CN119694132A - A highway high-risk section early warning method and system - Google Patents
A highway high-risk section early warning method and system Download PDFInfo
- Publication number
- CN119694132A CN119694132A CN202510192505.0A CN202510192505A CN119694132A CN 119694132 A CN119694132 A CN 119694132A CN 202510192505 A CN202510192505 A CN 202510192505A CN 119694132 A CN119694132 A CN 119694132A
- Authority
- CN
- China
- Prior art keywords
- risk
- data
- environment
- vehicle
- real
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Traffic Control Systems (AREA)
Abstract
本发明涉及智能交通安全管控技术领域，公开了一种公路高危路段预警方法及系统，包括：利用车辆端和路侧传感器采集实时环境及车辆数据，将实时环境及车辆数据上传至云端平台进行数据融合，获取融合数据；在云端平台建立风险判定模型，通过增量训练实时更新风险判定模型，利用风险判定模型对融合数据进行分析，生成综合风险评分；根据综合风险评分触发分级预警，并在车辆端进行提前预警，将预警结果进行可视化展示。本发明方法提高了安全检测系统对多变气象与车辆行为的适应能力，最终有效减少事故发生，提升驾驶体验与交通效率。
The present invention relates to the field of intelligent traffic safety control technology, and discloses a highway high-risk section early warning method and system, including: using vehicle-side and roadside sensors to collect real-time environment and vehicle data, uploading the real-time environment and vehicle data to a cloud platform for data fusion, and obtaining fused data; establishing a risk determination model on the cloud platform, updating the risk determination model in real time through incremental training, using the risk determination model to analyze the fused data, and generating a comprehensive risk score; triggering a graded early warning according to the comprehensive risk score, and giving an early warning on the vehicle side, and visually displaying the early warning results. The method of the present invention improves the adaptability of the safety detection system to changeable weather and vehicle behavior, and ultimately effectively reduces the occurrence of accidents, and improves driving experience and traffic efficiency.
Description
Technical Field
The invention relates to the technical field of intelligent traffic safety control, in particular to a highway high-risk road section early warning method and system.
Background
With the rapid development of intelligent transportation systems (INTELLIGENT TRANSPORTATION SYSTEM, ITS) and internet of vehicles (V2X), road safety monitoring and early warning are gradually changing from traditional static layout to dynamic sensing and real-time feedback. The existing traffic management means mainly depend on single data sources such as fixed cameras, roadside weather stations and the like to collect road conditions or weather information. However, in complex environments such as highways and mountain dangerous roads, it is difficult to accurately sense instantaneous traffic flow changes and microscopic meteorological conditions simply by means of fixed monitoring. Meanwhile, due to network bandwidth and difficulty in fusion of multi-source heterogeneous data, most existing systems often cannot realize real-time quantitative evaluation of sudden risk factors such as vehicle motion trail, road surface icing, rapid reduction of visibility and the like. In recent years, big data and cloud computing platforms have been tried to be applied in urban signal control and high-speed charging scheduling, but related technical systems for carrying out fine early warning on highway high-risk road sections are not completely mature, and an overall solution with instantaneity and prediction accuracy is lacking.
Aiming at the monitoring and early warning requirements of high-risk road sections, the traditional method is to manually set a specific threshold or warning sign based on historical accident statistics and basic meteorological information. However, such means cannot timely master the dynamic traffic flow, and is difficult to comprehensively analyze by combining the running data of the vehicle, so that early warning lag or excessive warning is easily caused. In addition, some prior arts have tried to integrate vehicle-mounted sensors with road side monitoring data, but often lack a systematic incremental training mechanism, and cannot quickly adjust risk judging parameters under continuously updated road conditions and meteorological conditions, so that prediction reliability is still insufficient under conditions of sudden rain and fog, road snow or extreme topography. Especially when multi-source data fusion, vehicle end and road side end linkage are needed, and real-time cloud end uploading is performed for deep analysis, the problems of low data processing efficiency, limited risk quantification capability, single early warning response mode and the like of the prior art still exist, and high standard requirements of highways and complex road conditions on safety management and control are difficult to comprehensively meet.
Disclosure of Invention
The present invention has been made in view of the above-described problems.
Therefore, the invention solves the technical problem that the existing highway high-risk road section early warning technology has limitations in the aspects of multi-source data real-time fusion, dynamic updating of a model and hierarchical early warning linkage.
The highway high-risk road section early warning method comprises the following steps of collecting real-time environment and vehicle data by using vehicle end and road side sensors, uploading the real-time environment and the vehicle data to a cloud platform for data fusion, and obtaining fusion data;
establishing a risk judgment model on a cloud platform, updating the risk judgment model in real time through incremental training, and analyzing the fusion data by using the risk judgment model to generate a comprehensive risk score;
and triggering grading early warning according to the comprehensive risk score, carrying out early warning on a vehicle end, and visually displaying the early warning result.
