Multi-dimensional traffic safety anomaly analysis and early warning system, method and program product

Abstract

The invention relates to the technical field of traffic safety early warning, in particular to a multi-dimensional traffic safety anomaly analysis and early warning system and method and a program product. The server side comprises a multi-dimensional data acquisition module, a data preprocessing module, a feature extraction module, an anomaly detection module, a behavior analysis module, a risk assessment module and a risk assessment module, and the risk assessment module is used for analyzing the abnormal data of the road environment and the vehicle behavior analysis data according to the output abnormal data of the road environment and the vehicle behavior analysis data. The position information corresponding to the abnormal position on the current road is identified, the risk level corresponding to the current road is determined based on a preset risk assessment strategy, and a corresponding early warning strategy is formulated based on the corresponding risk level; and the sending module is used for sending a corresponding early warning signal to the mobile device end in the preset range of the abnormal position according to the formulated early warning strategy.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic safety early warning, and in particular to a multi-dimensional traffic safety abnormality analysis early warning system, method and program product. Background Art

[0002] With the rapid development of intelligent transportation systems, navigation systems have become an indispensable part of modern vehicles. Existing navigation systems usually provide route planning, real-time traffic updates and other functions by integrating multiple data sources to help drivers avoid congestion and dangerous areas. However, when faced with sudden abnormal road conditions, the warning mechanism of existing navigation systems still has some limitations.

[0003] Currently, when the navigation system detects that there may be an abnormality in the road environment, it usually makes a preliminary judgment by asking the user what kind of abnormality is happening ahead. This method relies on the user's immediate feedback and needs to collect multiple feedbacks to determine whether an early warning is needed. This method has the following problems in emergency situations:

[0004] In an emergency road accident, the driver may not be able to provide accurate information feedback in a timely manner, especially when an accident has already occurred, information feedback is almost impossible, resulting in the problem of delayed information feedback.

[0005] In order to receive warning information ahead, vehicles need to install professional assisted driving systems or dedicated on-board terminals and other equipment, which not only increases costs, but also has a low penetration rate, reducing the effectiveness of the warning.

[0006] Due to its reliance on user feedback and multiple confirmations, warning information often has a high delay and is unable to respond promptly and effectively to emergencies such as heavy fog, bridge and tunnel collapse, icy and slippery road surfaces, and vehicle accidents.

[0007] Therefore, there is an urgent need for a multi-dimensional traffic safety anomaly analysis and warning system, method and program product that can automatically detect road environment anomalies without user feedback, and send warning information to vehicles in a timely and accurate manner, thereby improving the accuracy of road environment anomaly detection and the timeliness of warning, and reducing the cost of road environment anomaly detection. Summary of the invention

[0008] One of the purposes of the present invention is to provide a multi-dimensional traffic safety anomaly analysis and warning system, method and program product, which can automatically detect road environment anomalies without user feedback, and send warning information to vehicles in a timely and accurate manner, thereby improving the accuracy of road environment anomaly detection and the timeliness of warning, and reducing the cost of road environment anomaly detection.

[0009] In order to achieve the above objectives, a multi-dimensional traffic safety anomaly analysis and early warning system is provided, including a server and a mobile device;

[0010] The server includes:

[0011] A multi-dimensional data acquisition module is used to obtain in real time the vehicle driving data information collected by each mobile device on the current road, wherein the vehicle driving data information includes vehicle position information and vehicle motion state data collected by an accelerometer and a gyroscope on the vehicle, wherein the vehicle motion state data includes vehicle speed, vehicle acceleration and vehicle driving direction;

[0012] A data preprocessing module is used to perform preprocessing operations on vehicle driving data acquired in real time;

[0013] A feature extraction module, for extracting key features from the vehicle driving data according to the vehicle driving data after the preprocessing operation;

[0014] An anomaly detection module is used to input the extracted key features as input data into the road anomaly detection model based on a preset road anomaly detection model, and output corresponding road environment anomaly data, wherein the road environment anomaly data includes road anomaly conditions at various locations of the current road;

[0015] The behavior analysis module is used to analyze the vehicle behavior of each vehicle on the road according to the extracted key features and based on the pre-set behavior detection strategy to form corresponding vehicle behavior analysis data;

[0016] The risk assessment module is used to identify the location information corresponding to the abnormal location on the current road based on the output road environment abnormality data and vehicle behavior analysis data, and determine the risk level corresponding to the current road based on the preset risk assessment strategy, and formulate a corresponding early warning strategy based on the corresponding risk level;

[0017] The sending module is used to send a corresponding warning signal to a mobile device within a preset range of the location where the abnormality exists according to the formulated warning strategy.

