Road marking drawing method and system based on image recognition

Abstract

The invention relates to the technical field of road marking, in particular to a road marking drawing method and system based on image recognition. The method comprises the following steps: acquiring a historical track data set, a road design parameter set and an environment image set of a target road; analyzing the historical track data set, and determining track repetition; determining geometric constraint conditions according to the road design parameter set; according to the environment image set, dangerous source distribution characteristics are determined; and determining a marking scheme according to the track repetition degree, the geometric constraint condition and the hazard source distribution characteristics. The method is helpful for increasing the visual distance of the marking under extreme conditions, reducing out-of-control accidents caused by overspeed, reducing roadside hazard source danger avoiding space contradictions, generating a marking scheme, preventing compensation intensity from exceeding physical limitation of a lane, and reducing the decision error rate at a marking abrupt change position.

Description

Technical Field

[0001] The present application relates to the technical field of road markings, and in particular to a method and system for drawing road markings based on image recognition. Background Art

[0002] Road markings are a core element of traffic guidance systems, and their design rationality directly impacts driving safety and traffic efficiency. Traditional methods for marking mountain roads, which include continuous curves, steep slopes, and high-risk roadside environments, are often used.

[0003] However, the accident-inducing rate of existing road marking schemes is higher than that of ordinary road sections. Therefore, how to solve the significant technical defects of the current road marking drawing method in the application of complex terrain sections has become a core issue. Summary of the Invention

[0004] The present application provides a method and system for drawing road markings based on image recognition to solve the above problems.

[0005] In a first aspect, the present application provides a road marking drawing method based on image recognition, the method comprising: obtaining a historical trajectory data set, a road design parameter set, and an environmental image set of a target road; analyzing the historical trajectory data set to determine trajectory repeatability; determining geometric constraints based on the road design parameter set; determining hazard source distribution characteristics based on the environmental image set; and determining a road marking scheme based on the trajectory repeatability, the geometric constraints, and the hazard source distribution characteristics.

[0006] This solution extracts historical vehicle trajectory datasets, which helps improve trajectory classification accuracy. Extracting road design parameter sets helps improve the separation of trajectory features related to vehicle types. Extracting environmental image sets helps provide data support for hazard source distribution characteristics, compensating for the shortcomings of traditional static identification. Analyzing historical trajectory datasets helps eliminate GPS drift and abnormal driving behavior data, while effectively reducing trajectory range extraction errors. Calculating trajectory repeatability helps improve the effectiveness of road marking warnings and reduce the critical safety redundancy caused by trajectory competition between small and articulated vehicles. Generating geometric constraints helps reduce skidding accident rates and misjudgment rates. Determining hazard source distribution characteristics helps improve the visible distance of road markings under extreme conditions, reduce uncontrolled accidents caused by speeding, and reduce conflicts in roadside hazard avoidance space. Generating road marking solutions helps avoid compensation intensity exceeding the physical limitations of the lane and reduces decision-making errors at sudden road marking points.

[0007] Optionally, analyzing the historical trajectory data set to determine trajectory repetition includes: analyzing the historical trajectory data set to determine the type of traveling vehicle and the driving trajectory set of the corresponding type; analyzing the driving trajectory set for each type of traveling vehicle to determine the spatial deviation distribution of the corresponding driving vehicle type; establishing a spatiotemporal four-dimensional filtering model, and based on the spatiotemporal four-dimensional filtering model, eliminating abnormal data according to the spatial deviation distribution, and determining the trajectory range of the driving trajectory set after elimination according to the spatial deviation distribution; and determining the trajectory repetition based on the trajectory range.

[0008] Optionally, the types of traveling vehicles include articulated vehicles and small vehicles, and determining the trajectory range of the driving trajectory set after elimination based on the spatial deviation distribution includes: analyzing the driving trajectory set of the articulated vehicle and the driving trajectory set of the small vehicle for the articulated vehicle to obtain a first vehicle analysis result of the articulated vehicle and a second vehicle analysis result of the small vehicle; determining a turning envelope of each articulated vehicle based on the first vehicle analysis result; determining a turning boundary of the articulated vehicle based on the turning envelope; and determining the trajectory range of the driving trajectory set after elimination based on the turning boundary and the second vehicle analysis result and the spatial deviation distribution.

[0009] Optionally, determining the geometric constraints based on the road design parameter set includes: parsing the road design parameter set to obtain the curve radius, longitudinal slope, and superelevation transverse slope; analyzing the turning envelope and the spatial deviation distribution of small vehicles to determine the dynamic width requirement; correcting the dynamic width requirement based on the coupling relationship between the longitudinal slope and the superelevation transverse slope to obtain the actual dynamic width requirement; and determining the actual dynamic width requirement as the geometric constraints.

[0010] Optionally, determining the hazard source distribution characteristics based on the environmental image set includes: analyzing the environmental image set to determine road information and meteorological information; determining the distance and height difference of roadside cliffs based on the road information; determining the average annual frequency of foggy days and the duration of a single fog based on the meteorological information; determining the fog type based on the meteorological information; determining the driving impact of each type of fog based on the fog type and the meteorological information; determining high-risk area characteristics based on the distance and height difference of the roadside cliffs; determining visibility hazard characteristics based on the average annual frequency of foggy days, the driving impact and the duration of a single fog; and determining the high-risk area characteristics and the visibility hazard characteristics as hazard source distribution characteristics.

[0011] Optionally, determining the turning envelope of each articulated vehicle based on the first vehicle analysis result includes: determining the rear axle offset data of the articulated vehicle based on the first vehicle analysis result; predicting the vehicle load status based on the rear axle offset data; determining the dynamic envelope range based on the vehicle load status and the rear axle offset data; and correcting the dynamic envelope range based on the coupling relationship between the rear axle offset data and the super-high cross slope to obtain the turning envelope of each articulated vehicle.

[0012] Optionally, the line marking scheme is determined according to the trajectory repeatability, the geometric constraints and the hazard source distribution characteristics, including: determining a number of line marking positions according to the trajectory range; determining the line marking confidence of each line marking position according to the trajectory repeatability; determining the minimum allowable width according to the geometric constraints; determining the compensation intensity according to the hazard source distribution characteristics; determining the optimal line marking position and compensation scheme according to the line marking confidence of each line marking position, the minimum allowable width and the compensation intensity, to obtain a line marking scheme.

[0013] Optionally, after determining the optimal line marking position and compensation scheme based on the line marking confidence of each line marking position, the minimum allowable width and the compensation strength, and obtaining the line marking scheme, the method further includes: determining a vehicle dynamics constraint set based on the historical trajectory data set; constructing a driving psychological expectation model based on the historical trajectory data set; analyzing the coupling relationship between the vehicle dynamics constraint set and the driving psychological expectation model, dynamically compensating and correcting the line marking scheme, and generating a final line marking drawing scheme.