The method for early warning the highway high-risk road section comprises the steps of collecting real-time environment and vehicle data, wherein a vehicle-mounted sensor is used for collecting the running state and surrounding environment data of a vehicle;
And arranging a sensing device on the high-risk road section to collect road condition and meteorological data.
The fusion data comprises extracting key feature vectors aiming at the characteristics of the highway high-risk road section, and carrying out weight adjustment on the key feature vectors according to real-time environment and vehicle data;
Setting key environment dimensions according to road condition and meteorological data, wherein the key environment dimensions comprise meteorological conditions, road states and road section characteristics, and establishing an environment tag library according to the key environment dimensions, wherein each environment tag corresponds to one road condition and meteorological data;
Acquiring historical data of environment and vehicle data, and setting initial weight of key feature vector as according to the historical data ;
Mapping the collected real-time environment and vehicle data into an environment tag library, adjusting the weight of the key feature vector according to the environment tag, and adjusting the weight after adjustmentAnd multiplying the obtained product by the key feature vector to obtain fusion data.
As a preferable scheme of the highway high-risk road section early warning method, the risk judging model comprises an input layer, a hidden layer and an output layer, wherein the input layer receives the fusion data as the input of the risk judging model;
Extracting deeper characteristic association by the hidden layer according to the fusion data, dividing the hidden layer into different modules according to functions, and analyzing one type of real-time environment and vehicle data by each module;
the output layer outputs the comprehensive risk score;
according to the output of the risk judgment model and the calculation error of the real risk score, training the risk judgment model through an environment weighted mean square error loss function, wherein the training is expressed as follows:
wherein, Representing an environmentally weighted mean square error loss function; representing the number of samples during training of the risk judgment model; the representation is for the first Environmental weighting factors for the individual samples; Represent the first True risk score for each sample; Representation model pair number Comprehensive risk scores output by the individual samples; Represent the first Samples.
As a preferable scheme of the highway high-risk road section early warning method, the environment weighting factors comprise category punishment weights and environment weights, wherein the category punishment weights represent the road high-risk road section early warning method aiming at the first roadClass penalty coefficients for the individual samples, the penalty coefficients being adjusted according to the gap between the model predicted composite risk score and the true risk score:
wherein, Representing class penalty weights; Representing class penalty weight magnification;
the environmental weight is adjusted according to the real-time environment and the environment label obtained by mapping the vehicle data to the environment label library, and the adjusted weight is used for adjusting the environment label And initial weightIs used to determine the environmental weight of the vehicle, expressed as:
wherein, The weight of the environment is represented by the weight of the environment,Indicating the ambient weight magnification.
As an optimized scheme of the highway high-risk road section early warning method, the real-time updating the risk judging model through incremental training comprises the steps of setting a fixed time window according to real-time environment and vehicle data, and enabling the time window to pass each timeTriggering one-time incremental training;
Window the time Mapping the real-time environment and the vehicle data in the model to an environment tag library, acquiring environment tags of the real-time environment and the vehicle data, determining the correlation degree between the current data and a hidden layer module of the risk judgment model based on the environment tags, updating only a module with high correlation degree, and setting the learning rate of a module with low correlation degree to be zero;
if the presence of continuity is detected The un-updated module is trained for a second increment, and the relevance judgment is not performed, so that a time window is usedUpdating the corresponding module by the data in the memory;
Recording the prediction precision of the risk judgment model after each update is completed, and adjusting the learning rate of the risk judgment model module according to the prediction precision;
setting a global update period Summarizing all time windows in a period every time a global update period passesTraining and updating the whole risk judgment model, and unifying weight association among the calibration modules.
The method for warning the highway high-risk road section according to the invention comprises the steps of comparing the comprehensive risk score with a preset risk threshold, dividing the risk level into four types of risk-free, low-risk, high-risk and extremely high-risk, determining risk data according to real-time environment, vehicle data and risk labels, and visually displaying the risk data and the risk labels;
when the risk level is risk-free, the highway high-risk road section early warning is not carried out;
when the risk level is low risk, early warning is carried out through a vehicle-mounted display and voice, and a speed limit warning is sent out through a road side information board;
When the risk level is high risk, linking with vehicle navigation on the basis of low risk early warning measures, early warning vehicles containing high risk road sections in the navigation route in advance, and prompting to re-plan the route;
When the risk level is extremely high, the road section closing measures are implemented on the basis of the high risk prediction measures, and the staff is informed to prepare corresponding risk treatment measures.