[0018] Technical principles and effects of this solution: In this solution, first, the vehicle driving data information collected by the mobile device on the road is acquired in real time based on the current road. Specifically, the vehicle driving data information includes vehicle position information, and vehicle motion state data collected by the accelerometer and gyroscope on the vehicle, where the vehicle motion state data includes vehicle speed, vehicle acceleration, and vehicle driving direction. The vehicle driving data is collected by the mobile device without the need to install special equipment, which greatly reduces the cost of road environment anomaly detection.

[0019] Then, the acquired vehicle driving data information is preprocessed, and the preprocessing operation makes the vehicle driving data information more effective and greatly improves the anti-interference ability of the vehicle driving data information.

[0020] After that, the key features of the vehicle driving data after the preprocessing operation can be extracted, and the extracted key features can be input into the pre-set road anomaly detection model to realize the output of road environment anomaly data, and then the road anomaly conditions at various locations on the current road can be known.

[0021] Of course, at the same time, according to the key features, the vehicle behavior of each vehicle on the road can also be analyzed based on the preset behavior detection strategy, thereby realizing the formation of vehicle behavior analysis data.

[0022] Then, based on the road abnormalities at various locations on the current road and the vehicle behaviors of each vehicle on the road, namely the road environment abnormality data and vehicle behavior analysis data, the location information corresponding to the abnormal locations on the current road can be identified. After that, the risk level of the current road can be determined and evaluated according to the preset risk assessment strategy, so as to formulate the corresponding early warning strategy, and send a warning signal to the mobile device within the preset range of the abnormal location as soon as possible, thereby realizing the corresponding traffic abnormality alarm and improving the driving safety of vehicles on the road.

[0023] 1. The early data collection of the vehicle is to collect data of multiple dimensions including vehicle location information and vehicle motion status data collected by the accelerometer and gyroscope on the vehicle. The vehicle motion status data includes vehicle speed, vehicle acceleration and vehicle driving direction, which realizes the collection and integration of multi-dimensional data of the vehicle, improves more comprehensive vehicle motion status and road environment information, and provides reliable data support for subsequent early warnings. At the same time, the vehicle's own sensors are used to collect vehicle driving data information, which greatly reduces the cost of early data collection for road environment anomaly detection.

[0024] 2. Through the setting of road anomaly detection model and corresponding vehicle behavior analysis strategy, rapid and accurate detection of road environmental conditions and vehicle behavior is achieved, thereby realizing rapid detection and rapid warning of road anomalies, and automatically detecting road environmental anomalies without user feedback, and sending warning information to vehicles in a timely and accurate manner, thereby improving the accuracy of road environmental anomaly detection and the timeliness of warning.

[0025] Furthermore, the server also includes:

[0026] A historical data acquisition module is used to acquire historical data related to the current road, wherein the historical data related to the current road includes historical key features and historical road abnormalities;

[0027] Model building module, used to build a road anomaly detection model;

[0028] A training module, used to divide the acquired historical related data into a first training set and a first validation set, and input the historical key features in the first training set as input data into the road anomaly detection model, and use the historical road anomaly conditions in the first training set as output data of the road anomaly detection model to train the road anomaly detection model;

[0029] A verification module is used to, after each training is completed, input the historical key features in the first verification set as input data into the trained road anomaly detection model, output the corresponding first prediction result, and calculate the prediction accuracy corresponding to the road anomaly detection model after this training based on the first prediction result and the historical road anomaly corresponding to the historical key features in the first verification set, based on a preset prediction accuracy calculation formula, and determine the relationship between the prediction accuracy and the preset accuracy threshold. If the prediction accuracy is greater than or equal to the preset accuracy threshold, the road anomaly detection model after this training is determined to be a valid model; otherwise, it is determined to be an invalid model, and the road anomaly detection model continues to be trained;

[0030] The prediction accuracy calculation formula is:

[0031]

[0032] Wherein, YZ is the prediction accuracy, S is the total number of historical road anomalies in the first verification set, s is the number of historical anomalies that are the same as the first prediction result in the first verification set, and α is the corresponding weighting coefficient. The size of α is related to the number of training times. The smaller the number of training times, the smaller the value corresponding to α.

