A kind of cumulative transfer coefficient calculation based on multi-vehicle light signal and related equipment

Abstract

The application provides a kind of based on the cumulative transfer coefficient calculation method of multi-vehicle light signal and related equipment, belongs to intelligent driving environment perception technical field.The method includes: data preprocessing, by vanishing point search and ray matching determination vehicle lane attribution and calculate longitudinal distance;Build influence transfer path, upstream vehicle is divided into direct influence layer and indirect influence layer;Respectively calculate brake channel and steering channel transmission coefficient and weighted fusion is direct transmission coefficient;Time series smoothing processing suppresses detection noise;According to lane recursive calculation, the cumulative transfer coefficient of each upstream vehicle is introduced based on the attenuation coefficient of distance;Output self cumulative transfer coefficient, immediately preceding vehicle brake probability and maximum risk source tracing result.The application does not need to retrain model or install V2X equipment, can be embedded as post-processing module in existing perception system, can be in advance to predict the brake intention of immediately preceding vehicle, increase the reaction time of ego vehicle, applicable to straight dense traffic following scenario.

Description

Technical Field

[0001] This application relates to the field of intelligent driving environment perception technology, and in particular to a method and related equipment for calculating the cumulative transmission coefficient based on multi-vehicle light signals. Background Technology

[0002] Following a dense flow of traffic on a straight road is a typical scenario in intelligent driving environmental perception systems. The core challenge lies in the fact that the longitudinal deceleration and lateral lane-changing intentions of upstream vehicles are transmitted step by step along the traffic flow, eventually accumulating and affecting the vehicle itself. Accurate identification of this traffic flow transmission effect is directly related to the safety of following the vehicle and the timeliness of decision-making.

[0003] Currently, most vehicle perception and decision-making systems only focus on the current light signal status of the vehicle immediately in front, including brake lights, turn signals, and hazard lights, without fully considering the cascading and cumulative effects of light signals from multiple vehicles upstream. Due to this limitation, vehicles typically wait until the lights of the vehicle immediately in front illuminate before responding, resulting in insufficient buffer time and increasing the risk of rear-end collisions.

[0004] Some studies attempt to use multi-vehicle light signals for risk prediction. Existing technologies include patents that trigger active safety mechanisms by detecting the brake lights of the vehicle ahead, as well as technical solutions that use vehicle light detection to predict steering and braking actions. Some researchers have also proposed vehicle driving intention recognition models based on taillight signals, combining deep learning and color feature extraction methods. However, these solutions typically require retraining deep learning models or adding V2X communication equipment, resulting in high computational costs, limited generalization ability, and significant deployment difficulties. Especially for special light signals such as hazard lights, existing methods lack unified risk transmission handling rules.

[0005] Furthermore, existing risk quantification methods based on vehicle spacing fail to adequately address the propagation path, attenuation patterns, and multi-path accumulation issues of light signals in traffic flow, and do not fully utilize the intentions of vehicles further upstream. Regarding the tracing of the maximum risk source, existing solutions mostly employ fixed constant thresholds, failing to establish a dynamic correlation with the actual risk intensity.

[0006] Therefore, there is an urgent need for a method that does not change the existing perception model, has high computational efficiency, and can be embedded as a post-processing module to uniformly process brake lights, turn signals, and hazard lights, quantify the cumulative impact of multiple vehicle light signals, and provide a more accurate risk assessment basis for the vehicle. Summary of the Invention

[0007] In view of the problems that existing technologies mainly focus on the light signals of the vehicle immediately in front while ignoring the upstream cumulative effect, the need to retrain the model or add communication equipment to utilize the light signals of multiple vehicles, resulting in high costs, and the lack of unified processing rules for the special light signals of hazard lights, the main purpose of this application is to propose a method for calculating the cumulative transmission coefficient based on the light signals of multiple vehicles, an electronic device, a storage medium, and a program product. This application mainly solves one of the following technical problems: (1) Without changing the existing vehicle detection and light recognition model, it only uses the light signal information of multiple vehicles through the post-processing module to analyze the impact of upstream vehicles on the vehicle in layers, predict the braking intention of the vehicle immediately in front, and distinguish between direct and indirect sources of influence. (2) It quantifies the step-by-step transmission path, attenuation law, and cumulative impact of multiple paths on the vehicle in dense traffic flow, and uniformly processes the risk transmission logic of hazard lights in different lane scenarios. (3) It avoids the instantaneous misjudgment caused by single-frame detection errors, such as reflection, shadow, and detection frame jitter, improves the stability of the risk transmission system, and realizes the dynamic tracing of the maximum risk source, making the risk source judgment more reasonable.

[0008] To achieve the above objectives, one aspect of this application proposes a method for calculating the cumulative transmission coefficient based on multi-vehicle headlight signals, the method comprising: The data preprocessing step involves obtaining vehicle detection box information and light status information output by the target detection model, determining the lane to which each vehicle belongs and the longitudinal distance between vehicles based on the detection box information, and assigning a stable number to each vehicle. The influence transmission path construction steps are as follows: Based on the relative position relationship of vehicles, a one-way hierarchical transmission rule is set. Starting from the upstream vehicle, effective vehicle pairs are searched layer by layer downstream to construct all effective transmission paths with the self-vehicle as the endpoint. The upstream vehicles are divided into first-layer vehicles that have a direct transmission relationship with the self-vehicle and second-layer vehicles that only indirectly affect the self-vehicle through the first-layer vehicles. The transmission coefficient calculation steps are as follows: calculate the braking channel transmission coefficient and the steering channel transmission coefficient separately, and then weight and fuse the braking channel transmission coefficient and the steering channel transmission coefficient to obtain the direct transmission coefficient between the vehicle pairs; wherein, the braking channel transmission coefficient is calculated based on the brake light status of the preceding vehicle and the longitudinal distance between the two vehicles, and the steering channel transmission coefficient is calculated based on the turn light status of the preceding vehicle, the relative lane positions of the two vehicles, the light status of the following vehicle, the lateral distance between the two vehicles, and the lateral overlap. The temporal smoothing step performs temporal weighted smoothing on the direct transfer coefficients of multiple consecutive frames to suppress single-frame detection noise; The chain-like accumulation step calculates the cumulative transmission coefficient of each upstream vehicle by recursively calculating the cumulative transmission coefficient of each vehicle along the transmission direction of the lane, and introduces an attenuation coefficient based on longitudinal distance in the recursion process. The output step outputs the vehicle's cumulative transmission coefficient, the braking probability of the adjacent vehicle ahead, and the tracing result of the maximum risk source based on the cumulative transmission coefficient.

[0009] In some embodiments, determining the lane to which each vehicle belongs based on the detection frame information includes: The horizontal coordinate of the vanishing point is fixed at half the width of the image, and the candidate vanishing points are traversed within the corresponding vertical coordinate interval with a fixed step size. For each candidate vanishing point, generate a set of rays originating from that candidate vanishing point and covering the road surface area; Calculate the vertical distance from each vehicle's grounding point to each ray, and match the vehicle to the ray with the smallest distance. If the distance is less than the dynamic distance threshold, the vehicle is determined to belong to the corresponding ray. Count the number of valid rays with matching vehicles under each candidate vanishing point, and merge adjacent valid rays with an angle difference less than the merging threshold into the same lane. The number of merged rays is the number of lanes corresponding to the candidate vanishing point. The candidate vanishing point with the fewest rays after merging is selected as the optimal vanishing point, and the vehicle number and ray direction corresponding to each lane are output accordingly.