The invention relates to a highway high-risk road section early warning system adopting any one of the methods, wherein an acquisition module acquires real-time environment and vehicle data by using a vehicle end and a road side sensor, and uploads the real-time environment and the vehicle data to a cloud platform for data fusion to acquire fusion data;
the processing module establishes a risk judgment model on the cloud platform, updates the risk judgment model in real time through incremental training, and analyzes the fusion data by using the risk judgment model to generate a comprehensive risk score;
and the early warning module triggers hierarchical early warning according to the comprehensive risk score, performs early warning at the vehicle end and performs visual display on the early warning result.
A computer device comprising a memory and a processor, the memory storing a computer program comprising steps for implementing the method according to any one of the invention when the processor executes the computer program.
A computer readable storage medium having stored thereon a computer program comprising steps for carrying out the method according to any one of the present invention when said computer program is executed by a processor.
The method has the beneficial effects that the vehicle-side and road-side sensors are used for acquiring multidimensional real-time data, and the multidimensional real-time data are fused at the cloud for processing, so that road risk factors are timely identified and quantified. Based on the incremental training model, the weight can be dynamically updated along with new data, so that the risk judgment precision is remarkably improved, and the dependence on fixed rules is reduced. The importance of extreme environments and serious errors on model training is highlighted by carrying out environment weighting and class punishment loss calculation on the fusion data, so that potential high-risk road sections are rapidly positioned. The grading early warning strategy depends on the comprehensive risk scoring threshold value, can prompt through a vehicle-mounted display or a road side information board, and can also be in closed linkage with a navigation system and a road, so that the traffic safety of vehicles is enhanced. The modularized increment training and the global period correction are combined, the high-correlation module is updated in a concentrated mode while the low-correlation characteristic is kept stable, the adaptability to variable weather and vehicle behaviors is further improved, the accident occurrence is finally effectively reduced, and the driving experience and the traffic efficiency are improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is an overall flowchart of a highway high-risk road section early warning method according to an embodiment of the present invention.
Detailed Description
So that the manner in which the above recited objects, features and advantages of the present invention can be understood in detail, a more particular description of the invention, briefly summarized above, may be had by reference to the embodiments, some of which are illustrated in the appended drawings. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
Embodiment 1, referring to fig. 1, provides a highway high-risk section early warning method according to an embodiment of the present invention, including:
And S1, acquiring real-time environment and vehicle data by using a vehicle end and a road side sensor, uploading the real-time environment and the vehicle data to a cloud platform for data fusion, and acquiring fusion data.
Further, real-time environment and vehicle running information are acquired from the vehicle end and the road side end so as to comprehensively reflect the dynamic conditions of the high-risk road section. Specifically, the vehicle end mainly records the running speed, acceleration, steering angle and surrounding environment image or obstacle information of the vehicle in real time through various monitoring devices such as a vehicle-mounted radar, a camera, a GPS, an accelerometer, a steering angle sensor and the like. The data can reflect the running state of the current vehicle and the surrounding traffic situation, and provide important basis for subsequent risk judgment. Meanwhile, a vehicle-mounted communication unit can be deployed in the vehicle, so that the data can be rapidly transmitted to the rear-end platform in a V2I or V2X mode after being acquired.
Aiming at the road side of the high-risk road section, multiple types of sensing equipment are arranged to realize three-dimensional perception of the external environment. The sensing equipment comprises, but is not limited to, intelligent monitoring cameras, laser radars, geomagnetic sensors, meteorological sensors (such as temperature, humidity, visibility detection and the like), and special sensing devices (such as snow cover, icing, ponding detectors and the like) capable of detecting road surface conditions. In addition, a radar device may be provided on the road side to monitor information such as the speed, distance, and flow rate of vehicles passing by and coming from the vehicle.
And finally uploading the multi-source data to a data processing platform or a cloud end in real time through a wired or wireless network for preliminary aggregation and formatting processing to form fusion data for a subsequent risk judgment model. Through the multi-source sensing and dynamic interaction, the recognition accuracy and early warning timeliness of potential hazards of the highway high-risk road sections can be remarkably improved.
Further, according to the collected multi-source environment and vehicle data, aiming at the characteristics of the highway high-risk road section, key feature vectors are extracted, and weight adjustment is carried out on the key feature vectors according to the real-time environment and the vehicle data.
Specifically, a key environment dimension is set according to road conditions and meteorological data, the key environment dimension comprises meteorological conditions, road states and road section characteristics, an environment tag library is built according to the key environment dimension, and each environment tag corresponds to one road condition and meteorological data.