[0014] In a second aspect, the present application provides a road marking drawing system based on image recognition, the system comprising: A data acquisition module is used to obtain a historical trajectory dataset, a road design parameter set, and an environmental image set of the target road; A data analysis module, configured to analyze the historical trajectory data set and determine trajectory repetitiveness; a parameter analysis module, configured to determine geometric constraints based on the road design parameter set; An image analysis module, configured to determine the distribution characteristics of hazard sources based on the environmental image set; The scheme determination module is used to determine the marking scheme according to the trajectory repeatability, the geometric constraints and the hazard source distribution characteristics.

[0015] In a third aspect, the present application provides an electronic device, comprising: a memory and a processor; The memory is used to store program instructions; The processor is used to call and execute program instructions in the memory to perform the method as described in any one of the first aspects. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0017] Figure 1 A schematic diagram of an application scenario provided in one embodiment of the present application; Figure 2 A flowchart of a method for drawing road markings based on image recognition provided in one embodiment of the present application; Figure 3 A flowchart of another road marking system based on image recognition provided in one embodiment of the present application; Figure 4 A schematic structural diagram of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0018] To make the purpose, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0019] In this document, the term "and / or" simply describes a relationship between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document, unless otherwise specified, generally indicates an "or" relationship between the related objects.

[0020] The embodiments of the present application are described in further detail below with reference to the accompanying drawings.

[0021] The accident-inducing rate of existing road marking schemes is higher than that of ordinary road sections. Therefore, how to solve the significant technical defects of the current road marking drawing method in the application of complex terrain sections has become a core issue.

[0022] Based on this, the present application provides a road marking drawing method and system based on image recognition, which obtains a historical trajectory dataset, a road design parameter set, and an environmental image set for the target road; analyzes the historical trajectory dataset to determine trajectory repetition; determines geometric constraints based on the road design parameter set; determines hazard source distribution characteristics based on the environmental image set; and determines a road marking scheme based on trajectory repetition, geometric constraints, and hazard source distribution characteristics. Extracting historical vehicle trajectory datasets helps improve trajectory classification accuracy. Extracting road design parameter sets helps improve the separation of vehicle type-related trajectory characteristics. Extracting environmental image sets helps provide data support for hazard source distribution characteristics, compensating for the shortcomings of traditional static recognition. Analyzing historical trajectory datasets helps eliminate GPS drift and abnormal driving behavior data, while effectively reducing trajectory range extraction errors. Calculating trajectory repetition helps improve the warning effect of road markings and reduce the critical safety redundancy caused by trajectory competition between small vehicles and articulated vehicles. Generating geometric constraints helps reduce the rate of skidding accidents and misjudgment rates. Determining the distribution characteristics of hazard sources helps improve road marking visibility under extreme conditions, reduce speeding-related accidents, and alleviate spatial conflicts associated with roadside hazard avoidance. Generating road marking solutions helps prevent compensation strength from exceeding the physical limitations of lanes and reduces decision-making errors at sudden changes in markings.

[0023] Figure 1 This is a schematic diagram of an application scenario provided by this application. When marking roads in mountainous areas, the method provided by this application is applied. Specifically, the method provided by this application is applied to any server, and the server interacts with the vehicle-mounted GPS, roadside radar, video surveillance equipment, database, and drones in mountainous areas to obtain historical trajectory data sets collected by the vehicle-mounted GPS, roadside radar, and video surveillance equipment, road design parameter sets retrieved from the database, and environmental image sets collected from drone aerial photography. The above historical trajectory data sets, road design parameter sets, and environmental image sets are carefully analyzed, and a marking scheme is determined based on the results of the analysis to avoid compensation strength exceeding the physical limitations of the lane and reduce the error rate in decision-making at sudden changes in the marking. For specific implementation methods, please refer to the following embodiments.

[0024] Figure 2 This is a flow chart of a method for drawing road markings based on image recognition provided by an embodiment of the present application. The method of this embodiment can be applied to the server in the above scenario. Figure 2 As shown, the method includes: S201, obtaining a historical trajectory dataset, a road design parameter set, and an environmental image set of a target road; The target road can be a mountain road or a complex terrain section with continuous curves, steep slopes and high-risk roadside environments.

[0025] The historical trajectory data set may be a set of vehicle driving trajectory data over a period of time in the past. This period of time may be selected based on experience or determined by assumptions.

[0026] The road design parameter set may be a set of digital parameters of road geometric features.

[0027] The environment image set may be a road environment multimodal image dataset.

[0028] Specifically, a historical vehicle trajectory dataset for the target road is collected using data acquisition equipment such as on-board GPS, roadside radar, and video surveillance equipment. Road design parameters such as curve radius, longitudinal slope, and superelevation transverse slope are extracted from road design files in a database. Multi-view road images are acquired through drone aerial photography. A semantic segmentation model is constructed based on these multi-view road images. This semantic segmentation model is then used to extract an environmental image set that includes roadside cliff boundaries, guardrail locations, and meteorological characteristics.

[0029] S202, analyzing the historical trajectory data set to determine the trajectory repetitiveness; Trajectory repetitiveness can be an indicator to quantify the degree of spatial overlap of vehicle trajectories.

[0030] Specifically, cluster analysis was performed on the historical trajectory dataset to distinguish between clusters of articulated and small vehicles. Turning phase characteristics and the vehicle's straight-line retention rate were then extracted from these trajectories. A three-dimensional coordinate system was established based on the road centerline, and timestamps were superimposed to form a spatiotemporal four-dimensional filtering model. A dynamic filtering threshold was set based on the vehicle type, and a Kalman filter was used to smooth the trajectories, removing outliers that exceeded the dynamic filtering threshold. Spatial overlay analysis was then performed on the filtered historical trajectory dataset to calculate the trajectory reproducibility for each location.

[0031] S203. Determine geometric constraints based on the road design parameter set; Geometric constraints can be the physical limitations imposed by road alignment on vehicle movement.

[0032] Specifically, the envelope width is corrected based on the turning radius and superelevated cross slope. The standard deviation of the spatial deviation distribution of the historical trajectory dataset is extracted to calculate the dynamic width requirement. A correlation function is established between the longitudinal slope and the dynamic width requirement. Based on this correlation function, geometric constraints are generated for each road segment.

[0033] S204. Determine the distribution characteristics of the hazard sources based on the environmental image set; The distribution characteristics of hazardous sources can be quantitative parameters of the spatial distribution laws of hazardous sources.