[0010] In some embodiments, the dynamic distance threshold The calculation formula is:

[0011] in, The dynamic distance threshold is... The coordinates of the center point of the vehicle's bottom are shown in the image. The image coordinates of the vanishing point. This is the preset scaling factor.

[0012] In some embodiments, the longitudinal distance between the vehicles is calculated in the following manner:

[0013]

[0014] in, The longitudinal distance from the vehicle to the onboard camera. For the camera's equivalent focal length, The height of the camera's optical center from the ground. The vertical coordinate of the center point of the vehicle's bottom is shown in the image. The vertical coordinate of the vanishing point image. The minimum constant to prevent numerical overflow; longitudinal spacing between front and rear vehicles It is determined by the absolute value of the difference in the longitudinal distance between the two vehicles.

[0015] In some embodiments, the unidirectional hierarchical transmission rule includes: Within the same lane, only adjacent vehicles in front and behind are allowed to transmit influence directly; vehicles in between are not allowed to be bypassed. Only longitudinally adjacent vehicles with no intermediate vehicles are allowed to directly transfer influence between adjacent lanes; transmission across multiple lanes is prohibited. The transmission direction is along the traffic flow direction, and it is transmitted only from the vehicle in front to the vehicle behind.

[0016] In some embodiments, the braking channel transmission coefficient The calculation formula is:

[0017] in, , This indicates the status of the brake lights of the vehicle in front; 1 represents the indicator when the lights are on and 0 represents the indicator when they are off. This is the dual-flash enhancement factor; The initial value of the transmission coefficient based on the longitudinal distance is calculated using the following formula:

[0018] in, The longitudinal distance between the front and rear vehicles. The longitudinal half-effect distance; The braking channel transfer coefficient is only effective between vehicles in the same lane and longitudinally adjacent; when the hazard lights of the vehicle in front are detected to be on, the braking channel transfer coefficient is not limited to the same lane, and the vehicle in front is allowed to calculate the transfer coefficient for the vehicle in the same lane and the vehicle in the adjacent lane that is longitudinally adjacent to it.

[0019] In some embodiments, the steering channel transmission coefficient The calculation formula is:

[0020] in, This is the interaction coefficient, and its value is determined based on the lane position of the vehicle in front relative to the vehicle behind and the light status of the vehicle behind. The lateral overlap between the vehicle in front and the vehicle behind; The lateral distance between the two vehicles; This is the lateral half-effect distance; When the vehicle in front has not turned on the turn signal on one side, When the vehicle in front turns on its hazard lights, .

[0021] In some embodiments, the interaction coefficient Determined according to the following rules: When the vehicle in front and the vehicle behind are in the same lane: the vehicle behind has no lights on. = 0.5, when the following vehicle illuminates its turn signal on the same side. = 0.3, when the following vehicle illuminates its opposite turn signal. = 0.7; When the vehicle in front is in the adjacent lane and its turn signal is pointing towards the lane where the vehicle behind is located (i.e., the target lane): the vehicle behind has no turn signal. = 1.0, when the following vehicle illuminates its turn signal on the same side. = 0.8, when the following vehicle illuminates its opposite turn signal. = 0.4; When the vehicle in front is in the adjacent lane and its turn signal is in the opposite lane (i.e., the opposite lane) from the vehicle behind: = 0.

[0022] In some embodiments, the lateral overlap Calculated in the following way: Actual lateral coordinates of the center point of the bottom of the front vehicle Centered on standard vehicle width Determine the lateral section of the vehicle ahead for width. ; The actual lateral coordinates of the center point of the bottom of the vehicle Centered on standard vehicle width Determine the lateral section of the rear vehicle for the width. ; Horizontal overlap length ; lateral overlap .

[0023] In some embodiments, the formula for calculating the direct transfer coefficient is:

[0024] in, , This represents the weighting coefficient for the corresponding channel.

[0025] In some embodiments, the calculation formula for the time-weighted smoothing process is:

[0026] in, For the first Direct transfer coefficients after frame smoothing For the first Direct transfer coefficients after frame smoothing The original calculated value for the current frame. Historical retention factor; If the original calculated value becomes unavailable due to detection failure in the current frame, then retain... .

[0027] In some embodiments, the formula for calculating the cumulative transmission coefficient of each upstream vehicle is as follows:

[0028] in, For vehicles The cumulative transfer coefficient, For all those that can directly affect vehicles upstream vehicle collection, For vehicles right The direct transmission coefficient, For vehicles and The attenuation coefficient between them; The recursion is performed in order of longitudinal distance between vehicles, from farthest to nearest, with the cumulative transfer coefficient of the upstream vehicle being the highest value. .

[0029] In some embodiments, the vehicle cumulative transfer coefficient The calculation formula is:

[0030] in, This refers to the set of all upstream vehicles that can directly affect the vehicle itself. For vehicles The cumulative transfer coefficient, For vehicles The direct transmission coefficient to the vehicle. For vehicles The attenuation coefficient between the vehicle and the vehicle itself.

[0031] In some embodiments, the tracing of the maximum risk source includes: For the first level vehicle Calculate each of its upstream vehicles right Influence intensity candidate values ,in For vehicles arrive The direct value passed, For vehicles The cumulative transfer coefficient, For vehicles and The attenuation coefficient between them; calculate The direct value passed to the vehicle is used as its own candidate value; Compare all candidate values ​​and take the source vehicle corresponding to the maximum value as the output of the maximum risk source; if its own candidate value is the maximum, it is determined to be the risk source; if its own candidate value and the upstream vehicle's candidate value are tied for the maximum, they are both marked as the joint maximum risk source.

[0032] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described above.

[0033] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described above.

[0034] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer program product, including a computer program that, when executed by a processor, implements the method described above.

[0035] The embodiments of this application include at least the following beneficial effects: This application provides a method, electronic device, storage medium, and program product for calculating the cumulative transmission coefficient of multi-vehicle light signals. This solution constructs a hierarchical transmission path of upstream vehicle light signals along the traffic flow queue, independently calculates the dual-channel transmission coefficients of braking and steering, and then weights and fuses them. Combined with a time-series smoothing and distance attenuation recursive mechanism, it achieves accurate quantification of the cumulative impact of upstream multi-vehicle light signals. Its direct beneficial effects are: on the one hand, it does not require retraining the perception model or adding V2X communication equipment; it can be embedded into the existing system as a post-processing module, significantly reducing deployment costs and integration difficulty. On the other hand, by analyzing the cumulative effect of the indirect influence layer on the adjacent vehicle, it can predict the braking probability of the adjacent vehicle before its own brake lights are lit, thus gaining valuable reaction time for the vehicle's active safety system. At the same time, combined with dual-channel differentiated weights and dynamic risk source tracing, it effectively distinguishes between longitudinal rear-end collision and lateral cutting-in risks, improving the accuracy and stability of risk judgment in dense following scenarios. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating a method for calculating the cumulative transmission coefficient based on multi-vehicle headlight signals, provided in an embodiment of this application. Figure 2 This is a flowchart of the calculation of the dual-channel transmission coefficients for the braking and steering channels in the implementation of this application; Figure 3 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0038] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0039] like Figure 1 As shown, this embodiment provides a method for calculating the cumulative transmission coefficient based on multi-vehicle light signals, specifically including the following steps: Step S1: Data Preprocessing. The number of lanes is estimated and the lane to which a vehicle belongs is determined by vanishing point search and ray matching methods. The longitudinal distance between vehicles is calculated using a pinhole model, and a stable number is assigned to each vehicle using a lightweight tracking method.