The weather conditions are classified into sunny, rainy, snowy, foggy, typhoon and the like, the road conditions are classified into dry, wet and slippery, frozen, snow and the like, and the road section characteristics are classified into flat, continuous curves, long downhill slopes, mountain areas, watery cliffs and the like. Each label in the environment label library consists of meteorological conditions, road states and road section characteristics, different severity degrees are set by each label characteristic through a threshold value or a judging rule, a real-time environment perception vector is mapped into one or more environment labels according to a predefined threshold value or rule, for example, a medium-level rainy day, a wet road section and a continuous curve, the threshold value of the environment label library is set based on the local specific condition of a high-risk road section, and the collected real-time data is compared with the threshold value to quickly determine the environment label of the current road section.
Acquiring historical data of environment and vehicle data, and setting initial weight of key feature vector as according to the historical dataThe weight is used to identify the relative importance of each feature in risk determination under general circumstances (e.g., sunny days, straight road segments).
Mapping the collected real-time environment and vehicle data into an environment tag library for matching, judging the environment type of the current road section, calling a corresponding weight bias mapping table according to the environment tag, carrying out weight adjustment on key feature vectors, and carrying out weight adjustment after adjustmentAnd multiplying the obtained product by the key feature vector to obtain fusion data.
The adjustment value of the weight bias mapping table to the initial vector can be preset based on expert experience or statistical analysis (for example, the importance of the road surface humidity and the braking frequency is improved by 20 percent in rainy days), or can be continuously learned by the system in long-term operation and stored in the weight bias mapping table to form a self-adaptive mechanism.
It should be noted that, to avoid that the weights of some features are excessively enlarged or reduced in a single correction, the adjustment amounts of the weights are constrained by setting upper and lower threshold values, and when the system detects that the environmental label changes frequently (such as from "light rain" to "heavy rain" to "no rain" in a short time), a smoothing strategy is introduced, so as to gradually correct the weights, so as to prevent excessive oscillation of the model caused by excessively frequent environmental fluctuation.
The method not only carries out static correction on single weather or road conditions, but also supports multi-label superposition, and distributes a strategy with the highest weight superposition or priority according to the actual dangerous degree, thereby realizing the characteristic tuning of finer granularity. In the risk prediction of the highway high-risk road section, the characteristic weight can be adjusted in real time according to complex and changeable weather and road conditions, and the most representative characteristic is amplified or highlighted, so that the overall accuracy and the robustness of the risk judgment model are improved.
And S2, establishing a risk judgment model on the cloud platform, updating the risk judgment model in real time through incremental training, and analyzing the fusion data by using the risk judgment model to generate a comprehensive risk score.
The risk judging model comprises an input layer, a hidden layer and an output layer, wherein the input layer receives the fusion data as input of the risk judging model, the hidden layer extracts deeper characteristic association according to the fusion data, the hidden layer is divided into different modules according to functions, each module analyzes real-time environment and vehicle data, and the output layer outputs comprehensive risk scores.
And calculating errors according to the output of the risk judgment model and real risk scores, wherein the real risk scores are obtained through expert evaluation based on real risk events (such as vehicle accidents, faults and the like) which occur in historical data, and are used for reflecting the actual risk level. Training the risk judgment model through an environment weighted mean square error loss function can effectively highlight the weight of severe environment or key errors, and the loss function is expressed as:
wherein, Representing an environmentally weighted mean square error loss function; representing the number of samples during training of the risk judgment model; the representation is for the first Environmental weighting factors for the individual samples; Represent the first True risk score for each sample; Representation model pair number Comprehensive risk scores output by the individual samples; Represent the first Samples.
Further, the environmental weighting factor includes a class penalty weight and an environmental weight, and the value of the environmental weighting factor takes the product of the class penalty weight and the environmental weight.
Wherein the class penalty weight is used to distinguish the effect of the model's error type on the loss at the time of prediction, and the environment weight is used to amplify (or shrink) the error duty ratio under different environment scenarios. The environmental weight and the class penalty are focused on "which environment is wrong" and "what type of error", respectively, and these two dimensions do not conflict, and often need to be considered simultaneously to obtain a weighting strategy which meets the actual requirements.
The category penalty weight representation is for the firstClass penalty coefficients for the individual samples, the penalty coefficients being adjusted according to the gap between the model predicted composite risk score and the true risk score:
wherein, Representing class penalty weights; And (5) representing class punishment weight magnification.
The environmental weight is adjusted according to the real-time environment and the environment label obtained by mapping the vehicle data to the environment label library, and the adjusted weight is used for adjusting the environment labelAnd initial weightIs used to determine the environmental weight of the vehicle, expressed as:
wherein, The weight of the environment is represented by the weight of the environment,Indicating the ambient weight magnification. If the environmental weight and the class penalty are over large, gradient explosion or unstable training can be caused after accumulation, and measures such as maximum penalty coefficient limitation or gradient clipping are needed in practice.
