory and the vehicle trajectory by combining the real-time motion trajectory of the bus.

[0061] As one implementation of the above steps: Based on the direction of the motion vector when the dynamic target disappears in the adjacent visible sub-region, determine whether it points to the area of ​​missing vision. If it points to the area of ​​missing vision, it means that the dynamic target may be moving into the vehicle's blind spot. At this time, linear extrapolation is used to generate a straight-line predicted trajectory along the direction of its motion vector, starting from the last visible position of the target, to simulate the potential motion path of the dynamic target after entering the blind spot.

[0062] It is important to understand that linear extrapolation refers to the process of predicting the position or trajectory of a target at a future moment by assuming that its motion trend remains unchanged, given that the target's motion state, such as position, speed, and direction, is known. This is achieved by using mathematical modeling methods to extend the target's motion path along its current motion direction.

[0063] In one embodiment, the specific implementation process of linear extrapolation is as follows:

[0064] Input conditions: Known position coordinates of the target at the last visible time. The known motion vector of the target includes the velocity magnitude. With direction angle .

[0065] Assuming the target maintains uniform linear motion after entering the blind zone, along the direction of its motion vector, Starting from a point, a linear predicted trajectory is generated, with any point on the trajectory... It can be represented as ,in It is a time variable, representing the passage of time after the target enters the blind zone.

[0066] The real-time bus trajectory is determined based on the steering wheel angle and vehicle speed.

[0067] In the specific implementation of the above scheme, in order to achieve high-precision calculation of the real-time motion trajectory of the bus, the real-time rotational speed of the four wheels of the vehicle and the steering angle of the front wheels are collected, whereby the wheel speed is used to estimate the longitudinal speed of the vehicle. ,in Indicates the first The angular velocity of each wheel Indicates the first The tire radius of each wheel Indicates the wheel number. .

[0068] The front wheel steering angle is used to calculate the vehicle's heading angle. Specifically, the heading angle is expressed as follows: in Indicates the front wheel steering angle. Indicates wheelbase.

[0069] Vehicle location update ,in Indicates the vehicle is in The coordinates of the travel position at any given moment. Indicates time, Indicates the sampling time interval. Indicates the vehicle is in The heading angle at any given moment.

[0070] The real-time bus trajectory can be obtained based on the vehicle's updated position.

[0071] It is important to know that the above analysis of the bus's trajectory uses data from the vehicle's own sensors, such as wheel speed and steering angle, to calculate the vehicle's trajectory. This is a method of dead reckoning that does not rely on an external positioning system and achieves high-precision calculation of the bus's trajectory.

[0072] The predicted trajectory of a dynamic target entering a region outside the field of view is geometrically compared with the trajectory of the bus to determine whether the two trajectories intersect. If they intersect, it means that the dynamic target and the vehicle may appear in the same spatial location at some point in the future, which means there is a potential collision risk. At this time, the shortest spatial distance between the two trajectories is calculated and compared with the set safety threshold. If the shortest spatial distance is less than the safety threshold, a conflict time window is calculated based on the shortest spatial distance and the relative speed between the target and the vehicle. When there is no intersection, it means that the dynamic target and the vehicle are unlikely to appear in the same spatial location in the future, and the probability of a collision is small. The shortest spatial distance between the two trajectories is calculated again. If the shortest spatial distance is less than the safety threshold, the spatial distance between the two trajectories is updated. If the distance is detected to be continuously shortening in multiple consecutive time steps, the conflict time window calculation is retried.

[0073] The shortest spatial distance mentioned above refers to the Euclidean distance between the two closest points on the paths of the predicted trajectory of the dynamic target and the trajectory of the bus. This distance is used to measure the spatial proximity between the target and the vehicle.