[0034] Specifically, the distance from the cliff edge to the lane line is obtained from a collection of environmental images. This is combined with a LiDAR point cloud to generate a digital elevation model and calculate the height difference. Radiation fog and advection fog are determined based on the image color temperature. A visibility attenuation model is then established based on these two factors. The duration of a single fog episode and the average annual frequency of foggy days are collected through meteorological monitoring. The visibility attenuation model and the duration of a single fog episode are calculated to determine the driving impact coefficient. The calculated cliff distance and height difference are used to identify high-risk area characteristics. Based on these high-risk area characteristics, the driving impact coefficient and the average annual frequency of foggy days are calculated to determine the distribution characteristics of hazard sources.

[0035] S205. Determine the marking scheme based on trajectory repeatability, geometric constraints, and hazard source distribution characteristics.

[0036] The marking scheme may be a standard scheme for implementing traffic markings.

[0037] Specifically, based on the offset between the vertex of the convex hull of the trajectory range and the road centerline, a set of candidate positions is generated according to the average distance interval. The trajectory repeatability is used to determine the road marking confidence. Based on the geometric constraints, the minimum allowable width is determined by the larger value of the road's physical boundary and the dynamic width requirement. The compensation strength is determined by normalizing the dimensionless indicator based on the quantitative value of safety resource investment calculated based on the distribution characteristics of the hazard source. Based on the calculated road marking confidence, minimum allowable width, and compensation strength, an objective function is established to maximize the practicality and adaptability of the road marking while satisfying safety and geometric constraints. The road marking position and parameter combination that best balances the road marking confidence, minimum allowable width, and compensation strength is selected under the premise of satisfying safety, geometric constraints, and cost constraints in the objective function to generate the final road marking solution.

[0038] This solution extracts historical vehicle trajectory datasets, which helps improve trajectory classification accuracy. Extracting road design parameter sets helps improve the separation of trajectory features related to vehicle types. Extracting environmental image sets helps provide data support for hazard source distribution characteristics, compensating for the shortcomings of traditional static identification. Analyzing historical trajectory datasets helps eliminate GPS drift and abnormal driving behavior data, while effectively reducing trajectory range extraction errors. Calculating trajectory repeatability helps improve the effectiveness of road marking warnings and reduce the critical safety redundancy caused by trajectory competition between small and articulated vehicles. Generating geometric constraints helps reduce skidding accident rates and misjudgment rates. Determining hazard source distribution characteristics helps improve the visible distance of road markings under extreme conditions, reduce uncontrolled accidents caused by speeding, and reduce conflicts in roadside hazard avoidance space. Generating road marking solutions helps avoid compensation intensity exceeding the physical limitations of the lane and reduces decision-making errors at sudden road marking points.

[0039] In some embodiments, a historical trajectory data set is analyzed to determine the type of traveling vehicle and the corresponding type of driving trajectory set; for each type of traveling vehicle, the driving trajectory set is analyzed to determine the spatial deviation distribution of the corresponding traveling vehicle type; a spatiotemporal four-dimensional filtering model is established, and based on the spatiotemporal four-dimensional filtering model, abnormal data is eliminated according to the spatial deviation distribution, and the trajectory range of the driving trajectory set after elimination is determined according to the spatial deviation distribution; and the trajectory repeatability is determined based on the trajectory range.

[0040] The type of traveling vehicle can be a small vehicle, an articulated vehicle, or other vehicle types classified based on the physical parameters of the vehicle.

[0041] The corresponding type may be the vehicle type matched by the classification rules.

[0042] The driving trajectory set may be a set of historical trajectory data generated by the same type of driving vehicle on the target road.

[0043] The corresponding traveling vehicle type may be a vehicle type corresponding to the trajectory dataset.

[0044] The spatial deviation distribution may be a statistical distribution of the lateral offsets of the vehicle trajectory set relative to the road centerline.

[0045] The spatiotemporal four-dimensional filtering model can be a trajectory cleaning model that integrates timestamps, longitude, latitude, and elevation.

[0046] Abnormal data can be trajectory data that does not meet geometric constraints.

[0047] The trajectory range may be a spatial boundary formed by valid trajectory points retained after filtering.

[0048] Specifically, the system uses historical trajectory datasets to extract vehicle characteristic parameters such as vehicle length, wheelbase, and maximum steering angle for each trajectory. Based on these vehicle characteristic parameters, a vehicle classification rule base is established: vehicles with a larger wheelbase and the presence of articulation-section signature signals are articulated vehicles; vehicles with a smaller wheelbase and a larger maximum steering angle are small vehicles. The classification results are then correlated with the corresponding historical trajectory dataset to generate a trajectory set. Using the road centerline as the baseline, a local coordinate system is established, and the temporal and spatial coordinates of the trajectories of the same vehicle type are normalized. The road cross-sections are divided into average distance intervals, and the lateral offset distribution of the trajectory points corresponding to the vehicle type within each section is calculated, i.e., the spatial deviation distribution. Based on the statistical distribution theory of spatiotemporal data, a spatiotemporal four-dimensional filtering model is established by integrating time series analysis (such as Kalman filtering) with spatial clustering methods (such as DBSCAN). Dynamic thresholding (such as the 3σ criterion) and road geometry (such as curve radius) are used to adjust the filter strength to identify and eliminate anomalous trajectory data. A dynamic threshold is then set based on the spatial deviation distribution. Trajectory points exceeding the dynamic threshold are marked as anomalous data. If a series of anomalous points are present, the entire trajectory is eliminated. Furthermore, enhanced filtering is applied to sharp curves, reducing the dynamic threshold to generate a trajectory range. Convex hull analysis is performed on the filtered trajectory range to generate a spatial envelope covering the trajectory points. The curvature continuity of the spatial envelope is optimized based on the road longitudinal slope and superelevation transverse slope. Based on the optimized spatial envelope, the critical safety distance between the spatial envelope and the road boundary is calculated. The road centerline width is expanded based on the spatial deviation distribution. The overlap complexity of the trajectories of each vehicle type within the expanded baseline is then calculated.

[0049] This solution generates a set of driving trajectories, helping to quantify the differences in driving characteristics for each vehicle type, eliminate misjudgment of articulated vehicle trajectories, and improve the effectiveness of curve marking warnings. Statistical analysis of spatial deviation distribution helps reduce the rate of side-slip accidents. Establishing a spatiotemporal four-dimensional filtering model helps filter out abnormal data in inclement weather, addressing trajectory distortion during peak hours and inclement weather, reducing trajectory point drift caused by visibility degradation, shortening driver trajectory correction delays, and effectively reducing the rate of speeding and loss of control accidents. Determining the overlap complexity of trajectories helps increase the continuity of visual guidance in sections where markings suddenly change, reducing brake response delays.