[0033] Beneficial effects: In this scheme, by using a smaller α in the early training stage, it is possible to converge to a preliminary feasible model more quickly, reducing unnecessary training time and consumption of computing resources. As the number of training times increases, the value of α gradually increases, which can prompt the model to gradually improve its accuracy, thereby achieving high accuracy and high standards of the model. It can also prevent the accuracy threshold from being reached too early and stopping training, which helps to train a better and more reliable model.

[0034] Furthermore, the preset behavior detection strategy includes the following steps:

[0035] The extracted key features are input into a vehicle behavior detection model constructed based on a preset model building strategy, and the corresponding vehicle behavior analysis data is output;

[0036] The pre-set model building strategy includes:

[0037] Acquire historical vehicle data of the current road, wherein the historical vehicle data includes historical key features of the current road and corresponding historical vehicle behavior analysis data;

[0038] Build a vehicle behavior detection model;

[0039] The acquired historical vehicle data is divided into a second training set and a second validation set, and the historical key features in the second training set are input into the constructed vehicle behavior detection model. The historical vehicle behavior analysis data in the second training set is used as input data of the vehicle behavior detection model to train the vehicle behavior detection model.

[0040] After each training is completed, the historical key features in the second verification set are input into the trained vehicle behavior detection model, and the corresponding second prediction result is output. Based on the second prediction result and the corresponding historical vehicle behavior analysis data in the second verification set, it is determined whether the corresponding vehicle behavior detection model is a valid model. If it is a valid model, the training of the vehicle behavior detection model is completed. Otherwise, the training of the vehicle behavior detection model continues.

[0041] Further, the risk assessment strategy includes the following steps:

[0042] According to the output road environment abnormality data and vehicle behavior analysis data, identify the location information corresponding to the location where the current road has abnormality, the type of environmental abnormality corresponding to each location information, and the type of vehicle behavior abnormality corresponding to the vehicle at each location;

[0043] According to the environmental anomaly type corresponding to each location information and the vehicle behavior anomaly type corresponding to the vehicle at each location, the risk value of the corresponding environmental anomaly type and the risk value of the vehicle behavior anomaly type are matched, and the comprehensive risk value at the corresponding location is calculated based on the preset comprehensive risk value calculation formula, and the corresponding risk level is matched through the risk value based on the level matching strategy;

[0044] The risk value calculation formula is:

[0045]

[0046] In the formula, FX is the comprehensive risk value at the corresponding position, H is the risk value corresponding to the abnormal environment type at the corresponding position, and Y is the risk value corresponding to the abnormal vehicle behavior at the corresponding position. μ is the corresponding weight value;

[0047] The level matching strategy is as follows:

[0048] When 0 < FX ≤ X, the corresponding risk level is the first level;

[0049] When X < FX ≤ Z, the corresponding risk level is the second level;

[0050] When Z < FX ≤ R, the corresponding risk level is the third level;

[0051] When FX > R, the corresponding risk level is the fourth level; the X, Z, and R are dynamically changing interval points, and the dynamically changing interval points are associated with the number of abnormal positions on the current road.

[0052] Beneficial effects: In this solution, by comprehensively considering the risk values of environmental anomaly types and vehicle behavior anomaly types to calculate the comprehensive risk value, the risk level corresponding to the road can be more accurately evaluated and determined. At the same time, when matching the UI level, dynamically changing interval points are set, and the division of the risk level can be dynamically adjusted according to the number of abnormal positions on the road, making the evaluation more refined.

[0053] Furthermore, the server further includes:

[0054] A road division module, configured to divide the current road into a plurality of data collection points based on a preset distance interval according to the basic path of the current road;

[0055] An optimization module, configured to select an optimal collection point set of the current road based on an improved optimization algorithm according to a plurality of data collection points of the current road and historical related data;

[0056] The multi-dimensional data collection module is further configured to select vehicle driving data information with the same position information as each data collection point in the optimal collection point set of the current road from the obtained vehicle driving data information and input it into the data preprocessing module.

[0057] Beneficial effects: In this solution, through the settings of the road division module and the optimization module, the reasonable division of the road and the determination of the optimal data collection points are realized. By selecting the optimal collection point set of the current road to select the vehicle driving data information on the current road, and using the vehicle driving data information at these data collection points for subsequent early warning and anomaly analysis, the data processing efficiency is greatly improved.

[0058] Furthermore, the optimization module includes:

[0059] The initial screening module is used to randomly generate N initial populations according to several data collection points of the current road. The individuals of the initial population are the sets formed by the data collection points. Based on the preset restriction conditions, the individuals of the initial population are screened and judged. If the conditions are met, the corresponding initial population is a feasible solution. Otherwise, it is an infeasible solution and the corresponding initial population is eliminated.