[0074] The aforementioned safety threshold sets a critical distance; when the shortest spatial distance between a dynamic target and a vehicle is less than this threshold, a high risk of collision is considered to exist. For example, statistical analysis of a large amount of traffic accident data can be used to identify the typical minimum distance between the vehicle and the target before an accident, which can then be used as a reference for the safety threshold.

[0075] The aforementioned conflict time window can be obtained by dividing the shortest spatial distance by the relative speed between the target and the vehicle. This conflict time window reflects the time interval from the current moment to the expected occurrence of a potential conflict, and is used to quantify the response time that the driver or autonomous driving system can take evasive action. The shorter the conflict time window, the faster the dynamic target approaches the vehicle or the less remaining safe time, meaning that the system faces a higher risk of time urgency and needs to make an effective response in a shorter time to avoid a possible collision.

[0076] It is important to understand that when two trajectories intersect, the shortest spatial distance usually occurs near the intersection. Using the shortest spatial distance for calculation can accurately reflect the spatial interval between the dynamic target and the vehicle in case of a potential collision. This helps the system to issue a timely warning before a potential collision occurs, providing the driver with sufficient reaction time to take emergency braking or other evasive actions, thereby effectively avoiding the occurrence of a collision.

[0077] When two tracks do not intersect, the shortest spatial distance refers to the distance between the two closest points between the two non-intersecting tracks. If this distance is greater than the set safety threshold, there is no direct risk of collision. If it is less than the safety threshold, there may still be potential danger, and it is necessary to continue monitoring the trend of the relative position changes of the two tracks. If it is found that the shortest spatial distance between the two tracks is continuously decreasing, it is still necessary to calculate the conflict time window through the shortest spatial distance to assess the degree of potential collision risk and decide whether to issue a warning. In this way, the warning strategy can be dynamically adjusted to ensure that timely and accurate risk warnings are provided in any situation, helping the driver or autonomous driving system to make the best decision.

[0078] When the target is not pointing towards a blind spot, it means that the target is not moving towards the vehicle's blind spot, but may be turning towards other visible sub-areas. At this time, target search is performed in other visible sub-areas pointed to by its motion vector. If the target is detected again in other visible sub-areas, continuous tracking of it is resumed, and its trajectory and motion vector are updated to repeatedly determine whether it is pointing towards a blind spot. If the target is not detected, area monitoring and re-identification attempts are made based on the target's last movement trend until it is confirmed that the target has left the system's field of vision or reappears.

[0079] See Figure 3 As shown, S6. Based on the conflict time window, blind zone marking and warning are performed on the areas with missing vision.

[0080] Specifically, the area of ​​missing vision is divided according to the conflict time window as follows:

[0081] High-risk sub-zone: If the conflict time window is less than the lower limit of the warning range, it is determined that there is a high risk of collision between the target and the vehicle.

[0082] Medium-risk sub-zone: The conflict time window is within the warning range;

[0083] Low-risk sub-zone: The conflict time window is greater than the upper limit of the warning range, indicating that there is a low probability of collision between the target and the vehicle.

[0084] The aforementioned warning range for the conflict time window is an interval used to define different risk levels within the conflict time window. The design of this warning range must comprehensively consider the average reaction time of the driver or autonomous driving system and the time required for the vehicle to brake at the current speed. For example, at high speeds, due to the longer braking distance, the warning range should be appropriately increased to provide sufficient warning time.

[0085] When setting the warning range based on this consideration, the lower limit of the warning range is first calculated by summing the average reaction time of the driver or autonomous driving system and the time required for the vehicle to brake at the current speed. This value reflects the shortest time required for the system to take effective avoidance measures from detecting a potential collision risk, ensuring sufficient time for response and braking in emergency situations. The upper limit can be set to be a certain percentage larger than the lower limit to provide a certain safety margin. This value provides additional time margin, enabling the system to issue warnings at an earlier time, thereby improving the timeliness and reliability of the warnings.