The environment weight and the adjusted weight are used for the methodAll judge based on the environmental label in the environmental label library, and amplify the magnification by the environmental weightObtaining the adjustment value of the environmental weight can effectively reduce the calculation amount, and further deepen the influence and association of environmental factors on the output of the risk judgment model.
The feature weights are often adjusted individually for each feature, and the loss weighting factors are the overall amplification or reduction of the prediction error for the whole sample, the "granularity" of the two weights are different, the coverage is also different, so that the two weights cannot be multiplexed into one value, and for real-time environment and vehicle data, there are multiple differencesIs generally chosen relative to the problem ofWith the greatest adjustment amplitudeTo calculate the environmental weights becauseThe larger the variation amplitude is, the more severe the corresponding environmental condition is, and when the variation amplitude is used for calculating the loss function, the punishment of extreme weather prediction errors can be increased.
Moreover, the loss function is optimized through the environment weighting factors, so that the penalty coefficient of the risk judgment model is larger when the extreme weather prediction deviates, and the penalty coefficient is smaller when the good weather deviates, and the prediction capacity for the extreme weather is enhanced. For example, when the environment is extremely dangerousTake 2) and true risk score = high risk but predicted as low risk1.5), Then:
In the loss function, the error of the sample is amplified by 3 times, and the model can obviously adjust the related weight in the back propagation process so as to quickly correct the serious missing report. When the environment is relatively safe and the situation is misreported (the real non-high risk, the model gives a higher risk score), if only a smaller punishment coefficient (such as 1.2) is set, the whole magnification is not excessive.
Still further, the updating the risk determination model in real time by incremental training includes setting a fixed time window according to the real-time environment and the vehicle data, each time window passingAn incremental training is triggered. The time windowFor a short time window, for example, the last five minutes or ten minutes, the latest conditions of road conditions, weather and the like are obtained, the model is subjected to incremental training, and the prediction accuracy of the risk judgment model for short-time risk factor changes is improved.
Window the timeThe real-time environment and the vehicle data in the model are mapped to an environment tag library, environment tags of the real-time environment and the vehicle data are obtained, the correlation degree between the current data and the hidden layer module of the risk judgment model is determined based on the environment tags, only the module with high correlation degree is updated, and the learning rate of the module with low correlation degree is set to be zero.
Specifically, if the batch of data is found to be mainly related to weather changes, such as sudden heavy rain, higher updating priority is only given to modules which are highly related to weather determination, modules which are irrelevant or weakly related to traffic flow are not updated, and if certain novel characteristic combinations (such as night+rain and snow+curve+large flow) appear in the batch of data, higher learning rate is set on the connection weights of the corresponding modules, so that the model is more quickly adapted to new situations.
If the presence of continuity is detectedAnd performing incremental training on the non-updated modules, and not performing correlation judgment to update the corresponding modules by using the data in the time window.
And after each updating is completed, recording the prediction precision of the risk judgment model, and adjusting the learning rate of the risk judgment model module according to the prediction precision. After the system completes one small batch of incremental training, the system deploys new parameters to an online inference module and observes the predicted performance of the new parameters on actual data in the next time period. If the prediction effect (such as accuracy, error rate, etc.) of the model remains steadily elevated or maintained at a high level after several consecutive incremental updates, it is indicated that the local update is effective. When some modules of the model are found to have errors increased due to too frequent or improper updating, the system can actively reduce the learning rate of the module or temporarily freeze the updating of the module to allow other modules to compensate, and can unfreeze the module when the environment changes for a new time so as to prevent excessive interference to the learned knowledge in normal situations.
Setting a global update periodSummarizing all time windows in a period every time a global update period passesTraining and updating the whole risk judgment model, and unifying weight association among the calibration modules. The purpose of global fine tuning is to further correct long-term local accumulated errors and unify weight correlations among calibration modules while ensuring adaptation to the latest scene. When new large-scale historical traffic data or special scene data (such as severe weather and congestion data of special holidays) are available, offline supplementary training is performed again to strengthen the overall generalization performance of the model, and then the model is deployed in an online system again.
The invention adopts the mode of 'environment matching + priority identification + local training', effectively reduces disturbance to irrelevant modules, improves learning rate or updating frequency on key modules, and remarkably enhances pertinence and efficiency of real-time early warning of high-risk road sections.
Secondly, some existing incremental learning technologies do not have special global fine tuning links after updating small batches of data, and the invention can integrate more real traffic scene data in a large-scale time range by periodically carrying out centralized examination and retraining on incremental updating results, comprehensively correct network parameters and give consideration to short-term response and long-term optimization. Specifically, the incremental update of the short-term window corresponds to sudden changes of weather, road conditions and the like, and the global update corresponds to changes of seasons, periodic traffic and the like, and the combination of the two ensures the prediction accuracy of the risk judgment model.