[0050] In some embodiments, for articulated vehicles, the driving trajectory sets of articulated vehicles and the driving trajectory sets of small vehicles are analyzed to obtain a first vehicle analysis result for the articulated vehicles and a second vehicle analysis result for the small vehicles; based on the first vehicle analysis result, the turning envelope of each articulated vehicle is determined; based on the turning envelope, the turning boundary of the articulated vehicle is determined; based on the turning boundary, the trajectory range of the driving trajectory set after elimination is determined according to the second vehicle analysis result and the spatial deviation distribution.

[0051] An articulated vehicle may be a multi-axle transport vehicle having an articulated structure.

[0052] The first vehicle analysis result may be an analysis result of an articulated vehicle driving trajectory set.

[0053] The second vehicle analysis result may be an analysis result of a small vehicle driving trajectory set.

[0054] The turning envelope may be a dynamic spatial boundary generated based on the first vehicle analysis results of the articulated vehicle.

[0055] The turning boundary can be the final safety boundary generated by integrating the safety redundancy compensation mechanism, super-high cross slope, and the overlapping area of ​​the trajectories of articulated vehicles and small vehicles on the basis of the turning envelope.

[0056] Specifically, the maximum lateral offset, turning radius extreme point, and rear axle offset dynamic parameters of the articulated vehicle's driving trajectory set are extracted; thus, a trajectory curvature change map of the articulated vehicle is constructed, and the sudden change characteristics of the trailer trajectory at continuous curves are marked, which is the first vehicle analysis result. The average lateral offset, speed change gradient, and braking point distribution density of the small vehicle's driving trajectory set are statistically analyzed; based on vehicle kinematics theory (such as the inverse relationship between speed and curvature), the vehicle's ,in is the friction coefficient, =turn radius) and driving behavior statistics (using historical trajectory data to fit an empirical curvature-speed regression equation), the trajectory deformation characteristics caused by vehicle speed adjustments in curves are quantified. A trajectory curvature-speed correlation model for small vehicles is established, and the trajectory compression characteristics in curves are quantified, which is the result of the second vehicle analysis. Based on the second vehicle analysis results, trajectory points with the highest lateral offset values ​​in the articulated vehicle's driving trajectory are selected as envelope reference points. The spatial distribution weights of the envelope reference points are dynamically adjusted based on the load status. Several envelope reference points are then connected to generate an initial convex hull boundary. Based on the longitudinal slope, the initial convex hull boundary is longitudinally stretched and corrected. The dynamic parameters of the rear axle offset of the articulated vehicle are then superimposed to generate the turning envelope for each articulated vehicle. The lateral distance between the turning envelope and the road boundary is calculated. If the distance is too small, a safety redundancy compensation mechanism is triggered. The safety redundancy compensation mechanism is adjusted based on the superelevated cross slope to generate a slope-dependent turning boundary. The turning boundary of the articulated vehicle is superimposed on the analysis results of the second vehicle to generate an initial trajectory range. The initial trajectory range is then subjected to probability-weighted smoothing according to the probability density function of the spatial deviation distribution to determine the trajectory range of the final driving trajectory set.

[0057] This solution determines the first vehicle analysis results for articulated vehicles and the second vehicle analysis results for small vehicles, helping to provide a dynamic parameter benchmark for turning envelope generation, reducing the misjudgment rate of trailer trajectories at curves and improving the accuracy of marking warning matching. Generating a turning envelope helps avoid the problem of insufficient critical safety distance between the driving trajectories of articulated vehicles and small vehicles. Generating turning boundaries helps reduce the rate of side-slip accidents and the rate of decision-making errors caused by sudden changes in markings. It also eliminates trajectory deviations caused by excessive cross slopes and improves the continuity of markings at curves. Determining the trajectory range helps trigger the driver's trajectory correction in advance, reducing the rate of loss of control accidents caused by speeding.

[0058] In some embodiments, the road design parameter set is analyzed to obtain the curve radius, longitudinal slope, and superelevation transverse slope; the turning envelope and the spatial deviation distribution of small vehicles are analyzed to determine the dynamic width requirement; based on the coupling relationship between the longitudinal slope and the superelevation transverse slope, the dynamic width requirement is corrected to obtain the actual dynamic width requirement; and the actual dynamic width requirement is determined as a geometric constraint.

[0059] The curve radius may be the radius of curvature of the centerline of the road in the horizontal plane.

[0060] The longitudinal slope can be the longitudinal slope value of the road.

[0061] The superelevation cross slope can be the transverse inclination of the road surface at the curve.

[0062] The dynamic width requirement may be an initial value for the minimum lane width.

[0063] The coupling relationship can be the interaction mechanism between the longitudinal slope and the superelevation cross slope.

[0064] The actual dynamic width requirement may be the corrected final lane width value.

[0065] Specifically, the original road design parameter set is obtained, and three core geometric parameters, namely, curve radius, longitudinal slope, and superelevation transverse slope, are extracted. Based on the turning envelope of articulated vehicles, the maximum lateral width of the turning envelope is calculated. The driving trajectory set of small vehicles is spatially clustered, and driving trajectory points with spatially deviated distributions are screened to calculate the width of the small vehicle trajectory density area. The dynamic width requirement is calculated by combining the maximum lateral width, the trajectory density area width, and the safety redundancy factor. A compensation coefficient formula is established for the coupling relationship between longitudinal slope and superelevation transverse slope. When the compensation coefficient is too high, the slope-related compensation mode is triggered. The dynamic width requirement is multiplied by the compensation coefficient to generate a corrected dynamic width requirement. An additional longitudinal stretch compensation is added to downhill sections. The lateral spacing between the corrected dynamic width requirement and the road boundary is verified. If the lateral spacing is insufficient, the boundary is expanded by the difference, thus generating the final actual dynamic width requirement. The actual dynamic width requirement is bound to the curve radius, longitudinal slope, and superelevation slope to form a dynamic geometric constraint tuple containing the curve radius, longitudinal slope, superelevation slope, and actual width. According to road design specifications, the actual width is converted into marking design parameters: lane spacing is calculated based on the actual width. The edge marking offset is also calculated based on the actual width and safety redundancy factor. Geometric constraints are generated based on the lane spacing and edge marking offset.

[0066] This solution extracts the curve radius, longitudinal slope, and superelevated transverse slope to provide basic variables for dynamic correction. Based on the turning envelope of the articulated vehicle, the maximum lateral width of the turning envelope is calculated to ensure that the marking design completely covers the maximum lateral offset of the articulated vehicle and reduce the driver's misjudgment of the towing trajectory. The width of the dense trajectory area of ​​small vehicles is calculated to identify the risk area of ​​trajectory overlap and repair defects caused by insufficient safety redundancy space. Calculating the dynamic width requirement helps to achieve dynamic compensation of the side-slip accident rate of the marking width on downhill sections. Generating the final actual dynamic width requirement helps to reduce side-slip accidents caused by fixed lane width, offset the vehicle center of gravity offset effect caused by superelevated transverse slope, and avoid vehicles crossing the line due to centrifugal force. Generating geometric constraints helps to reduce the probability of marking warning failure and helps the marking scheme adapt to road geometry changes in real time.