[0060] The restrictions are:

[0061]

[0062] Where M is the total number of data collection bags, M max is the maximum allowed quantity, ρ k For a specific location R k The density of surrounding data collection points, ρ mix is the minimum permissible density;

[0063] A fitness calculation module, used to calculate the first fitness and the second fitness of the selected population;

[0064] The first fitness is calculated as follows:

[0065]

[0066] In the formula, F1 is the first fitness, D1 is the total value of the data collection points corresponding to the population, and x i is the vehicle passing rate of the i-th data collection point in the population, c i is the number of historical vehicles passing through the i-th data collection point in the population, C is the total number of historical vehicles passing through, w1, w2 are the corresponding weighting coefficients, B is a random number between 0 and 1, b y is a random threshold;

[0067] The second fitness is calculated as follows:

[0068]

[0069] In the formula, F2 is the second fitness, D2 is the distance difference between data collection points in the population, and d pq is the distance between the qth data collection point and the pth data collection point;

[0070] A selection module is used to select a population whose first fitness is less than or equal to a preset first fitness threshold within a first number of iterations;

[0071] When the first number of iterations is exceeded, selecting a population whose second fitness is less than or equal to a preset second fitness threshold;

[0072] The crossover mutation module is used to cross and mutate the selected population to obtain the offspring population;

[0073] A loop module is used to continue executing the fitness calculation module after obtaining the offspring population until a preset number of iterations is met;

[0074] The output module is used to output the offspring population as the optimal solution set for multi-objective optimization.

[0075] Beneficial effects: In this solution, multiple indicators can be considered at the same time through multi-objective optimization, achieving higher quality data collection points. When identifying anomalies, it is not necessary to use all data, and accurate anomaly warning and judgment can be achieved. At the same time, by setting random numbers, the algorithm can be prevented from converging too quickly and falling into the problem of local optimal solution.

[0076] Furthermore, the server also includes:

[0077] The precise module is used to determine the data collection point corresponding to the position information based on the position information corresponding to the position where the abnormality exists on the current road identified when the determined risk level is the second level, the third level or the fourth level, and select the vehicle driving data information within the preset range where the corresponding data collection point is located from the vehicle driving data information obtained by the multi-dimensional data acquisition module, and determine the position corresponding to the initial point position where the corresponding data collection point has an abnormality based on the vehicle driving data information within the range where the corresponding data collection point is located, and upload it to the server, wherein the initial point position is the initial position corresponding to the abnormality of the data collection point.

[0078] Beneficial effect: In this solution, a preliminary abnormality judgment is first made on the optimal collection point set. After the abnormal location is determined, a more accurate positioning is achieved through the setting of the precise module, that is, the surrounding data is judged based on the data collection point corresponding to the abnormal location, so as to identify the initial point that caused the abnormality of the data collection point, so that the operator on the server can better and more comprehensively understand the abnormal situation.

[0079] The present invention also provides a multi-dimensional traffic safety anomaly analysis and early warning method, using the above-mentioned multi-dimensional traffic safety anomaly analysis and early warning system.

[0080] The present invention also provides a multi-dimensional traffic safety anomaly analysis and early warning program product, which includes a computer program. When the computer program is executed, the above-mentioned multi-dimensional traffic safety anomaly analysis and early warning system is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] Figure 1This is a logic block diagram of a multi-dimensional traffic safety anomaly analysis and early warning system in Embodiment 1 of the present invention;

[0082] Figure 2 This is another logical block diagram of the multi-dimensional traffic safety anomaly analysis and early warning system in the first embodiment of the present invention. DETAILED DESCRIPTION

[0083] The following is further described in detail through specific implementation methods:

[0084] Embodiment 1

[0085] A multi-dimensional traffic safety abnormality analysis and early warning system, basically Figure 1 and Figure 2 As shown, it includes a server side and a mobile device side;

[0086] The server includes:

[0087] The multi-dimensional data acquisition module is used to obtain in real time the vehicle driving data information collected by each mobile device on the current road, the vehicle driving data information includes the vehicle position information, and the vehicle motion state data collected by the accelerometer and gyroscope on the vehicle, the vehicle motion state data includes the vehicle speed, vehicle acceleration and vehicle driving direction; in this embodiment, the monitoring video data on the vehicle is also acquired.