And S3, triggering grading early warning according to the comprehensive risk score, carrying out early warning on a vehicle end, and visually displaying the early warning result.
And triggering grading early warning according to the comprehensive risk score, carrying out early warning through a vehicle-mounted navigation system at the vehicle end, and visually displaying the early warning result. In the invention, on the basis of the comprehensive risk score generated by the risk judgment model, the score is compared with the preset risk threshold value, so that the potential risk degree of the current road section or the vehicle driving scene is primarily divided, and four grades of no risk, low risk, high risk and extremely high risk are formed. By combining the real-time environment, the vehicle data and the environment labels, the comprehensive risk score can be further corrected or verified, and the specific extreme environments (such as heavy rain and strong wind areas at night) can be additionally screened to form more accurate early warning results. And then, the early warning result and the corresponding environment label are visually displayed together, so that a driver, road side equipment or a management platform can conveniently and rapidly identify dangerous conditions.
Specifically, when the risk level is risk-free, highway high-risk road section early warning is not performed, but the system still records related data for subsequent incremental training and model optimization.
If the risk level is low risk, the driver is warned in advance through the vehicle-mounted display and the voice prompt, and a speed limit warning can be sent out on the roadside information board at the same time to remind other vehicles entering the road section of keeping a safe vehicle distance and a proper vehicle speed.
When the risk level is increased to high risk, the system is further linked with the vehicle navigation on the basis of low risk early warning measures, and the system sends a warning of high risk of a front road section to a driver in advance through comprehensive analysis of the vehicle destination and road section information and provides a suggestion of re-planning a route or decelerating to drive.
If the comprehensive risk score and the environment label jointly indicate that the risk level is extremely high, a more severe coping strategy is started on the basis of a high-risk early warning scheme, including implementing temporary traffic control such as road section sealing and the like, and synchronously notifying related emergency or maintenance departments to prepare corresponding risk treatment measures so as to reduce the possibility of accident occurrence and expansion to the greatest extent.
In the process, the vehicle-mounted navigation system can dynamically update the navigation path or the speed limiting scheme by combining the model output and the records in the environment tag library, so that the vehicle can avoid dangerous areas as much as possible or take corresponding relieving measures in a severe environment. When the management center or the road side equipment receives the extremely high risk information pushed by the system, the management center or the road side equipment can also execute the corresponding processes of warning broadcasting, instruction guiding, personnel evacuation and the like in cooperation with the actual situation of the site. In the whole, through multistage early warning and visualization means, drivers and traffic management personnel can clearly grasp specific risk conditions of highway high-risk road sections, and corresponding protection or shunt actions are adopted in advance, so that high-efficiency management and optimization of highway traffic safety are realized.
Embodiment 2, in an exemplary embodiment, also provides a highway high-risk road section early warning system, which comprises an acquisition module, a processing module and an early warning module.
The acquisition module is connected with the processing module, the acquisition module uploads real-time environment and vehicle data acquired by the vehicle end and the road side sensor to the cloud platform through a wired or wireless network, and the processing module performs fusion, cleaning and formatting operation on the data after receiving the data and performs comprehensive analysis according to the risk judgment model.
The processing module is connected with the acquisition module and the early warning module, updates the risk judgment model in real time by utilizing incremental training, outputs the comprehensive risk score, and transmits the comprehensive risk score and corresponding risk information to the early warning module so that the later triggers hierarchical early warning and performs visual display according to the comprehensive risk score and the corresponding risk information.
The early warning module is connected with the vehicle-mounted/road side terminal, decides whether early warning or higher level management and control is carried out according to the level of the comprehensive risk score, and sends early warning instructions or visual information to the vehicle-mounted navigation system, the vehicle-mounted display terminal or the road side information board. In extremely high risk situations, higher level response measures such as road segment closure instructions may also be sent to the management center or emergency department.
Through the transmission of the data and the instructions, the three modules form a closed loop system flow from bottom to top (data acquisition, data processing, risk judgment and early warning execution), and the road safety monitoring efficiency is continuously improved on the basis of real-time iteration and modularized updating.
Still further, the acquisition module is deployed at the vehicle end and the road side end of the high-risk road section, and is used for acquiring multi-source real-time data, and transmitting the multi-source real-time data to the cloud platform after preliminary integration.