[0086] For high-risk sub-areas, an audible and visual directional alarm is triggered. A directional sound beam is generated through the phase array speaker built into the door. The focus of the sound beam locks onto the coordinates of the target's disappearance, realizing spatial perception-guided auditory warning. When necessary, it can be combined with visual cues such as red flashing warnings to enhance the driver's perception of high-risk blind spots.

[0087] For medium-risk sub-areas, the system triggers local image enhancement display, automatically retrieves camera footage from the location where the target disappears, and displays the original view and wide-angle extended view on the in-vehicle display screen in a split-screen format, providing wider field of view coverage and spatial perception assistance to help the driver further confirm the blind spot status.

[0088] Low-risk sub-regions are only continuously monitored in the background, without actively triggering early warnings.

[0089] This invention dynamically divides blind spot areas into high-risk, medium-risk, and low-risk sub-regions based on a conflict time window, and implements differentiated early warning strategies according to the risk level of each sub-region, thereby achieving refined hierarchical management of potential collision risks. This method not only dynamically assesses the threat level based on the spatiotemporal relationship between the target and the vehicle, but also adopts corresponding early warning forms such as visual cues and auditory alarms for different risk levels, thereby improving the response accuracy of the early warning system and driving safety. Through this mechanism, the system can adaptively identify and classify the potential collision risks of blind spot targets in complex traffic environments, ensuring timely, reliable, and clearly graded risk warnings for drivers or autonomous driving systems in different driving scenarios, assisting them in making scientific and efficient driving decisions.

[0090] Example 2

[0091] See Figure 4 As shown, the present invention proposes a blind spot warning system for buses, including the following modules: a field of view generation module: generating a first panoramic field of view coverage area under straight-line conditions and generating a second panoramic field of view coverage area under turning conditions through a wide-angle camera group deployed around the vehicle body.

[0092] Field of view missing identification module: connected to the field of view generation module, used to spatially overlay the first panoramic field of view coverage area and the second panoramic field of view coverage area, and mark the field of view missing area of ​​the second panoramic field of view coverage area relative to the first panoramic field of view coverage area.

[0093] Dynamic target detection module: connected to the field of view missing recognition module, it performs an adjacent region search operation within the second panoramic field of view coverage area based on the spatial location and contour boundary information of the field of view missing area to determine the visible area adjacent to the missing area, and then performs dynamic target detection on the visible area adjacent to the field of view missing area, recording the last spatial coordinates and motion vector of the target at the moment of disappearance.

[0094] Blind Spot Risk Analysis Module: Connected to the Dynamic Target Detection Module, it is used to generate a predicted trajectory line based on the motion vector, starting from the last spatial coordinates, and calculate the conflict time window between the predicted trajectory and the vehicle trajectory by combining the real-time motion trajectory of the bus.

[0095] Blind Spot Risk Warning Module: Connected to the blind spot risk analysis module, it is used to mark and warn of blind spots in areas with missing vision based on conflict time windows.

[0096] Example 3

[0097] This invention proposes a device comprising a processor, and memory and a network interface connected to the processor; the network interface is connected to non-volatile memory in a server; when the processor is running, it retrieves a computer program from the non-volatile memory through the network interface and runs the computer program through the memory to execute a blind spot warning method for buses according to this invention.

[0098] Example 4

[0099] This invention proposes a storage medium on which a computer program is burned, and the computer program, when running in the memory of a server, implements the blind spot warning method for buses described in this invention.

[0100] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0101] Those skilled in the art will recognize that the algorithmic steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0102] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0103] 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.