[0088] A road division module is used to divide the current road into a number of data collection points based on the basic path of the current road and the preset distance interval;

[0089] The optimization module is used to select the optimal collection point set of the current road based on several data collection points of the current road and historical related data based on the improved optimization algorithm;

[0090] The optimization module includes:

[0091] The initial screening module is used to randomly generate N initial populations according to several data collection points of the current road. The individuals of the initial population are the sets formed by the data collection points. Based on the preset restriction conditions, the individuals of the initial population are screened and judged. If the conditions are met, the corresponding initial population is a feasible solution. Otherwise, it is an infeasible solution and the corresponding initial population is eliminated.

[0092] The restrictions are:

[0093]

[0094] Where M is the total number of data collection bags, M max is the maximum allowed quantity, ρ k For a specific position P kThe density of surrounding data collection points, ρ mix is the minimum permissible density;

[0095] A fitness calculation module, used to calculate the first fitness and the second fitness of the selected population;

[0096] The first fitness is calculated as follows:

[0097]

[0098] In the formula, F1 is the first fitness, D1 is the total value of the data collection points corresponding to the population, and x i is the vehicle passing rate of the i-th data collection point in the population, c i is the number of historical vehicles passing through the i-th data collection point in the population, C is the total number of historical vehicles passing through, w1, w2 are the corresponding weighting coefficients, B is a random number between 0 and 1, b y is a random threshold;

[0099] The second fitness is calculated as follows:

[0100]

[0101] In the formula, F2 is the second fitness, D2 is the distance difference between data collection points in the population, and d pq is the distance between the qth data collection point and the pth data collection point;

[0102] A selection module is used to select a population whose first fitness is less than or equal to a preset first fitness threshold within a first number of iterations;

[0103] When the first number of iterations is exceeded, selecting a population whose second fitness is less than or equal to a preset second fitness threshold;

[0104] The crossover mutation module is used to cross and mutate the selected population to obtain the offspring population;

[0105] A loop module is used to continue executing the fitness calculation module after obtaining the offspring population until a preset number of iterations is met;

[0106] The output module is used to output the offspring population as the optimal solution set for multi-objective optimization.

[0107] The multi-dimensional data acquisition module is also used to select vehicle driving data information whose vehicle position information is the same as the position information of each data collection point in the optimal collection point set from the acquired vehicle driving data information according to the selected optimal collection point set of the current road, and input it into the data preprocessing module.

[0108] The data preprocessing module is used to perform preprocessing operations on the vehicle driving data acquired in real time; in this embodiment, the preprocessing operations include filtering, denoising, and data synchronization.

[0109] A feature extraction module, for extracting key features from the vehicle driving data according to the vehicle driving data after the preprocessing operation;

[0110] An anomaly detection module is used to input the extracted key features as input data into the road anomaly detection model based on a preset road anomaly detection model, and output corresponding road environment anomaly data, wherein the road environment anomaly data includes road anomaly conditions at various locations of the current road;

[0111] The server also includes:

[0112] A historical data acquisition module is used to acquire historical data related to the current road, wherein the historical data related to the current road includes historical key features and historical road abnormalities;

[0113] The model building module is used to build a road anomaly detection model; the model can be built by methods such as SVM (Support Vector Machine), K-means Clustering, Isolation Forest and BP neural network. In this embodiment, the road anomaly detection model is built by BP neural network technology. First, a three-layer BP neural network model is built, including an output layer, a hidden layer and an input layer. In this embodiment, the obstacle position, movement trajectory, traffic flow, average speed, location information and acceleration change information are used as inputs of the input layer. Therefore, the input layer has a total of 6 nodes, and the output is the corresponding historical road anomaly, including traffic accidents, traffic jams, and natural disasters. Therefore, there are a total of 3 nodes. For the hidden layer, this embodiment uses the following formula to determine the number of hidden layer nodes: Where L is the number of nodes in the hidden layer, n is the number of nodes in the input layer, m is the number of nodes in the output layer, and u is a number between 1 and 10. In this embodiment, u is 4, so there are 7 nodes in the hidden layer. BP neural network usually uses Sigmoid differentiable function and linear function as the excitation function of the network. This paper selects the S-type tangent function tansig as the excitation function of the hidden layer neurons. The prediction model selects the S-type logarithmic function tansig as the excitation function of the output layer neurons.