[0040] In one embodiment, the method for determining the lane to which each vehicle belongs includes: fixing the horizontal coordinate of the vanishing point to half the image width, and traversing candidate vanishing points within the corresponding vertical coordinate interval with a fixed step size; for each candidate vanishing point, generating a set of rays originating from the candidate vanishing point and covering the road surface area; calculating the vertical distance from each vehicle's ground point to each ray, matching the vehicle to the ray with the smallest distance, and determining that the vehicle belongs to the corresponding ray if the distance is less than a dynamic distance threshold; counting the number of valid rays with matching vehicles under each candidate vanishing point, merging adjacent valid rays with an angle difference less than a merging threshold into the same lane, and the number of merged rays is the number of lanes corresponding to the candidate vanishing point; selecting the candidate vanishing point with the fewest merged rays as the optimal vanishing point, and outputting the vehicle number and ray direction corresponding to each lane accordingly.

[0041] Dynamic distance threshold The calculation formula is:

[0042] in, The dynamic distance threshold is... The coordinates of the center point of the vehicle's bottom are shown in the image. The image coordinates of the vanishing point. This is the preset scaling factor.

[0043] The longitudinal distance between vehicles is calculated in the following way:

[0044]

[0045] in, The longitudinal distance from the vehicle to the onboard camera. For the camera's equivalent focal length, The height of the camera's optical center from the ground. The vertical coordinate of the center point of the vehicle's bottom is shown in the image. The vertical coordinate of the vanishing point image. The minimum constant to prevent numerical overflow; longitudinal spacing between front and rear vehicles It is determined by the absolute value of the difference in the longitudinal distance between the two vehicles.

[0046] Step S2: Construct the impact transmission path. Set a one-way hierarchical transmission rule, allowing direct transmission of impact only between adjacent vehicles in the same lane and between longitudinally adjacent vehicles in adjacent lanes.

[0047] In one embodiment, the specific rules include: within the same lane, only adjacent vehicles in front and behind are allowed to directly transfer influence, and skipping intermediate vehicles is not allowed; between adjacent lanes, only longitudinally adjacent vehicles with no intermediate vehicles in their respective lanes are allowed to directly transfer influence, and transferring influence across multiple lanes is prohibited; the direction of transfer is along the traffic flow direction, and it is only transferred from the preceding vehicle to the following vehicle.

[0048] Starting from the upstream vehicles, the search proceeds layer by layer to construct all valid transmission paths pointing to the vehicle, and the upstream vehicles are divided into the direct influence layer and the indirect influence layer.

[0049] Step S3: Calculate the transfer coefficients. Calculate the vehicle-to-vehicle transfer coefficients for the braking channel and the steering channel respectively, and then fuse the two channels into a direct transfer coefficient through linear weighting.

[0050] Specifically, the braking channel transmission coefficient The calculation formula is:

[0051] in, , This indicates the status of the brake lights of the vehicle in front; 1 represents the indicator when the lights are on and 0 represents the indicator when they are off. This refers to the hazard light enhancement factor, calculated when the hazard lights of the vehicle in front are on. = 0.5, otherwise = 0. The initial value of the transmission coefficient based on the longitudinal distance is calculated using the following formula:

[0052] in, The longitudinal distance between the front and rear vehicles. The longitudinal half-effect distance.

[0053] The braking channel transfer coefficient is only effective between longitudinally adjacent vehicle pairs in the same lane; when the hazard lights of the preceding vehicle are detected to be on, the braking channel transfer coefficient is not limited to the same lane, and the preceding vehicle is allowed to calculate the transfer coefficient for the longitudinally adjacent vehicle in the same lane and the vehicle in the adjacent lane.

[0054] Specifically, the steering channel transmission coefficient The calculation formula is:

[0055] in, This is the interaction coefficient, and its value is determined based on the lane position of the vehicle in front relative to the vehicle behind and the light status of the vehicle behind. The lateral overlap between the vehicle in front and the vehicle behind; The lateral distance between the two vehicles; This is the lateral half-effective distance. When the vehicle in front has not activated its turn signal on one side, When the vehicle in front turns on its hazard lights, .

[0056] In some embodiments, the interaction coefficient F is determined according to the following rules: When the vehicle in front and the vehicle behind are in the same lane: the vehicle behind has no lights on. = 0.5, when the following vehicle illuminates its turn signal on the same side. = 0.3, when the following vehicle illuminates its opposite turn signal. = 0.7; When the vehicle in front is in the adjacent lane and its turn signal is pointing towards the lane where the vehicle behind is located (i.e., the target lane): the vehicle behind has no turn signal. = 1.0, when the following vehicle illuminates its turn signal on the same side. = 0.8, when the following vehicle illuminates its opposite turn signal. = 0.4; When the vehicle in front is in the adjacent lane and its turn signal is in the opposite lane (i.e., the opposite lane) from the vehicle behind: = 0.

[0057] In some embodiments, lateral overlap Calculated in the following way: Actual lateral coordinates of the center point of the bottom of the front vehicle Centered on standard vehicle width Determine the lateral section of the vehicle ahead for width. ; The actual lateral coordinates of the center point of the bottom of the vehicle Centered on standard vehicle width Determine the lateral section of the rear vehicle for the width. ; Horizontal overlap length ; lateral overlap .

[0058] Specifically, the formula for calculating the direct transfer coefficient is as follows:

[0059] in This is the weighting coefficient for the braking channel. This is the weighting coefficient for the turning channel. > ,and and The value of is such that the maximum comprehensive coefficient for a single pair of vehicles is equal to the cumulative upper limit of 100.

[0060] Step S4: Temporal smoothing and chained accumulation. Multi-frame transfer coefficients are exponentially weighted smoothed to suppress detection noise; risks are accumulated by lane, and distance-based attenuation is considered in the recursion.

[0061] Specifically, the calculation formula for time-weighted smoothing is as follows:

[0062] in, For the first Direct transfer coefficients after frame smoothing For the first Direct transfer coefficients after frame smoothing The original calculated value for the current frame. This is the historical retention factor, with a value ranging from 0.6 to 0.8.

[0063] If the original calculated value becomes unavailable due to detection failure in the current frame, then retain... .

[0064] Specifically, the cumulative transmission coefficient for each upstream vehicle is calculated recursively by lane:

[0065] in, For vehicles The cumulative transfer coefficient, For all those that can directly affect vehicles upstream vehicle collection, For vehicles right The direct transmission coefficient, For vehicles and The attenuation coefficient between them.

[0066] The recursion is performed in order of longitudinal distance between vehicles, from farthest to nearest, with the cumulative transfer coefficient of the upstream vehicle being the highest value. .

[0067] Step S5: Output the results. Output the directly passed value, the probability of braking of the vehicle immediately preceding it, and the maximum risk source.