[0104] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for blind spot warning of a passenger vehicle, characterized in that, Includes the following steps: S1. A first panoramic field of view coverage area is generated in the straight-line driving condition by a wide-angle camera group deployed around the vehicle body, and a second panoramic field of view coverage area is generated in the turning condition. S2. Spatially overlay the first panoramic field of view coverage area and the second panoramic field of view coverage area, and mark the field of view missing area of ​​the second panoramic field of view coverage area relative to the first panoramic field of view coverage area; S3. Based on the spatial location and contour boundary information of the missing area, perform an adjacent area search operation within the second panoramic field of view coverage to determine the visible area adjacent to the missing area. S4. Perform dynamic target detection on the visible area adjacent to the missing area of ​​vision, and record the last spatial coordinates and motion vector of the target when it disappears; S5. Starting from the last spatial coordinates, generate the predicted trajectory line according to the motion vector, and calculate the conflict time window between the predicted trajectory and the vehicle trajectory by combining the real-time motion trajectory of the bus. S6. Based on the conflict time window, blind spot zoning and warning are issued for areas with missing vision.

2. The method for blind spot warning of a passenger vehicle as described in claim 1, characterized in that: The specific implementation of S1 is as follows: When the vehicle is traveling straight, the wide-angle camera array deployed around the vehicle body is activated, and each camera simultaneously acquires images. Construct a rectangular projection coordinate system parallel to the ground with the longitudinal axis of the vehicle body as the axis of symmetry; Distortion correction and coordinate transformation are performed on the original wide-angle images output by each camera, and each image is projected from its original viewpoint onto the aforementioned rectangular projection coordinate system; The corrected images are stitched together to form a continuous rectangular field of view; When the turn signal is activated or the steering wheel angle exceeds a preset threshold, the wide-angle camera group located around the vehicle is activated simultaneously to start real-time image acquisition. The projected coordinate system, with the longitudinal axis of the vehicle body as the axis of symmetry, is dynamically adjusted according to the steering angle of the front wheels. The original images output by each wide-angle camera are subjected to distortion correction and coordinate transformation and then mapped to a projection coordinate system that is dynamically adjusted according to the steering angle of the vehicle's front wheels. The corrected images are stitched together to form a continuous curved edge field of view.

3. The method for blind spot warning of a passenger vehicle as described in claim 1, characterized in that: S2 includes the following: The closed envelope boundaries of the outer contours of the first and second panoramic view coverage areas are extracted respectively. Using the closed envelope boundary of the first panoramic field of view coverage area as a reference, the closed envelope boundary of the second panoramic field of view coverage area is geometrically superimposed on it, and the first panoramic field of view coverage area that is not covered by the second panoramic field of view coverage area is extracted as the field of view missing area.

4. The method for blind spot warning of a passenger vehicle as described in claim 1, characterized in that: S3 includes the following: Based on the spatial location and contour boundary information of the missing field of view area, the contour boundary of the missing field of view area is marked within the second panoramic field of view coverage area. Based on the contour boundary of the missing field of view region, perform boundary adjacency analysis to identify image regions that share a common boundary with the missing field of view region, and further trace the boundary features of these image regions until continuous image regions without image stitching are identified. These continuous regions that are not affected by stitching and share a common boundary with the missing field of view region are defined as the adjacent visible range. The adjacent visible range is divided according to the image acquisition source to which it belongs, generating adjacent visible sub-regions.

5. The method for blind spot warning of a passenger vehicle as described in claim 1, characterized in that: S4 includes the following: In real time, moving objects entering the area are identified within the adjacent visible sub-regions surrounding the area of ​​visual loss. Once a dynamic target is detected, its trajectory is continuously tracked. If a dynamic target disappears from its visible sub-region in the current image frame during tracking, a target matching search operation is performed on the spatial adjacent region of the visible sub-region within a set time window to determine whether the target has migrated to an adjacent field of view. If the target is not detected again in the adjacent region, the time when the target disappeared is recorded and the adjacent visible sub-region where it disappeared is located. Starting from the moment of disappearance, a set number of historical image frames are traced backward to extract the spatial location information of the target in consecutive frames; Calculate the displacement vector of the centroid between adjacent frames for the target's position coordinates in consecutive frames.