[0067] In some embodiments, a set of environmental images is analyzed to determine road information and meteorological information; based on the road information, the distance and height difference of roadside cliffs are determined; based on the meteorological information, the average annual frequency of foggy days and the duration of a single fog episode are determined; based on the meteorological information, the fog type is determined; based on the fog type and meteorological information, the driving impact of each type of fog is determined; based on the distance and height difference of roadside cliffs, the characteristics of high-risk areas are determined; based on the average annual frequency of foggy days, the driving impact and the duration of a single fog episode, the visibility hazard characteristics are determined; the high-risk area characteristics and the visibility hazard characteristics are determined as hazard source distribution characteristics.

[0068] The road information may be road geometric feature data obtained by parsing an environment image set.

[0069] Meteorological information can be meteorological state parameters such as fog concentration level, precipitation probability, wind speed, etc. extracted from environmental image sets, meteorological sensor data, and image time series information.

[0070] The roadside cliff distance can be the shortest horizontal distance from the road edge line to the nearest cliff or valley edge.

[0071] Height difference can be the vertical elevation difference between the road plane and the bottom of a cliff or hazardous terrain.

[0072] The average annual frequency of foggy days can be the statistical value of the cumulative number of days with foggy weather in the target area in a year.

[0073] The duration of a single fog event can be the continuous time length from the formation to the dissipation of a single fog event.

[0074] The fog type may be a classification result of the fog.

[0075] Fog properties can be a combination of physical characteristics of fog types.

[0076] Driving impact can be a quantitative indicator of the reduction in driver perception and control capabilities caused by fog.

[0077] The characteristics of high-risk areas can be a composite risk characterization parameter composed of the distance to the roadside cliff and the height difference.

[0078] The visibility hazard characteristic can be a comprehensive meteorological risk indicator that integrates the average annual frequency of foggy days, the duration of a single fog event, and the impact on driving.

[0079] Specifically, the environmental image set is fused at the pixel level, and road information such as road surface texture, shoulder contour, and slope morphological features are extracted through a convolutional neural network. At the same time, meteorological sensor data is matched with image time series information to determine meteorological information. A depth map is generated based on the binocular camera image, and the horizontal projection length from the road edge to the cliff boundary, i.e., the roadside cliff distance, is calculated. Point cloud elevation difference, determine the elevation difference between the road surface and the bottom of the cliff, that is, the height difference. Associate the spatiotemporal labels of the environmental image set with the meteorological database, and count the average annual frequency of foggy days and the duration of a single fog. Based on meteorological information, classify fog types according to the fog formation mechanism: combine the image features of day and night temperature difference and wind speed to determine radiation fog; combine the humidity gradient and wind direction persistence to determine advection fog; determine uphill fog through terrain elevation difference and humidity diffusion pattern. Extract the fog concentration gradient value of the environmental image set through image colorimetry analysis, combine the visibility measured value obtained by the meteorological sensor, and match the predefined fog type classification threshold; based on the matching results, retrieve the light scattering coefficient and humidity saturation of the corresponding fog type from the historical fog database. According to the light scattering coefficient of the fog type, use The dynamic visibility distance is calculated using the law; based on the dynamic visibility distance, the trajectory correction error rate is calculated using a piecewise function model; based on the trajectory correction error rate, standard visibility distance, and dynamic visibility distance, a driving influence coefficient is generated using weighted fusion. A high-risk threshold is defined based on the distance and height difference from roadside cliffs; the boundaries of the danger zone are demarcated based on the threshold, and high-risk area characteristics are generated. The visibility hazard index is calculated based on the average annual frequency of foggy days, the driving influence coefficient, and the duration of a single fog episode; based on the visibility hazard index, a visibility-line marking correction relationship is established; and the visibility hazard characteristics are determined by combining the visibility hazard index and the visibility-line marking correction relationship. The high-risk area characteristics are spatiotemporally matched with the visibility hazard characteristics; based on the matching results, road sections that meet both high cliff risk and high visibility risk are marked as composite hazard sources, and the dynamic line marking enhancement mode is activated. This generates a hazard source distribution characteristic containing four-dimensional parameters: hazard type, spatial location, duration, and risk level.

[0080] Through this solution, extracting road information helps improve the recognition accuracy of road boundaries; determining meteorological information provides data support for time-space matching for hazard source quantification. Determining height difference helps improve the positioning accuracy of high-risk areas. Statistical analysis of the average annual frequency of foggy days and the duration of a single fog episode helps avoid the problem of visual distance attenuation. Determining the fog type helps reduce the decision-making error rate and avoid the misjudgment rate of mixed fog. Calculating dynamic visual distance using the law helps reduce driver prediction errors. Determining the driving impact coefficient for each type of fog helps reduce the rate of speeding and loss of control accidents. Generating high-risk area features helps reduce rollover accidents and mitigate the lack of safety redundancy caused by trajectory competition between articulated vehicles and small vehicles. Determining visibility hazard features helps reduce driver speed perception distortion. Generating hazard source distribution features helps shorten the response time for dynamic adjustments to road marking schemes, reducing driver decision-making errors.

[0081] In some embodiments, the rear axle offset data of the articulated vehicle is determined based on the first vehicle analysis results; the vehicle load status is predicted based on the rear axle offset data; the dynamic envelope range is determined based on the vehicle load status and the rear axle offset data; the dynamic envelope range is corrected based on the coupling relationship between the rear axle offset data and the super-high cross slope to obtain the turning envelope of each articulated vehicle.

[0082] The rear axle offset data may be the lateral displacement of the rear axle centerline relative to a theoretical driving trajectory during the driving process of the articulated vehicle.

[0083] The vehicle load status may be a vehicle load level.

[0084] The dynamic envelope range may be a vehicle turning space requirement boundary.

[0085] Specifically, the real-time rear axle offset is obtained through the on-board displacement sensor, and the rear axle offset data is calculated in combination with the reference wheelbase in the first vehicle analysis result. The load level is divided based on the rear axle offset data, and the random forest algorithm is used to fuse the number of axle load sensors and the rear axle offset data to predict the vehicle load status. The minimum turning radius is calculated based on the rear axle offset data corresponding to the vehicle load status and the first vehicle analysis result; the minimum turning radius is used as a reference, and the safety redundancy factor is superimposed to generate the dynamic envelope range. Based on the minimum turning radius, the equivalent offset increment caused by the superelevated cross slope is calculated; based on the equivalent offset increment, the lateral expansion of the envelope is corrected; the dynamic envelope range and the lateral expansion of the envelope are integrated to generate the turning envelope of each articulated vehicle.