The vehicle-mounted intelligent monitoring system comprises a vehicle-mounted radar, a camera, a GPS, an accelerometer, a steering angle sensor and the like, wherein the vehicle-mounted intelligent monitoring system comprises a vehicle-mounted radar, a road side, a road platform and a processing module, wherein the vehicle-mounted radar, the camera, the GPS, the accelerometer, the steering angle sensor and the like can be used for acquiring the running state and surrounding environment information of the vehicle, the road side is used for detecting road ponding, icing, terrain change and the like by arranging an intelligent monitoring camera, a laser radar, a meteorological sensor, a road side radar and other equipment for monitoring road surface conditions on a high-risk road section, and the multi-source data are transmitted to the cloud platform in a wired (such as optical fiber) or wireless (such as V2X) mode so as to provide comprehensive and accurate real-time data support for the processing module.
The processing module runs in the cloud platform and mainly completes data fusion, model training update and risk score calculation, and is a core analysis center of the whole system.
Specifically, the multi-source real-time data uploaded by the acquisition module is subjected to operations such as formatting, noise filtering, time stamp alignment and the like, and the processed data is used as input of a risk judgment model;
establishing a network model comprising an input layer, a hidden layer and an output layer, and updating model parameters in real time through incremental training;
Setting a fixed time window according to the acquired real-time environment and vehicle data, triggering small batch incremental training when each time window arrives, and focusing on extreme environments or serious errors by using mechanisms such as environment weighted mean square error and the like;
and digitally quantifying the potential risk degree of the current road section, and sending the result to the early warning module.
And the early warning module triggers early warning measures of corresponding grades according to the comprehensive risk scores and the real-time environment information output by the processing module, and visually presents the risk data to a driver or a traffic management department.
Specifically, the comprehensive risk score is compared with a preset risk threshold value, and four grades of no risk, low risk, high risk and extremely high risk are automatically judged;
the road section in a high risk or extremely high risk state is pre-warned and prompted in advance through a vehicle navigation system, and a vehicle can be guided to re-plan a driving path if necessary;
The risk data, the environment label and other information are presented on a vehicle-mounted display screen, a road side information board or a background management platform interface, so that a driver and a manager can intuitively know the risk condition;
If the risk level reaches an extremely high level, the early warning module can also send warning information to a highway management center or an emergency department so as to coordinate and execute higher-level coping schemes such as road section sealing, personnel evacuation and the like.
Through the tight linkage among the three functional modules, the high-risk road section monitoring system can monitor, accurately judge and dynamically pre-warn the high-risk road section in real time, continuously improve the prediction precision through continuous incremental training of the cloud model on the basis of collecting rich multi-source data, and timely transmit potential danger to related parties by means of a grading pre-warning mechanism, so that the high-risk road section high-efficiency safety management is realized.
The above functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. The storage medium includes a U disk, a removable hard disk, a Read-only memory (ROM), a random access memory (RAM, randomAccessMemory), a magnetic disk, an optical disk, or other various media capable of storing program codes.
Logic and/or steps represented in the flowcharts or otherwise described herein, e.g., a ordered listing of executable instructions for implementing logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.
More specific examples (a non-exhaustive list) of the computer-readable medium include an electrical connection (an electronic device) having one or more wires, a portable computer diskette (a magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). In addition, the computer readable medium may even be paper or other suitable medium on which the program is printed, as the program may be electronically captured, via, for instance, optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner, if necessary, and then stored in a computer memory.
It is to be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above-described embodiments, the various steps or methods may be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, may be implemented using any one or combination of techniques known in the art, discrete logic circuits with logic gates for implementing logic functions on data signals, application specific integrated circuits with appropriate combinational logic gates, programmable Gate Arrays (PGAs), field Programmable Gate Arrays (FPGAs), and the like.
It should be noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention may be modified or substituted without departing from the spirit and scope of the technical solution of the present invention, which is intended to be covered in the scope of the claims of the present invention.