[0068] Self-accumulated transmission coefficient The calculation formula is:

[0069] in, This refers to the set of all upstream vehicles that can directly affect the vehicle itself. For vehicles The cumulative transfer coefficient, For vehicles The direct transmission coefficient to the vehicle. For vehicles The attenuation coefficient between the vehicle and the vehicle itself.

[0070] Probability of braking close to the vehicle in front The cumulative transmission coefficient received by the vehicle immediately preceding it is determined as follows: Cut off to the interval between 0 and 100 to obtain If the current frame detects the brake lights of the vehicle immediately preceding it are illuminated, then Otherwise, according to the piecewise linear function Mapped to :when hour, ;when hour, ;when hour, .

[0071] When tracing the source of the greatest risk, calculate the value of each upstream vehicle. right Influence intensity candidate values ,in For vehicles arrive The direct value passed, For vehicles The cumulative transfer coefficient, For vehicles and Attenuation coefficient between; calculation The direct transmission value of the vehicle is used as its own candidate value; all candidate values ​​are compared, and the source vehicle corresponding to the maximum value is taken as the output of the maximum risk source; if the vehicle's own candidate value is the largest, the risk source is determined to be itself; if the vehicle's own candidate value and the upstream vehicle's candidate value are tied for the largest, they are both marked as the joint maximum risk source.

[0072] Below, in conjunction with Figure 2 The solutions of the embodiments of this application will be described in detail and explained with specific application examples.

[0073] (1) Data acquisition and preprocessing 1.1) Vehicle light status extraction This embodiment uses an existing vehicle detection model as the perception module to extract the vehicle light status and obtain the left turn signal indicator. Right turn signal sign Brake light markings The value 1 indicates the lights are on, and 0 indicates they are off. Table 1 shows the interpretation of driving scenarios corresponding to different light combinations. In the experimental verification, the vehicle's light status was temporarily set to off; in actual system deployment, it can be directly replaced by the real-time vehicle light status output by the perception module. Furthermore, the vehicle position information output by the model is processed to obtain the center point coordinates of the vehicle detection box's grounding point, which serves as the basic input data for calculating the transfer coefficient.

[0074]

[0075] 1.2) Lane number estimation and vehicle lane assignment determination In single-frame pure visual perception output, when lane number estimation and vehicle affiliation determination are based solely on the bottom center point of the vehicle detection box, perspective effect causes the lateral position of distant vehicles on the image to be severely compressed. The difference in the horizontal coordinate of the bottom center point of near and far vehicles in the same lane is significant. Traditional clustering methods based on Euclidean distance or horizontal coordinate thresholds are difficult to obtain stable and accurate lane division results.

[0076] To address this requirement, this embodiment proposes a lane segmentation method based on vanishing point search and ray matching. Leveraging the prior condition that the road vanishing point in a straight road scenario lies on the vertical line at the image center, the horizontal coordinate of the vanishing point is fixed at half the image width. Candidate vanishing points are traversed within the corresponding vertical coordinate interval with a fixed step size. For each candidate point, a set of rays originating from that point and evenly distributed between the lower left and lower right corners of the image is generated, completely covering the road surface area. Subsequently, the vertical distance from each vehicle's ground contact point to each ray is calculated, and the vehicle is matched to the ray with the smallest distance. If this distance is less than a set threshold... If the vehicle is within the corresponding ray, it is determined to belong to that ray; otherwise, it is determined that the vehicle is not covered by the road surface ray. Candidate vanishing points that can cover all vehicles are selected, and the number of valid rays matching a vehicle at each candidate point is counted. Rays with angle differences less than a threshold are then selected. Adjacent valid rays are merged into the same lane, and the number of rays after merging is the number of lanes corresponding to that candidate point. After traversing all candidate vanishing points, the one with the fewest merged rays is selected as the optimal vanishing point. Finally, the vehicle number and ray direction corresponding to each lane are output based on this, completing the vehicle lane assignment statistics.

[0077] For the vehicle itself, since it is outside the image detection box, and considering that rear-end collision risk is the main source of risk in car-following scenarios, it is necessary to accurately characterize the longitudinal relationship between the rear of the preceding vehicle and the front of the vehicle. The position of the preceding vehicle's rear is described by the center point of its detection box at the bottom, while the position of the vehicle's front is determined based on the assumption that the camera is centered and its optical axis is parallel to the road surface. The center point of the bottom of the image represents the ground contact point of the vehicle's front. Therefore, this study sets the vehicle's ground contact point at the center point of the bottom of the image, which corresponds to the road surface point directly below the camera, consistent with the assumption that the longitudinal centerline of the vehicle coincides with the vertical line at the center of the image in straight road scenarios.

[0078] Threshold for merging angle difference These are preset constants, determined through offline calibration. The calibration process is as follows: Select a typical straight road surface image. Within the road surface area at the bottom of the image, determine the intersection points of each lane line with the straight line containing the bottom edge of the image. Connect these intersection points with the true vanishing point to obtain the reference lines corresponding to each lane. Calculate the angle difference between adjacent reference lines within the same lane. Repeat the above measurement for multiple collected calibration images, and take the average of the measured angle differences as the merging threshold. After completing the measurement of multiple sets of straight road images, take the average of all effective angle differences as the final lane merging threshold. In this embodiment, .

[0079] Dynamic distance threshold Defined as:

[0080] in As a preset scaling factor, the coordinates of the center point image at the bottom of the vehicle are: The vanishing point image coordinates are Image coordinate system Using the top-left corner of the image as the origin, the x-axis points horizontally to the right, and the y-axis points vertically downwards, with units in pixels. This design makes the threshold proportional to the longitudinal distance of the vehicle, automatically adapting to scale changes in perspective projection: closer vehicles are allowed larger pixel offsets, while distant vehicles require more precise alignment. (Coefficient) As a fixed constant, this embodiment takes... This coefficient was determined through offline calibration in multiple scenarios.

[0081] Based on the lane statistics and vehicle classification results described above, the relative positions of each vehicle can be further determined. Within the same lane, vehicles are arranged from largest to smallest according to their bottom vertical coordinates to form a front-to-back queue. Between different lanes, the front and back positions can be determined by directly comparing the bottom vertical coordinates: vehicles with smaller coordinates are farther away, and those with larger coordinates are closer.

[0082] 1.3) Calculation of lateral and longitudinal distances of vehicles Based on lane assignment and vehicle sequencing, to quantify the actual longitudinal distance between vehicles, the absolute longitudinal distance between the camera installation height and the vanishing point is used as the calculation method. Let the height of the camera's optical center from the ground be... The camera's equivalent focal length is Based on the pinhole camera model and the assumption of a flat road surface, the longitudinal distance from the vehicle to the camera... satisfy:

[0083]

[0084] in The difference between the vertical coordinates of the vehicle's contact point and vanishing point in the image is given by the inverse of which is proportional to the actual distance. Pixels prevent numerical overflow. In the method presented in this paper, the vanishing point... Based on the vanishing point search process given in the previous section, the camera height... Focal length can be obtained through actual measurement. It can be obtained through camera calibration or approximate estimation.