[0114] A training module, used to divide the acquired historical related data into a first training set and a first validation set, and input the historical key features in the first training set as input data into the road anomaly detection model, and use the historical road anomaly conditions in the first training set as output data of the road anomaly detection model to train the road anomaly detection model;

[0115] A verification module is used to, after each training is completed, input the historical key features in the first verification set as input data into the trained road anomaly detection model, output the corresponding first prediction result, and calculate the prediction accuracy corresponding to the road anomaly detection model after this training based on the first prediction result and the historical road anomaly corresponding to the historical key features in the first verification set, based on a preset prediction accuracy calculation formula, and determine the relationship between the prediction accuracy and the preset accuracy threshold. If the prediction accuracy is greater than or equal to the preset accuracy threshold, the road anomaly detection model after this training is determined to be a valid model; otherwise, it is determined to be an invalid model, and the road anomaly detection model continues to be trained;

[0116] The prediction accuracy calculation formula is:

[0117]

[0118] Wherein, YZ is the prediction accuracy, S is the total number of historical road anomalies in the first verification set, s is the number of historical anomalies that are the same as the first prediction result in the first verification set, and α is the corresponding weighting coefficient. The size of α is related to the number of training times. The smaller the number of training times, the smaller the value corresponding to α.

[0119] The behavior analysis module is used to analyze the vehicle behavior of each vehicle on the road according to the extracted key features and based on the pre-set behavior detection strategy to form corresponding vehicle behavior analysis data;

[0120] The preset behavior detection strategy includes the following steps:

[0121] The extracted key features are input into a vehicle behavior detection model constructed based on a preset model building strategy, and the corresponding vehicle behavior analysis data is output;

[0122] The pre-set model building strategy includes:

[0123] Acquire historical vehicle data of the current road, wherein the historical vehicle data includes historical key features of the current road and corresponding historical vehicle behavior analysis data;

[0124] Construct a vehicle behavior detection model; The construction of the vehicle behavior detection model is mainly constructed using existing deep reinforcement learning and time series analysis techniques. In this embodiment, the time series analysis technology used is, specifically, a suitable time series model is selected, such as the autoregressive integrated moving average model (ARIMA), the hidden Markov model (HMM, this embodiment adopts the autoregressive integrated moving average model (ARIMA), and then the parameters of the model are fitted according to the corresponding historical vehicle data in the second training set, and then the fitted model is used to predict the corresponding second prediction result according to the historical key features in the second verification set, and then compared with the historical vehicle behavior analysis data in the second verification set, so as to realize the verification of the model.

[0125] The acquired historical vehicle data is divided into a second training set and a second validation set, and the historical key features in the second training set are input into the constructed vehicle behavior detection model. The historical vehicle behavior analysis data in the second training set is used as input data of the vehicle behavior detection model to train the vehicle behavior detection model.

[0126] After each training is completed, the historical key features in the second verification set are input into the trained vehicle behavior detection model, and the corresponding second prediction result is output. Based on the second prediction result and the corresponding historical vehicle behavior analysis data in the second verification set, it is determined whether the corresponding vehicle behavior detection model is a valid model. If it is a valid model, the training of the vehicle behavior detection model is completed. Otherwise, the training of the vehicle behavior detection model continues.

[0127] The risk assessment module is used to identify the location information corresponding to the abnormal location on the current road based on the output road environment abnormality data and vehicle behavior analysis data, and determine the risk level corresponding to the current road based on the preset risk assessment strategy, and formulate a corresponding early warning strategy based on the corresponding risk level; in this implementation, different risk levels correspond to different early warning strategies. For example, when the risk level is the first level, the corresponding early warning strategy is to display prompt information on the screen of the mobile device. When the risk level is the third level, multi-dimensional reminders are given through sound and vibration on the mobile device.

[0128] The risk assessment strategy includes the following steps:

[0129] According to the output road environment abnormality data and vehicle behavior analysis data, identify the location information corresponding to the location where the current road has abnormality, the type of environmental abnormality corresponding to each location information, and the type of vehicle behavior abnormality corresponding to the vehicle at each location;

[0130] Based on the environmental anomaly types corresponding to each location information point and the vehicle behavior anomaly types corresponding to the vehicles at each location, match the risk values of the corresponding environmental anomaly types and the risk values of the vehicle behavior anomaly types. Based on a preset comprehensive risk value calculation formula, calculate the comprehensive risk value at the corresponding location, and based on a level matching strategy, match the corresponding risk level through the risk value.

[0131] The risk value calculation formula is as follows:

[0132]

[0133] In the formula, FX is the comprehensive risk value at the corresponding location, H is the risk value corresponding to the environmental anomaly type at the corresponding location, Y is the risk value corresponding to the vehicle behavior anomaly at the corresponding location, μ is the corresponding weight value;

[0134] The level matching strategy is as follows:

[0135] When 0 < FX ≤ X, the corresponding risk level is the first level;

[0136] When X < FX ≤ Z, the corresponding risk level is the second level;

[0137] When Z < FX ≤ R, the corresponding risk level is the third level;

[0138] When FX > R, the corresponding risk level is the fourth level; X, Z, and R are dynamically changing interval points, and the dynamically changing interval points are associated with the number of abnormal positions on the current road.