[0086] This solution calculates rear axle offset data, providing the foundation for dynamic envelope calculations. Predicting vehicle load status helps reduce line-marking indentation errors in areas where articulated vehicles compete with smaller vehicles. Generating a dynamic envelope helps reduce skidding accidents and errors in predicting curve radius, shortening driver braking response delays. Generating a turning envelope for each articulated vehicle helps reduce decision-making errors.

[0087] In some embodiments, a plurality of marking line positions are determined based on the trajectory range; and a marking line confidence level of each marking line position is determined based on the trajectory repeatability, using the following formula: (1) in, Indicates the confidence level of the marking line; ; ; represents the trajectory discrete variance; Based on the geometric constraints, the minimum allowable width is determined using the following formula: (2) in, Indicates the minimum allowed width; ; The minimum permissible width for a single lane as specified in the national road design specifications, usually 3.5 meters; The compensation intensity is determined based on the distribution characteristics of the hazard sources; the optimal marking position and compensation plan are determined based on the marking confidence, minimum allowable width and compensation intensity of each marking position to obtain the marking plan.

[0088] The marking position may be a set of candidate marking placement points.

[0089] The road marking confidence can be an indicator that quantitatively represents the degree of match between the positions of several road markings and the actual driving path of the vehicle in the historical trajectory data.

[0090] The minimum permissible width may be the minimum lane physical space width.

[0091] The compensation intensity can be a marking parameter adjustment level set for different hazard source types.

[0092] The optimal marking position may be the optimal marking layout point.

[0093] The compensation scheme can be a set of marking parameter adjustment strategies.

[0094] Specifically, based on the spatial distribution characteristics of the trajectory range, a density clustering algorithm is used to identify high-frequency trajectory areas and eliminate discrete noise points; with the geometric center of the trajectory-dense area as the benchmark, several marking positions are generated according to the dynamic interval algorithm. Define the trajectory repeatability index, count the trajectory repeatability of several marking positions in the historical trajectory data, and screen out the marking confidence of each marking position. Based on the lateral force balance equation under the coupling effect of the longitudinal slope and the super-elevated transverse slope, calculate the slope correction coefficient, and then dynamically adjust the lane width reference value; thereby deriving the minimum allowable width. According to the distribution characteristics of the hazard source, a risk-visibility attenuation mapping table is constructed, and the compensation level is assigned according to the hazard source type; based on the real-time meteorological data, the compensation intensity is determined according to the compensation level. According to the marking confidence, minimum allowable width and compensation intensity of each marking position, a multi-objective optimization model is established by integrating multi-objective optimization theory, traffic engineering specifications, driving behavior models and empirical data, and then the Pareto optimal solution is used for screening to generate the final marking solution. Trajectory coverage frequency It indicates the ratio of a certain marking position to be covered by historical trajectories, reflecting the driver's actual path selection preference. Indicates the number of trajectory points that fall within the buffer zone (e.g. ±0.5m) of the candidate position of the marking line; Indicates the total number of trajectory points in this section; trajectory discrete variance penalty Indicates the correction of the confidence drop caused by the dispersion of trajectory distribution. The larger the value, the more significant the confidence decay. If the trajectory is highly concentrated ( ), the penalty term approaches 1, and the confidence is dominated by the coverage frequency. If the trajectory is discrete ( ), the penalty term approaches 0, reducing the confidence. Exponential function It can quickly attenuate the impact of high variance, which is consistent with the perception in driving behavior that "the higher the trajectory consistency, the stronger the effectiveness of the markings"; The specific setting needs to be given through multiple tests. The existing "Highway Engineering Technical Standards" Article 4.0.2 stipulates that when a fourth-level highway uses a single lane, the lane width should be 3.5 meters; when dynamic requirements , based on the dynamic calculation results, adapt to complex terrain; if the dynamic calculation value is insufficient or error , it is mandatory to take 3.5m to ensure legal compliance.

[0095] This solution eliminates discrete noise points, helping to reduce interfering trajectories caused by driver errors or temporary obstacle avoidance. Generating several marking positions helps reduce the trajectory conflict rate for each type of vehicle. Trajectory repeatability helps screen out the trajectory areas frequently used by drivers in actual driving. Determining the confidence level of the markings helps reduce the impact of abnormal trajectories such as speeding or underspeed on the confidence level and reduces the misjudgment rate of curve trajectories. Determining the minimum allowable width helps improve the adaptability of the marking width on downhill sections, reducing the rate of side-slip accidents and the incidence of critical dangerous states. Determining the compensation intensity helps shorten the delay time for driver trajectory correction, increase the safe distance between vehicles and roadside hazards, and reduce the driver's predicted error in the curve radius. Generating the final marking solution helps reduce the error rate in decision-making at sudden changes in markings and improve the timeliness of risk response.

[0096] In some embodiments, a vehicle dynamics constraint set is determined based on a historical trajectory data set; a driving psychological expectation model is constructed based on the historical trajectory data set; the coupling relationship between the vehicle dynamics constraint set and the driving psychological expectation model is analyzed, and the road marking scheme is dynamically compensated and corrected to generate a final road marking drawing scheme.

[0097] The vehicle dynamics constraint set may be a set of dynamic parameters including the type of traveling vehicle and load status.

[0098] The driving psychological expectation model can be a quantitative model constructed through analysis of driver cognitive characteristics and behavioral data.

[0099] The final marking drawing solution can be a marking parameter instruction set generated after data fusion and correction.

[0100] Specifically, the method separates trajectory samples of articulated vehicles and small vehicles from historical trajectory datasets. Then, based on the vehicle wheelbase, track width, and maximum steering angle, the turning envelope of the articulated vehicle is calculated to measure the spatial competition between the trajectories of articulated vehicles and small vehicles. The vehicle lateral offset threshold is derived by combining the curve radius and longitudinal slope gradient, thereby determining the vehicle dynamics constraint set. The driver's prediction error of the curve radius and the braking response delay are statistically analyzed to generate a prediction error threshold. A correlation mapping between the speed perception distortion rate and road marking warnings on longitudinal slope sections is established to generate a reaction time window. The correlation between the decision error rate at sudden changes in road markings and the spacing between road markings is analyzed to extract the driver's trajectory correction success rate for gradually changing road markings and generate a visual continuity requirement. Through a multidisciplinary approach based on cognitive psychology, human factors engineering, control theory, and empirical driving research, the prediction error threshold, reaction time window, and visual continuity requirement are dynamically coupled into a driving psychological expectation model. The vehicle dynamics constraint set is compared with the driver's psychological expectation model in real time. If conflicts persist, the road marking position is dynamically adjusted. The road marking parameters corresponding to the maximum value of the optimization objective function are iteratively solved through the gradient descent method. The road marking parameters are imported into the simulation environment and the road design parameters are set. If the trajectory deviation exceeds the standard, the node coordinates are called back until the standard is met. Vehicle trajectory data is collected in real time and the latest confidence level is calculated. The weight fluctuations are then smoothed through the sliding window algorithm to generate the final road marking drawing plan.