[0085] This embodiment of the method specifies that risk is transmitted unidirectionally only from the upstream vehicle to the downstream vehicle. The longitudinal distance from the vehicle in front to the camera on the vehicle is [missing information]. The longitudinal distance from the following vehicle to the camera on the vehicle itself; the vehicle in front is located further ahead in the direction of travel and is farther away from the vehicle itself, satisfying the condition... Therefore, the longitudinal distance between the front and rear vehicles We obtain the following by subtracting the absolute distances between the two vehicles:

[0086] like If the vehicle with the larger longitudinal distance is determined, then the vehicle in front is the vehicle in front, and the other vehicle is the vehicle behind.

[0087] Regarding lateral distance, based on the pinhole camera model and the assumption of a flat road surface, the lateral offset of the vehicle's bottom center point and the longitudinal distance satisfy a linear relationship. The height of the camera's optical center from the ground is... First, calculate the normalized lateral offset:

[0088] Then the actual lateral distance of the car relative to the camera for:

[0089] Furthermore, the lateral spacing between the two vehicles This is the absolute value of the difference between the actual lateral distances of the two:

[0090] To prevent numerical overflow or invalid calculations caused by extremely distant vehicles, a distance exceeding the limit judgment is added: if the longitudinal absolute distance of a vehicle is... Or lateral absolute distance If the vehicle is deemed invalid, it will be removed from the calculation and not included in any subsequent risk assessment or path construction.

[0091] 1.4) Vehicle cross-frame tracking A lightweight tracking method based on the intersection-union ratio (IU) and size similarity of detection boxes is adopted. That is, for each detection box in the current frame, its IU with all vehicle boxes in the previous frame is calculated, and candidate matches are made for detection boxes with an IU greater than 0.3. Size similarity is then calculated for these candidate matches.

[0092] in, and These are the width and height of the detection bounding box in the current frame, respectively. and These are the width and height of the vehicle frame corresponding to the previous frame, in pixels. If the highest similarity is greater than 0.5, the corresponding vehicle frame number is reused; otherwise, a new number is assigned. The consecutive loss count of unmatched vehicle frames in the previous frame is incremented by 1. Vehicle frames and their numbers that are consecutively lost for more than 3 frames are deleted.

[0093] (2) Construction of influence transmission paths between vehicles 2.1) Vehicle transfer rules In dense traffic following scenarios, the light signals of upstream vehicles intend to be transmitted through the queue and cumulatively affect the following vehicle. To quantify this cumulative effect and avoid excessive paths, this embodiment proposes a simplified transmission path setting rule based on the relative positional relationship of vehicles, as detailed below: (a) The method in this embodiment only considers vehicles traveling in the same direction and in motion. Other vehicles do not participate in dense following queues and are not included in the transmission path construction.

[0094] (b) One-way transmission: The transmission direction is along the traffic flow direction. The vehicle behind cannot affect the vehicle in front. Since there is no reverse action in physical space and driving logic, the influence is only transmitted from the vehicle in front to the vehicle behind.

[0095] (c) Continuous signal transmission within the same lane: In a close-following situation, vehicles in the same lane are closely arranged. The light signal of the vehicle in front will first act on the vehicle immediately behind it, and then be transmitted sequentially to the rear. If the signal is transmitted directly without skipping the middle vehicles, it does not conform to the signal propagation pattern of real traffic flow and deviates from the actual vehicle interaction logic.

[0096] (d) Longitudinal Adjacent Transmission in Adjacent Lanes: A vehicle's light signal can only be effectively perceived by the vehicle behind it in the immediately adjacent lane, and the longitudinal adjacent condition must be met: that is, there are no other vehicles between the two vehicles in their respective lanes. When there are two or more lanes between them or when there are vehicles in the middle of the same lane, the signal is difficult to transmit effectively. Therefore, only vehicles in adjacent lanes that are longitudinally adjacent are allowed to transmit the signal directly. Transmission across multiple lanes or skipping lanes is prohibited.

[0097] 2.2) Complete Path Construction and Hierarchical Division Starting from the upstream vehicle, search for valid adjacent vehicle pairs layer by layer downstream, constructing all possible influence transmission relationships. The specific rules are as follows: (a) The same-lane passing rule requires that only the immediately preceding and following vehicles be connected within the current lane, and that intermediate vehicles not be skipped; (b) The longitudinal adjacent transfer rule for adjacent lanes requires that a vehicle moves from one lane to the next vehicle in the adjacent lane where there are no other vehicles in the lane between the two vehicles and the vehicle in the adjacent lane, and the longitudinal coordinate of the vehicle in the adjacent lane is greater than that of the vehicle in front.

[0098] (c) The path search starts from the upstream vehicle and proceeds downwards layer by layer along the traffic flow direction, ending with the vehicle itself. During the search, only paths ending with the vehicle itself are retained; paths not ending with the vehicle itself are considered invalid paths and are not included in subsequent calculations.

[0099] (d) Based on whether there is a direct transmission relationship between the vehicle and the vehicle itself, upstream vehicles are divided into two levels. The first level consists of vehicles that have a direct transmission relationship with the vehicle itself, namely vehicles that are adjacent to the vehicle in the same lane or longitudinally adjacent to the vehicle in an adjacent lane; the second level consists of other upstream vehicles that indirectly affect the vehicle only through the vehicles in the first level.

[0100] This results in a set of effective transmission sequences from upstream vehicles to the vehicle itself, and each sequence is called a transmission path.

[0101] (3) Calculation of vehicle-to-vehicle transmission coefficient See Figure 2 The algorithm consists of two independent channels: braking and steering. The braking channel uses the brake light status of the preceding vehicle and the longitudinal distance between the two vehicles as core inputs to determine whether the braking effect is effectively transmitted and quantify its intensity. The steering channel uses the turn signal status of one side of the preceding vehicle, the lateral distance between the two vehicles, and the lateral overlap rate as core inputs to calculate the degree of lateral interference of the lane-changing intention on the following vehicle. The two channels are calculated independently without interference, and the final value of direct transmission between the two vehicles is obtained through weighted merging.

[0102] 3.1) Calculation of Braking Channel Transmission Coefficient To quantify the impact of the preceding vehicle's headlight status on the following vehicle, this embodiment defines a brake light influence item. This item is based on the status of the brake lights of the vehicle ahead. The calculation is used to indicate whether the vehicle ahead poses a braking risk input; the braking effect term is defined as:

[0103] in, The value is 1 when the brake light is on and 0 when it is off. The presence of the preceding vehicle's brake light indicates the existence of a risk input. Regardless of whether the following vehicle brakes, this risk objectively affects it. Therefore, the influencing factor directly uses the brake light status of the preceding vehicle, rather than the difference between the two vehicle states. This ensures that upstream risk is independent of downstream response, consistent with the actual physical logic of a car-following scenario.

[0104] Based on this, the calculation of the braking channel transmission coefficient needs to be determined in conjunction with lane position relationships. The braking channel only applies to vehicle pairs in the same lane that are longitudinally adjacent and have no intervening vehicles. If the vehicles in front and behind are not in the same lane, the braking channel transmission coefficient is directly set to zero. This is because the braking behavior of vehicles in adjacent lanes does not pose a direct longitudinal risk to the following vehicle; it is only transmitted through the steering channel when accompanied by a steering intention. Additionally, if the brake lights of the vehicle in front are not illuminated, the braking transmission coefficient is directly set to zero.