[0139] An accurate module, used when the determined risk level is the second level, the third level, or the fourth level. Based on the location information corresponding to the abnormal positions identified on the current road, determine the data collection points corresponding to the location information, select the vehicle driving data information within the preset range where the corresponding data collection points are located from the vehicle driving data information obtained from the multi-dimensional data collection module, and based on the vehicle driving data information within the range where the corresponding data collection points are located, determine and upload to the server the location corresponding to the initial point where the corresponding data collection point is abnormal. The initial point is the initial position corresponding to the abnormality of the data collection point.

[0140] A sending module, used to send corresponding warning signals to the mobile device terminals within the preset range of the abnormal positions according to the formulated warning strategy.

[0141] This embodiment also discloses a multi-dimensional traffic safety anomaly analysis and warning method, which uses the above-mentioned multi-dimensional traffic safety anomaly analysis and warning system.

[0142] This embodiment also discloses a multi-dimensional traffic safety anomaly analysis and early warning program product, which includes a computer program. When the computer program is executed, the above-mentioned multi-dimensional traffic safety anomaly analysis and early warning system is implemented.

[0143] Embodiment 2, compared with embodiment 1, the difference of this embodiment is that: the server also includes:

[0144] The multi-source heterogeneous warning module is used to access other environmental warning systems and obtain the multi-source heterogeneous warning signals corresponding to each other environmental warning system, and send the corresponding multi-source heterogeneous warning signals to each mobile device based on the obtained multi-source heterogeneous warning signals. The other environmental warning systems include earthquake warning systems, weather forecast warning systems, and meteorological monitoring warning systems. At this time, each mobile device can visualize the corresponding multi-source heterogeneous signals through its own screen.

[0145] The multi-source heterogeneous warning module can access different environmental warning systems, such as earthquake warning systems, weather forecast warning systems, meteorological monitoring warning systems, etc., thereby integrating various types of warning information. By accessing other environmental warning systems and acquiring data, some dangers can be known in advance without starting the corresponding vehicle. This integrated method can provide comprehensive environmental warning services to help users obtain more comprehensive warning information.

[0146] By accessing these early warning systems in real time, we can ensure that the early warning information is updated and accurate. Users can get the latest early warning information at the first time and improve their ability to respond to emergencies. At the same time, when designing the system, we took into account the possibility of connecting to more types of early warning systems in the future, such as flood early warning systems, geological disaster early warning systems, etc. With the development of new technologies, new early warning sources and services can be easily added.

[0147] The above is only an embodiment of the present invention. The common sense such as the known specific structure and characteristics in the scheme is described too much here. The ordinary technicians in the relevant field know all the common technical knowledge in the technical field of the invention before the application date or priority date, can know all the existing technologies in the field, and have the ability to apply the conventional experimental means before that date. The ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the enlightenment given by this application. Some typical known structures or known methods should not become obstacles for ordinary technicians in the relevant field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can be made, which should also be regarded as the scope of protection of the present invention, which will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

Claims

1. A multi-dimensional traffic safety abnormality analysis and early warning system, characterized by: Including server side and mobile device side; The server includes: A multi-dimensional data acquisition module is used to obtain in real time the vehicle driving data information collected by each mobile device on the current road, wherein the vehicle driving data information includes vehicle position information and vehicle motion state data collected by an accelerometer and a gyroscope on the vehicle, wherein the vehicle motion state data includes vehicle speed, vehicle acceleration and vehicle driving direction; A data preprocessing module is used to perform preprocessing operations on vehicle driving data acquired in real time; A feature extraction module, for extracting key features from the vehicle driving data according to the vehicle driving data after the preprocessing operation; An anomaly detection module is used to input the extracted key features as input data into the road anomaly detection model based on a preset road anomaly detection model, and output corresponding road environment anomaly data, wherein the road environment anomaly data includes road anomaly conditions at various locations of the current road; The behavior analysis module is used to analyze the vehicle behavior of each vehicle on the road according to the extracted key features and based on the pre-set behavior detection strategy to form corresponding vehicle behavior analysis data; The risk assessment module is used to identify the location information corresponding to the abnormal location on the current road based on the output road environment abnormality data and vehicle behavior analysis data, and determine the risk level corresponding to the current road based on the preset risk assessment strategy, and formulate a corresponding early warning strategy based on the corresponding risk level; The sending module is used to send a corresponding warning signal to a mobile device within a preset range of the location where the abnormality exists according to the formulated warning strategy.