[0101] This solution separates trajectory samples for articulated vehicles and small vehicles, helping to reduce vehicle trajectory misjudgment and improve road marking warning effectiveness. Determining the vehicle dynamics constraint set helps reduce skidding accidents and accidents on high-risk roads, increasing safety margins. Determining a driver's psychological expectation model helps shorten driver braking response delays, reducing cornering accidents, while also lowering speed perception distortion and minimizing speeding and loss-of-control accidents. Generating a final road marking plan helps maintain a stable, long-term decline in accident rates.

[0102] Figure 3 A schematic diagram of a road marking system based on image recognition is provided in one embodiment of the present application. Figure 3 As shown, the road marking system 300 based on image recognition in this embodiment includes: a data acquisition module 301 , a data analysis module 302 , a parameter analysis module 303 , an image analysis module 304 , and a solution determination module 305 .

[0103] The data acquisition module 301 is used to obtain a historical trajectory data set, a road design parameter set, and an environmental image set of a target road; A data analysis module 302 is used to analyze the historical trajectory data set to determine trajectory repetitiveness; A parameter analysis module 303 is used to determine geometric constraints based on the road design parameter set; An image analysis module 304 is configured to determine the distribution characteristics of hazard sources based on the environmental image set; The scheme determination module 305 is used to determine the marking scheme according to the trajectory repeatability, the geometric constraints and the hazard source distribution characteristics.

[0104] Optionally, when the data analysis module 302 analyzes the historical trajectory data set to determine the trajectory repetitiveness, it is used to: Analyze the historical trajectory data set to determine the type of traveling vehicle and the corresponding type of driving trajectory set; For each type of traveling vehicle, analyzing the set of driving trajectories to determine a spatial deviation distribution corresponding to the type of traveling vehicle; Establishing a spatiotemporal four-dimensional filtering model, eliminating abnormal data based on the spatiotemporal four-dimensional filtering model and according to the spatial deviation distribution, and determining a trajectory range of the driving trajectory set after elimination according to the spatial deviation distribution; The trajectory repeatability is determined according to the trajectory range.

[0105] Optionally, the types of traveling vehicles include articulated vehicles and small vehicles. When the data analysis module 302 determines the trajectory range of the eliminated traveling trajectory set according to the spatial deviation distribution, it is configured to: For the articulated vehicle, analyzing the driving trajectory set of the articulated vehicle and the driving trajectory set of the small vehicle to obtain a first vehicle analysis result of the articulated vehicle and a second vehicle analysis result of the small vehicle; determining a turning envelope of each articulated vehicle based on the first vehicle analysis result; determining a turning boundary of the articulated vehicle based on the turning envelope; Based on the turning boundary, according to the second vehicle analysis result and the spatial deviation distribution, a trajectory range of the driving trajectory set after elimination is determined.

[0106] Optionally, when determining geometric constraints based on the road design parameter set, the parameter analysis module 303 is configured to: Analyzing the road design parameter set to obtain curve radius, longitudinal slope, and superelevation transverse slope; Analyze the turning envelope and the spatial deviation distribution of small vehicles to determine the dynamic width requirement; According to the coupling relationship between the longitudinal slope and the superelevation transverse slope, the dynamic width requirement is modified to obtain an actual dynamic width requirement; The actual dynamic width requirement is determined as a geometric constraint.

[0107] Optionally, when determining the distribution characteristics of hazard sources based on the environmental image set, the image analysis module 304 is configured to: Analyzing the environmental image set to determine road information and weather information; Determining the distance and height difference of the roadside cliffs based on the road information; Determine the average annual frequency of foggy days and the duration of a single foggy day based on the meteorological information; determining the fog type according to the meteorological information; determining a driving impact of each type of fog based on the fog type and the meteorological information; Determine the characteristics of the high-risk area based on the distance and height difference of the roadside cliffs; Determine visibility hazard characteristics based on the average annual frequency of foggy days, the driving impact, and the duration of a single fog episode; The high-risk area characteristics and the visibility hazard characteristics are determined as hazard source distribution characteristics.

[0108] Optionally, when determining the turning envelope of each articulated vehicle based on the first vehicle analysis result, the data analysis module 302 is configured to: determining rear axle offset data of the articulated vehicle based on the first vehicle analysis result; predicting a vehicle load state based on the rear axle offset data; determining a dynamic envelope range according to the vehicle load state and the rear axle offset data; According to the coupling relationship between the rear axle offset data and the superelevated transverse slope, the dynamic envelope range is corrected to obtain the turning envelope of each articulated vehicle.

[0109] Optionally, when determining a marking scheme based on the trajectory repeatability, the geometric constraints, and the hazard source distribution characteristics, the scheme determination module 305 is configured to: Determining a plurality of marking line positions according to the trajectory range; Determining the confidence level of each marking line position according to the trajectory repeatability; Determining a minimum allowable width based on the geometric constraints; Determine the compensation intensity based on the distribution characteristics of the hazard sources; According to the marking confidence of each marking position, the minimum allowable width and the compensation strength, an optimal marking position and compensation scheme are determined to obtain a marking scheme.

[0110] Optionally, the road marking system based on image recognition further includes a solution generation module 306 for: determining a vehicle dynamics constraint set based on the historical trajectory dataset; Constructing a driving psychological expectation model based on the historical trajectory data set; The coupling relationship between the vehicle dynamics constraint set and the driving psychological expectation model is analyzed, and the road marking scheme is dynamically compensated and corrected to generate a final road marking drawing scheme.

[0111] The system of this embodiment can be used to execute the method of any of the above embodiments. Its implementation principles and technical effects are similar and will not be described in detail here.

[0112] Figure 4 A schematic diagram of the structure of an electronic device provided in one embodiment of the present application is shown in FIG. Figure 4 As shown, the electronic device 400 of this embodiment may include: a memory 401 and a processor 402.

[0113] The memory 401 stores a computer program that can be loaded by the processor 402 and execute the method in the above embodiment.

[0114] The processor 402 and the memory 401 are connected, for example, via a bus.