[0105] As an exception to the above rules, if the hazard lights of the vehicle ahead are detected to be on, it is considered an emergency warning event. In this case, the braking lane is no longer restricted by the same lane, allowing the vehicle to apply the transfer coefficient to vehicles immediately behind it, including vehicles in the same lane and vehicles in adjacent lanes, provided that they are longitudinally adjacent and there are no vehicles in between. For vehicles in the same lane, the original calculation method remains unchanged. This rule only applies to vehicles that simultaneously detect both brake lights and hazard lights; braking alone does not trigger this exception.

[0106] Based on the driving distance, a formula for calculating the braking effect transmission coefficient is constructed:

[0107] in: The coefficient for the transmission of the effects of braking between the two vehicles. The hazard light enhancement factor is set to 0.5 when the hazard lights are on and 0 otherwise. Considering that the initial factor cannot adequately reflect the risk of close following, a distance-based transfer factor with an initial value is introduced. The calculation formula is as follows:

[0108] In the formula, The initial values ​​for the transfer coefficients, The longitudinal driving distance between vehicles. The longitudinal half-effect distance is set to 10m. This value reduces the basic transmission coefficient to 0.5 when the longitudinal spacing is 10m. This attenuation is used to quantify the intensity attenuation of the direct response between the front and rear vehicles with distance; that is, the smaller the longitudinal spacing, the higher the initial value of the transmission coefficient; when the spacing approaches infinity, the coefficient approaches 0, which conforms to the actual driving safety law.

[0109] 3.2) Calculation of steering channel transmission coefficient When it is detected that the vehicle in front has only illuminated one side of its turn signal, its lateral impact on the vehicle behind depends not only on the turn signal status of the following vehicle itself, but also on the relative lane positions between the two vehicles. To accurately describe this relationship, an interaction coefficient is defined. This coefficient simultaneously encodes the lane position of the vehicle in front relative to the vehicle behind and the lighting response status of the vehicle behind, and is directly used for calculating the steering channel transmission coefficient.

[0110] The lane position of the vehicle in front relative to the vehicle behind can be categorized into three types: Same lane: The vehicle in front is in the same lane as the vehicle behind, with the vehicle in front traveling ahead. Target lane: The vehicle in front is in an adjacent lane, and its turn signal is pointing towards the lane of the vehicle behind, indicating that the vehicle in front is about to or intends to cut in front of the vehicle behind. Opposite lane: The vehicle in front is in an adjacent lane, but its turn signal is pointing away from the lane of the vehicle behind. In this case, the turn signal of the vehicle in front has no effective impact on the vehicle behind, and can be ignored. At the same time, the vehicle in front turns on its hazard lights. Set to 0. Interaction coefficient As shown in Table 2.

[0111]

[0112] Horizontal spacing While it can reflect the lateral proximity of two vehicles, it cannot distinguish between two states: a turn signal activated but not yet merged, and a vehicle partially entering its lane. For example, when a vehicle in an adjacent lane slowly merges, its lateral center position may be in the middle transition zone. The value is similar to some horizontally similar but non-intrusive scenarios, based solely on... Unable to distinguish between them. Therefore, this embodiment introduces lateral overlap. It directly measures the proportion by which the vehicle in front encroaches on the lateral space of the vehicle behind. Even same, The larger the value, the deeper the intrusion and the higher the risk, thus effectively distinguishing between two different levels of danger: lateral approach without intrusion and actual intrusion.

[0113] First, establish a road surface coordinate system: with the vertical projection point of the camera's optical center onto the ground as the origin. The axis points to the right side of the vehicle, i.e., laterally. The axis points directly in front of the vehicle, i.e., longitudinally, within the road surface. .

[0114] For bicycles, the lateral range is The standard vehicle width For the detected vehicles, the longitudinal distance from section 1.3) is used. and lateral distance The calculation formula yields the actual lateral coordinates of the center point at the bottom of the vehicle. To avoid errors caused by inaccurate detection frame width, this embodiment uses a standard vehicle width. The actual width of the vehicle. That is, for any vehicle, its lateral width is uniformly represented as... ,in This represents the actual lateral coordinates of the center point at the bottom of the vehicle.

[0115] Set up the front vehicle The horizontal boundary is The car behind The horizontal boundary is The horizontal overlap length is defined as:

[0116] The horizontal overlap is then:

[0117] The range of values ​​is , This indicates that the entire width of the narrower vehicle has been overlapped by the vehicle in front, meaning it has completely encroached upon the lane. Furthermore, the formula for calculating the original lane change transmission coefficient is as follows:

[0118] In the formula, The lateral distance between the two vehicles. This represents the horizontal overlap, with a value ranging from 0 to 1. The lateral half-effective distance is taken as 1.5m. This value is based on the 1.7m difference between the Chinese standard lane width of 3.5m and the width of a car of 1.8m. This ensures that the transmission coefficient decays rapidly when the lateral clearance is lower than the lane safety margin, accurately reflecting the risk of cutting in. Note that the turning lane is calculated normally when only one side of the turn signal is illuminated by the vehicle in front; when there is no turn signal, .

[0119] This algorithm does not separately classify lane-changing scenarios; instead, it distinguishes risk levels based on lateral spacing and lateral overlap, making it suitable for two typical lane-changing situations. The first type is when a vehicle from an adjacent lane cuts into the lane ahead from a distance. Gradually decrease, r The risk gradually increases from 0, propagating backward along the traffic flow; the second type occurs when a vehicle behind cuts close into the gap between the vehicle in front and the vehicle behind, at which point... extremely small and r If the value increases rapidly, the transmission coefficient will increase significantly, and this risk only directly affects the vehicle itself and is not transmitted to the rear.

[0120] 3.4) Direct transmission coefficient The braking and steering channels are calculated independently, and the final direct transfer value is obtained through linear weighting, as shown in the following formula:

[0121] in, , The weighting coefficients for the corresponding channels are set to 24 and 20 respectively. This set of values ​​ensures that the maximum comprehensive coefficient for a single pair of vehicles is exactly equal to the cumulative upper limit of 100. The weighting ratio of 6:5 reflects the judgment that longitudinal rear-end collision risk takes precedence over lateral cutting-in risk. Based on the calculated direct transmission value, which is a continuous value from 0 to 100, the intensity of the impact between vehicles can be quantified.

[0122] (4) Timing smoothing To reduce misjudgments caused by single-frame image detection errors, multiple frames of transfer coefficient data were continuously collected, and their variation over time was analyzed. To address the issues of incorrect lighting encoding and abrupt changes in transfer coefficients caused by reflections, shadows, and detection box jitter, a time-weighted smoothing mechanism was introduced to smooth the braking transfer coefficient and lane-change transfer coefficient separately. The time-weighted smoothing formula is as follows:

[0123] in For the first Direct transfer coefficients after frame smoothing The smoothing value of the previous frame. The original calculated value for the current frame. A historical retention factor of 0.7 is used to balance noise suppression and response timeliness; a typical value is 0.6~0.8. This method eliminates the need for frame delay, suppressing abnormal jitter in single frames while ensuring real-time response, thus improving the stability of the risk transmission system.