6. The method for blind spot warning of a passenger vehicle as described in claim 5, characterized in that: S5 specifically includes the following: Based on the direction of the motion vector of the dynamic target when it disappears in the adjacent visible sub-region, it is determined whether it points to the area of ​​missing vision. If it points to the area of ​​missing vision, linear extrapolation is used to generate a straight-line predicted trajectory along the direction of its motion vector, starting from the last visible position of the target. The real-time bus trajectory is determined based on the vehicle's steering wheel angle and vehicle speed. The predicted trajectory of the dynamic target entering the area of ​​missing vision is geometrically compared with the trajectory of the bus to determine whether the two trajectories intersect. If they intersect, the shortest spatial distance between the two trajectories is calculated and compared with the set safety threshold. If the shortest spatial distance is less than the safety threshold, the conflict time window is calculated based on the shortest spatial distance and the relative speed between the target and the vehicle. When there is no intersection, the shortest spatial distance between the two trajectories is calculated. If the shortest spatial distance is less than the safety threshold, the spatial distance between the two trajectories is updated. If the distance is detected to be continuously shortening in multiple consecutive time steps, the conflict time window calculation is re-triggered. When not pointing to a region with a missing field of view, target search is performed in other visible sub-regions pointed to by its motion vector direction. If the target is detected again in other visible sub-regions, continuous tracking of it is resumed and its trajectory and motion vector are updated. If the target is not detected, regional monitoring and re-identification attempts are made based on the target's last motion trend until it is confirmed that the target has left the system's field of view or reappears.

7. The method for blind spot warning of a passenger vehicle as described in claim 1, characterized in that: S6 includes the following: Based on the conflict time window, the area with missing vision is divided into: High-risk sub-region: The conflict time window is less than the lower limit of the warning range; Medium-risk sub-zone: The conflict time window is within the warning range; Low-risk sub-region: The conflict time window is greater than the upper limit within the warning range; For high-risk sub-areas, an audio-visual directional alarm is triggered, generating a directional sound beam through the phase array speaker built into the car door, and the focus of the sound beam locks onto the coordinates of the target's disappearance. For medium-risk sub-areas, trigger local image enhancement display, automatically retrieve camera footage from the location where the target disappears, and display the original view and wide-angle extended view on the vehicle display screen in a split-screen format; Low-risk sub-regions are only continuously monitored in the background, without actively triggering early warnings.

8. A blind spot warning system for buses, characterized in that, Includes the following modules: Field of view generation module: A wide-angle camera group deployed around the vehicle body generates a first panoramic field of view coverage area in straight driving conditions and a second panoramic field of view coverage area in turning conditions. Field of view missing identification module: Spatially overlay the first panoramic field of view coverage area and the second panoramic field of view coverage area, and mark the field of view missing area of ​​the second panoramic field of view coverage area relative to the first panoramic field of view coverage area; Dynamic target detection module: Based on the spatial location and contour boundary information of the missing area in the field of view, it performs an adjacent region search operation within the second panoramic field of view coverage to determine the visible area adjacent to the missing area, and then performs dynamic target detection on the visible area adjacent to the missing area, recording the last spatial coordinates and motion vector at the moment the target disappears. Blind spot risk analysis module: Starting from the last spatial coordinates, a predicted trajectory line is generated according to the motion vector, and the conflict time window between the predicted trajectory and the vehicle trajectory is calculated by combining the real-time motion trajectory of the bus. Blind spot risk warning module: Based on the conflict time window, blind spot zoning and warning are given for areas with missing vision.

9. A device, characterized in that: The device includes a processor, and memory and a network interface connected to the processor; the network interface is connected to non-volatile memory in a server; the processor, during operation, retrieves a computer program from the non-volatile memory via the network interface and runs the computer program via the memory to perform the method described in any one of claims 1-7.

10. A storage medium, characterized in that: The storage medium is programmed with a computer program, which, when run in the server's memory, implements the method described in any one of claims 1-7.

Images