[0115] Optionally, the electronic device 400 may further include a transceiver. It should be noted that in actual applications, the number of transceivers is not limited to one, and the structure of the electronic device 400 does not constitute a limitation on the embodiments of the present application.

[0116] Processor 402 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the present disclosure. Processor 402 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, or a combination of a DSP and a microprocessor.

[0117] A bus includes a path that transmits information between the components mentioned above. Examples include a PCI (Peripheral Component Interconnect) bus and an EISA (Extended Industry Standard Architecture) bus. Buses can be categorized as address buses, data buses, and control buses. For ease of illustration, the diagram uses a single thick line, but this does not imply a single bus or type of bus.

[0118] The memory 401 may be a ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, a RAM (Random Access Memory) or other type of dynamic storage device that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0119] The memory 401 is used to store application code for executing the solution of the present application, and the execution is controlled by the processor 402. The processor 402 is used to execute the application code stored in the memory 401 to implement the content shown in the above method embodiment.

[0120] Electronic devices include, but are not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. They may also include servers, etc. Figure 4 The electronic device shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0121] The electronic device of this embodiment can be used to execute the method of any of the above embodiments. The implementation principles and technical effects are similar and will not be described in detail here.

[0122] The present application also provides a computer-readable storage medium storing a computer program that can be loaded by a processor and execute the method in the above embodiment.

[0123] Those skilled in the art will appreciate that all or part of the steps in the above-described method embodiments can be implemented using hardware associated with program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

Claims

1. A road marking drawing method based on image recognition, characterized in that: include: Obtain the historical trajectory dataset, road design parameter set, and environmental image set of the target road; Analyzing the historical trajectory data set to determine trajectory repeatability; Determining geometric constraints based on the road design parameter set; determining the distribution characteristics of the hazard sources according to the environmental image set; A line marking scheme is determined based on the trajectory repeatability, the geometric constraints and the hazard source distribution characteristics.

2. The method according to claim 1, characterized in that The analyzing the historical trajectory data set to determine trajectory repetitiveness includes: Analyze the historical trajectory data set to determine the type of traveling vehicle and the corresponding type of driving trajectory set; For each type of traveling vehicle, analyzing the set of driving trajectories to determine a spatial deviation distribution corresponding to the type of traveling vehicle; Establishing a spatiotemporal four-dimensional filtering model, eliminating abnormal data based on the spatiotemporal four-dimensional filtering model and according to the spatial deviation distribution, and determining a trajectory range of the driving trajectory set after elimination according to the spatial deviation distribution; The trajectory repeatability is determined according to the trajectory range.

3. The method according to claim 2, characterized in that The types of traveling vehicles include articulated vehicles and small vehicles. The determining of the trajectory range of the eliminated traveling trajectory set according to the spatial deviation distribution includes: For the articulated vehicle, analyzing the driving trajectory set of the articulated vehicle and the driving trajectory set of the small vehicle to obtain a first vehicle analysis result of the articulated vehicle and a second vehicle analysis result of the small vehicle; determining a turning envelope of each articulated vehicle based on the first vehicle analysis result; determining a turning boundary of the articulated vehicle based on the turning envelope; Based on the turning boundary, according to the second vehicle analysis result and the spatial deviation distribution, a trajectory range of the driving trajectory set after elimination is determined.

4. The method according to claim 3, characterized in that Determining geometric constraints based on the road design parameter set includes: Analyzing the road design parameter set to obtain curve radius, longitudinal slope, and superelevation transverse slope; Analyze the turning envelope and the spatial deviation distribution of small vehicles to determine the dynamic width requirement; According to the coupling relationship between the longitudinal slope and the superelevation transverse slope, the dynamic width requirement is modified to obtain an actual dynamic width requirement; The actual dynamic width requirement is determined as a geometric constraint.

5. The method according to claim 1, wherein Determining the distribution characteristics of the hazard sources according to the environmental image set includes: Analyzing the environmental image set to determine road information and weather information; Determining the distance and height difference of the roadside cliffs based on the road information; Determine the average annual frequency of foggy days and the duration of a single foggy day based on the meteorological information; determining the fog type according to the meteorological information; determining a driving impact of each type of fog based on the fog type and the meteorological information; Determine the characteristics of the high-risk area based on the distance and height difference of the roadside cliffs; Determine visibility hazard characteristics based on the average annual frequency of foggy days, the driving impact, and the duration of a single fog episode; The high-risk area characteristics and the visibility hazard characteristics are determined as hazard source distribution characteristics.

6. The method according to claim 4, characterized in that Determining the turning envelope of each articulated vehicle based on the first vehicle analysis result includes: determining rear axle offset data of the articulated vehicle based on the first vehicle analysis result; predicting a vehicle load state based on the rear axle offset data; determining a dynamic envelope range according to the vehicle load state and the rear axle offset data; According to the coupling relationship between the rear axle offset data and the superelevated transverse slope, the dynamic envelope range is corrected to obtain the turning envelope of each articulated vehicle.

7. The method according to claim 2, characterized in that The determining of the marking scheme according to the trajectory repeatability, the geometric constraint conditions and the hazard source distribution characteristics includes: Determining a plurality of marking line positions according to the trajectory range; Determining the confidence level of each marking line position according to the trajectory repeatability; Determining a minimum allowable width based on the geometric constraints; Determine the compensation intensity based on the distribution characteristics of the hazard sources; According to the marking confidence of each marking position, the minimum allowable width and the compensation strength, an optimal marking position and compensation scheme are determined to obtain a marking scheme.

8. The method according to claim 1, characterized in that After determining the optimal marking position and compensation scheme based on the marking confidence of each marking position, the minimum allowable width, and the compensation strength, and obtaining the marking scheme, the method further includes: determining a vehicle dynamics constraint set based on the historical trajectory dataset; Constructing a driving psychological expectation model based on the historical trajectory data set; The coupling relationship between the vehicle dynamics constraint set and the driving psychological expectation model is analyzed, and the road marking scheme is dynamically compensated and corrected to generate a final road marking drawing scheme.

9. A road marking system based on image recognition, applied to the method according to any one of claims 1 to 8, characterized in that: include: A data acquisition module is used to obtain a historical trajectory dataset, a road design parameter set, and an environmental image set of the target road; A data analysis module, configured to analyze the historical trajectory data set and determine trajectory repetitiveness; a parameter analysis module, configured to determine geometric constraints based on the road design parameter set; An image analysis module, configured to determine the distribution characteristics of hazard sources based on the environmental image set; The scheme determination module is used to determine the marking scheme according to the trajectory repeatability, the geometric constraints and the hazard source distribution characteristics.

10. An electronic device, characterized in that: include: memory and processor; The memory is used to store program instructions; The processor is configured to call and execute program instructions in the memory to perform the method according to any one of claims 1 to 8.

Images