[0124] To prevent a rapid drop in coefficients due to vehicle detection loss or abnormal light recognition in the current frame, a detection failure hold mechanism is added: if the original value of the current frame is... If the detection fails and is unavailable or forcibly set to zero, the smoothing result of the previous frame will be used. No update is performed; the smoothing formula described above is only applied when the detection is normal. This protection ensures that the transfer coefficient remains continuous during brief occlusions or detection failures, better reflecting the actual risk situation.

[0125] (5) Chain accumulation of transmission coefficients Based on the chain-like relationship of the vehicle following queue and combined with a dynamic transmission coefficient, the impact is propagated layer by layer backward. To reduce computational complexity and simplify the process, this embodiment only considers a maximum of 7 upstream vehicles in front of the vehicle, with no more than 3 vehicles in the first layer, and any excess vehicles are ignored. The selection of 7 vehicles is based on actual perception constraints: when the vehicle is in the middle lane of a three-lane road, it can usually only reliably detect the vehicle immediately in front in the same lane. Therefore, 3 vehicles are selected from each of the left and right adjacent lanes, totaling 7 vehicles, covering the range of upstream vehicles that pose the main risk to the vehicle in dense following scenarios.

[0126] Based on determining the range of upstream vehicles, it is also necessary to quantify the actual impact of each preceding vehicle on the following vehicle. Therefore, a distance-related attenuation coefficient is introduced to characterize the influence of distance on the propagation intensity, as shown in the following formula:

[0127] in The attenuation coefficient characterizes the gradual decay of risk as it propagates within the queue; This is the distance attenuation coefficient. This represents the actual longitudinal distance between the two vehicles. This attenuation is used to simulate the step-by-step dissipation of risk during multi-hop propagation, and... The direct response attenuation differs, and both affect different calculation stages. The smaller the vehicle distance, the closer the transmission coefficient is to 1, and the stronger the risk impact of the preceding vehicle on the following vehicle. As the distance increases, the transmission coefficient decays exponentially, and the risk impact from distant vehicles gradually weakens, consistent with the law of road traffic risk propagation. In this embodiment... This value results in a single-stage transmission factor of approximately 0.6 at a spacing of 10m, decreasing to 0.37 at 20m, and below 0.23 at 30m, which is consistent with the law that the influence of upstream vehicles in traffic flow decreases exponentially with distance.

[0128] For vehicles in the same lane and adjacent lanes, upstream vehicles can influence the same downstream vehicle through multiple different paths. This embodiment's method starts with the upstream vehicle in each lane and recursively calculates the cumulative transmission value vehicle by vehicle in a direction from farthest to nearest. Specifically: (a) Recursive calculation: For the target vehicle to be calculated Its cumulative transmission coefficient The recursive formula is obtained by linearly summing the effects of all its front-wheel-drive vehicles, as follows:

[0129] in: For all those that can directly affect vehicles The upstream vehicle collection; The attenuation coefficient; For vehicles right The direct transmission coefficient, including the braking transmission coefficient. and steering transmission coefficient .

[0130] (b) Recursive order and initial conditions: Calculate vehicle by vehicle in order of longitudinal distance from the target vehicle, from farthest to closest. For the upstream vehicle without a front-wheel drive vehicle, its cumulative transfer coefficient... .

[0131] Based on the above recursive rules and the transfer conditions of adjacent lanes, the method in this embodiment calculates and accumulates the transfer coefficient for each vehicle in each lane from upstream to downstream, thereby obtaining the cumulative transfer coefficient of each vehicle in each lane.

[0132] (6) Output results The method in this embodiment outputs the following three results, which correspond to overall risk quantification, braking prediction of the vehicle immediately preceding it, and risk source location, respectively.

[0133] 6.1) Cumulative transmission coefficient of the vehicle

[0134]

[0135] The value is directly passed and marked in the upper right corner of the image. This represents the total cumulative impact of all vehicle light signals upstream on the vehicle after being transmitted in a chain. This coefficient is used for threshold judgment in the driving warning system.

[0136] 6.2) First-level vehicle braking probability make For the car that is right next to the car in front of it, This is the cumulative transmission coefficient experienced by the vehicle. Mapped to braking risk coefficient via piecewise linear function :

[0137]

[0138] in, right A threshold of 100 is set. A higher value indicates a greater probability that the vehicle immediately in front will brake in the near future. It is directly marked in the upper left corner of the detection frame immediately adjacent to the vehicle in front. Additionally, if the brake light indicator of the vehicle in front is detected in the current frame... Then directly order It no longer relies on cumulative transfer coefficient mapping. This output can provide a prediction immediately before the brake lights of the vehicle in front illuminate, giving the vehicle more reaction time.

[0139] 6.3) Tracing the source of greatest risk For the first level vehicle All its upstream vehicles right The candidate values ​​for the intensity of the influence are given by the following formula:

[0140] in The attenuation coefficient is... For vehicles arrive The value is passed directly. The greatest source of risk is defined as making Largest upstream vehicle At the same time, The direct transmission value to the vehicle itself is used as its own candidate value. All candidate values ​​are compared, and the source vehicle corresponding to the maximum value is taken as the output. If the vehicle's own candidate value is the largest, it is determined to be the source of risk, meaning there is no upstream dominant influence. If the vehicle's own candidate value and the upstream vehicle's candidate value are tied for the largest, both are marked as the largest risk source.

[0141] The judgment result will be determined in the first layer of vehicles. The detection frames are labeled: During the preprocessing stage, each detected vehicle is assigned a unique number, which is then labeled on the corresponding vehicle's detection frame. After the maximum risk source is determined, the determination result is attached to the first layer of vehicles based on the risk source vehicle number. The detection boxes are marked. If the greatest risk source is the vehicles on the second layer, then... The detection box is marked with the number of the second-layer vehicle; if the greatest risk source is the vehicle itself, then the vehicle number is marked; if it is a joint greatest risk source, then the numbers of multiple source vehicles are marked at the same time.

[0142] In summary, compared with the prior art, this application has at least the following advantages and beneficial effects: (a) Low deployment cost. No need to retrain the model or add V2X communication equipment. It can be directly embedded into the existing perception system as a post-processing module. The development cost is low, the integration is convenient, and the original detection model is not changed.

[0143] (b) It has a certain ability to predict in advance. By analyzing the direct and indirect impact layers of upstream traffic flow in a layered manner, it is possible to predict the deceleration or lane change trend of the vehicle in front before its own brake lights are turned on, based on the cumulative effect of the upstream light signals, thereby increasing the reaction time of the vehicle and helping to reduce the risk of rear-end collisions.

[0144] (c) Braking and steering are calculated and fused independently in dual channels. Combined with the exponential decay related to lateral overlap and distance, it can distinguish different lane-changing scenarios such as cutting in from a distance and cutting in from close range, as well as the risk of following another vehicle on a straight road. At the same time, the processing rules for hazard light signals are unified, reducing logical ambiguity.

[0145] (d) The temporal smoothing mechanism can suppress single-frame detection noise; the dynamic distance threshold adapts to the longitudinal distance to adapt to the scale effect of perspective projection; the maximum risk source is determined based on the actual transmission coefficient, so that the risk source identification is directly related to the actual impact intensity.

[0146] (e) Vehicle-by-vehicle recursive accumulation avoids path enumeration, only considers a limited number of upstream vehicles, has high computational efficiency, can meet real-time requirements, and is easy to deploy on mass-produced models.