Claims (8)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510192505.0A CN119694132A (en) | 2025-02-21 | 2025-02-21 | A highway high-risk section early warning method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510192505.0A CN119694132A (en) | 2025-02-21 | 2025-02-21 | A highway high-risk section early warning method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN119694132A true CN119694132A (en) | 2025-03-25 |
Family
ID=95035463
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202510192505.0A Pending CN119694132A (en) | 2025-02-21 | 2025-02-21 | A highway high-risk section early warning method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN119694132A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119989003A (en) * | 2025-04-15 | 2025-05-13 | 杭州诚道科技股份有限公司 | Traffic business management big data comprehensive analysis and judgment system |
| CN120012025A (en) * | 2025-04-16 | 2025-05-16 | 济南瑞源智能城市开发有限公司 | A road surface perception method, device and medium based on multimodal data |
| CN120220421A (en) * | 2025-05-07 | 2025-06-27 | 南京码极客科技有限公司 | Traffic incident dynamic risk prediction method, system, electronic device and storage medium |
| CN120580806A (en) * | 2025-06-19 | 2025-09-02 | 重交智安(浙江)科技发展有限公司 | A sensor optical cable guardrail safety alarm device for renovation and expansion projects |
| CN120873997A (en) * | 2025-09-29 | 2025-10-31 | 甘肃省交通规划勘察设计院股份有限公司 | Intelligent recognition method for continuous longitudinal slope traffic risk factors based on deep learning |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118470991A (en) * | 2024-06-14 | 2024-08-09 | 中建照明有限公司 | Intelligent traffic control system and method for smart city based on 5G |
| CN118644069A (en) * | 2024-04-24 | 2024-09-13 | 清华大学深圳国际研究生院 | A method and system for hierarchical management of heterogeneous terminals in an intelligent transportation system |
| CN119091616A (en) * | 2024-08-13 | 2024-12-06 | 深圳市孪生云计算技术有限公司 | A road traffic accident detection method and system based on roadside radar perception |
-
2025
- 2025-02-21 CN CN202510192505.0A patent/CN119694132A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118644069A (en) * | 2024-04-24 | 2024-09-13 | 清华大学深圳国际研究生院 | A method and system for hierarchical management of heterogeneous terminals in an intelligent transportation system |
| CN118470991A (en) * | 2024-06-14 | 2024-08-09 | 中建照明有限公司 | Intelligent traffic control system and method for smart city based on 5G |
| CN119091616A (en) * | 2024-08-13 | 2024-12-06 | 深圳市孪生云计算技术有限公司 | A road traffic accident detection method and system based on roadside radar perception |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119989003A (en) * | 2025-04-15 | 2025-05-13 | 杭州诚道科技股份有限公司 | Traffic business management big data comprehensive analysis and judgment system |
| CN120012025A (en) * | 2025-04-16 | 2025-05-16 | 济南瑞源智能城市开发有限公司 | A road surface perception method, device and medium based on multimodal data |
| CN120220421A (en) * | 2025-05-07 | 2025-06-27 | 南京码极客科技有限公司 | Traffic incident dynamic risk prediction method, system, electronic device and storage medium |
| CN120580806A (en) * | 2025-06-19 | 2025-09-02 | 重交智安(浙江)科技发展有限公司 | A sensor optical cable guardrail safety alarm device for renovation and expansion projects |
| CN120873997A (en) * | 2025-09-29 | 2025-10-31 | 甘肃省交通规划勘察设计院股份有限公司 | Intelligent recognition method for continuous longitudinal slope traffic risk factors based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11176845B2 (en) | Adaptive analysis of driver behavior | |
| CN119694132A (en) | A highway high-risk section early warning method and system | |
| US9898936B2 (en) | Recording, monitoring, and analyzing driver behavior | |
| US11170639B2 (en) | Transportation threat detection system | |
| KR102477885B1 (en) | Safety analysis management server for evaluating autonomous driving roads | |
| JP2020013537A (en) | Road surface condition estimation device and road surface condition estimation method | |
| CN111465972B (en) | System for calculating error probability of vehicle sensor data | |
| CN117238139B (en) | Real-time road condition early warning system based on meteorological data | |
| US20250209911A1 (en) | Systems and methods for training and or using machine learning models to detect traffic accidents on roads | |
| KR102709531B1 (en) | Video analysis based road risk control device and method thereof | |
| CN118366312B (en) | Traffic detection system and method | |
| CN120431727A (en) | Intelligent highway dispatching system based on air-ground collaborative data | |
| CN120653945B (en) | AI-based autonomous driving evaluation system | |
| Sundararajan et al. | Vehicle automation and weather: challenges and opportunities. | |
| CN119024381B (en) | Real-time travel path track big data analysis processing method based on Beidou positioning | |
| CN119672954B (en) | Expressway driving risk prediction method and system based on management and control unit | |
| CN115257812A (en) | Automatic driving confidence obtaining and reminding method and equipment | |
| US20250035449A1 (en) | Method and system of lane-level road congestion identification and forecasting for navigation application | |
| CN116935641B (en) | A method, device and control equipment for road condition risk early warning | |
| Ali et al. | Employment of instrumented vehicles to identify real-time snowy weather conditions on freeways using supervised machine learning techniques–A naturalistic driving study | |
| US12235119B2 (en) | Methods and apparatuses for late lane change prediction and mitigation | |
| CN120636166B (en) | A traffic emergency big data analysis system and data fusion method | |
| CN121849172A (en) | Performance determination method, device, equipment and medium for intelligent pilot function | |
| CN121747320A (en) | Intelligent driving environment abnormal state diagnosis method and system based on vehicle cloud cooperation | |
| CN119964373A (en) | A highway traffic monitoring system and method based on the Internet |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20250325 |
|
| RJ01 | Rejection of invention patent application after publication |