[0147] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0148] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0149] Please see Figure 3 , Figure 3 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 301 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 302 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 302 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 302 and is called and executed by the processor 301 using the methods described above in the embodiments of this application. Input / output interface 303 is used to implement information input and output; The communication interface 304 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 305 transmits information between various components of the device (e.g., processor 301, memory 302, input / output interface 303, and communication interface 304); The processor 301, memory 302, input / output interface 303, and communication interface 304 are connected to each other within the device via bus 305.

[0150] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0151] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0152] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0153] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0154] It is understood that the content of the above method embodiments is applicable to the embodiments of this program product. The specific functions implemented in the embodiments of this program product are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments. The executable computer program code or "code" used to perform the various embodiments can be written in high-level programming languages ​​such as C, C++, Python, Smalltalk, Java, JavaScript, Visual Basic, Structured Query Language (e.g., Transact-SQL), Perl, or in various other programming languages.

[0155] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0156] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0157] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0158] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0159] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0160] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0161] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0162] The units described above 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.

[0163] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0164] If the integrated unit is implemented as a software functional unit 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 all 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 multiple 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 of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0165] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for calculating the cumulative transmission coefficient based on multi-vehicle lighting signals, characterized in that, The method includes the following steps: Obtain vehicle detection box information and light status information output by the target detection model, determine the lane to which each vehicle belongs and the longitudinal distance between vehicles based on the detection box information, and assign a stable number to each vehicle. Based on the relative position relationship of vehicles, a one-way hierarchical transmission rule is set. Starting from the upstream vehicle, effective vehicle pairs are searched layer by layer downstream. All effective transmission paths with the self-vehicle as the endpoint are constructed. The upstream vehicles are divided into first-layer vehicles that have a direct transmission relationship with the self-vehicle and second-layer vehicles that only indirectly affect the self-vehicle through the first-layer vehicles. The braking channel transmission coefficient and the steering channel transmission coefficient are calculated separately, and the braking channel transmission coefficient and the steering channel transmission coefficient are weighted and fused to obtain the direct transmission coefficient between the vehicle pairs; wherein, the braking channel transmission coefficient is calculated based on the brake light status of the preceding vehicle and the longitudinal distance between the two vehicles, and the steering channel transmission coefficient is calculated based on the turn light status of the preceding vehicle, the relative lane positions of the two vehicles, the light status of the following vehicle, the lateral distance between the two vehicles, and the lateral overlap. The direct transfer coefficients of multiple consecutive frames are subjected to time-weighted smoothing to suppress single-frame detection noise; The cumulative transmission coefficient of each upstream vehicle is calculated recursively along the transmission direction of the lane, and an attenuation coefficient based on longitudinal distance is introduced in the recursive process. Based on the cumulative transmission coefficient, output the vehicle's cumulative transmission coefficient, the braking probability of the adjacent vehicle in front, and the tracing result of the maximum risk source.

2. The method according to claim 1, characterized in that, The step of determining the lane to which each vehicle belongs based on the detection frame information includes: The horizontal coordinate of the vanishing point is fixed at half the width of the image, and the candidate vanishing points are traversed within the corresponding vertical coordinate interval with a fixed step size. For each candidate vanishing point, generate a set of rays originating from that candidate vanishing point and covering the road surface area; Calculate the vertical distance from each vehicle's grounding point to each ray, and match the vehicle to the ray with the smallest distance. If the distance is less than the dynamic distance threshold, the vehicle is determined to belong to the corresponding ray. Count the number of valid rays with matching vehicles under each candidate vanishing point, and merge adjacent valid rays with an angle difference less than the merging threshold into the same lane. The number of merged rays is the number of lanes corresponding to the candidate vanishing point. The candidate vanishing point with the fewest rays after merging is selected as the optimal vanishing point, and the vehicle number and ray direction corresponding to each lane are output accordingly.

3. The method according to claim 1, characterized in that, The longitudinal distance between the vehicles is calculated in the following way: in, The longitudinal distance from the vehicle to the onboard camera. For the camera's equivalent focal length, The height of the camera's optical center from the ground. The vertical coordinate of the center point of the vehicle's bottom is shown in the image. The vertical coordinate of the vanishing point image. The minimum constant to prevent numerical overflow; longitudinal spacing between front and rear vehicles It is determined by the absolute value of the difference in the longitudinal distance between the two vehicles.

4. The method according to claim 1, characterized in that, The unidirectional hierarchical transmission rule includes: Within the same lane, only adjacent vehicles in front and behind are allowed to transmit influence directly; vehicles in between are not allowed to be bypassed. Only longitudinally adjacent vehicles with no intermediate vehicles are allowed to directly transfer influence between adjacent lanes; transmission across multiple lanes is prohibited. The transmission direction is along the traffic flow direction, and it is transmitted only from the vehicle in front to the vehicle behind.

5. The method according to claim 1, characterized in that, The braking channel transmission coefficient The calculation formula is: in, , This indicates the status of the brake lights of the vehicle in front; 1 represents the indicator when the lights are on and 0 represents the indicator when they are off. This is the dual-flash enhancement factor; The initial value of the transmission coefficient based on the longitudinal distance is calculated using the following formula: in, The longitudinal distance between the front and rear vehicles. The longitudinal half-effect distance; The braking channel transfer coefficient is only effective between vehicles in the same lane and longitudinally adjacent; when the hazard lights of the vehicle in front are detected to be on, the braking channel transfer coefficient is not limited to the same lane, and the vehicle in front is allowed to calculate the transfer coefficient for the vehicle in the same lane and the vehicle in the adjacent lane that is longitudinally adjacent to it.

6. The method according to claim 1, characterized in that, The steering channel transmission coefficient The calculation formula is: in, This is the interaction coefficient, and its value is determined based on the lane position of the vehicle in front relative to the vehicle behind and the light status of the vehicle behind. The lateral overlap between the vehicle in front and the vehicle behind; The lateral distance between the two vehicles; This is the lateral half-effect distance; When the vehicle in front has not turned on the turn signal on one side, When the vehicle in front turns on its hazard lights, .

7. The method according to claim 1, characterized in that, The calculation formula for the time-weighted smoothing process is as follows: in, For the first Direct transfer coefficients after frame smoothing For the first Direct transfer coefficients after frame smoothing The original calculated value for the current frame. Historical retention factor; If the original calculated value becomes unavailable due to detection failure in the current frame, then retain... .

8. The method according to claim 1, characterized in that, The formula for calculating the cumulative transmission coefficient of each upstream vehicle is as follows: in, For vehicles The cumulative transfer coefficient, For all those that can directly affect vehicles upstream vehicle collection, For vehicles right The direct transmission coefficient, For vehicles and The attenuation coefficient between them; The recursion is performed in order of longitudinal distance between vehicles, from farthest to nearest, with the cumulative transfer coefficient of the upstream vehicle being the highest value. .

9. The method according to claim 1, characterized in that, The vehicle cumulative transmission coefficient The calculation formula is: in, This refers to the set of all upstream vehicles that can directly affect the vehicle itself. For vehicles The cumulative transfer coefficient, For vehicles The direct transmission coefficient to the vehicle. For vehicles The attenuation coefficient between the vehicle and the vehicle itself.

10. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method according to any one of claims 1 to 9.

Images