2. A multi-dimensional traffic safety abnormality analysis and early warning system according to claim 1, characterized in that: The server also includes: A historical data acquisition module is used to acquire historical data related to the current road, wherein the historical data related to the current road includes historical key features and historical road abnormalities; Model building module, used to build a road anomaly detection model; A training module, used to divide the acquired historical related data into a first training set and a first validation set, and input the historical key features in the first training set as input data into the road anomaly detection model, and use the historical road anomaly conditions in the first training set as output data of the road anomaly detection model to train the road anomaly detection model; A verification module is used to, after each training is completed, input the historical key features in the first verification set as input data into the trained road anomaly detection model, output the corresponding first prediction result, and calculate the prediction accuracy corresponding to the road anomaly detection model after this training based on the first prediction result and the historical road anomaly corresponding to the historical key features in the first verification set, based on a preset prediction accuracy calculation formula, and determine the relationship between the prediction accuracy and the preset accuracy threshold. If the prediction accuracy is greater than or equal to the preset accuracy threshold, the road anomaly detection model after this training is determined to be a valid model; otherwise, it is determined to be an invalid model, and the road anomaly detection model continues to be trained; The prediction accuracy calculation formula is: Wherein, YZ is the prediction accuracy, S is the total number of historical road anomalies in the first validation set, s is the number of the first prediction results that are the same as the historical anomalies in the first validation set, α is the corresponding weighting coefficient, and the magnitude of α is associated with the number of training times. The smaller the number of training times, the smaller the value corresponding to α.

3. A multi-dimensional traffic safety abnormality analysis and early warning system according to claim 2, characterized in that: The preset behavior detection strategy includes the following steps: Input the extracted key features into the vehicle behavior detection model constructed based on the preset model construction strategy, and output the corresponding vehicle behavior analysis data; The preset model construction strategy includes: Obtain the historical vehicle data of the current road, where the historical vehicle data includes the historical key features of the current road and the corresponding historical vehicle behavior analysis data; Construct a vehicle behavior detection model; Divide the obtained historical vehicle data into a second training set and a second validation set, and input the historical key features in the second training set into the constructed vehicle behavior detection model. The historical vehicle behavior analysis data in the second training set is used as the input data of the vehicle behavior detection model to train the vehicle behavior detection model; After each training is completed, input the historical key features in the second validation set into the trained vehicle behavior detection model, output the corresponding second prediction results, and based on the second prediction results and the corresponding historical vehicle behavior analysis data in the second validation set, determine whether the corresponding vehicle behavior detection model is an effective model. If it is an effective model, the training of the vehicle behavior detection model is completed; otherwise, continue to train the vehicle behavior detection model.

4. A multi-dimensional traffic safety abnormality analysis and early warning system according to claim 3, characterized in that: The risk assessment strategy includes the following steps: According to the output road environment anomaly data and vehicle behavior analysis data, identify the location information corresponding to the abnormal positions on the current road, the environmental anomaly types corresponding to each location information, and the vehicle behavior anomaly types corresponding to the vehicles at each location; According to the environmental anomaly types corresponding to each location information and the vehicle behavior anomaly types corresponding to the vehicles at each location, match the risk values of the corresponding environmental anomaly types and the risk values of the vehicle behavior anomaly types. Based on the preset comprehensive risk value calculation formula, calculate the comprehensive risk value at the corresponding location, and based on the level matching strategy, match the corresponding risk level through the risk value; The risk value calculation formula is: In the formula, FX is the comprehensive risk value at the corresponding position, H is the risk value corresponding to the abnormal environment type at the corresponding position, and Y is the risk value corresponding to the abnormal vehicle behavior at the corresponding position. μ is the corresponding weight value; The level matching strategy is: When 0 < FX ≤ X, the corresponding risk level is the first level; When X < FX ≤ Z, the corresponding risk level is the second level; When Z < FX ≤ R, the corresponding risk level is the third level; When FX > R, the corresponding risk level is the fourth level; X, Z, and R are dynamically changing interval points, and the dynamically changing interval points are associated with the number of abnormal positions on the current road.

5. A multi-dimensional traffic safety abnormality analysis and early warning system according to claim 4, characterized in that: The server further includes: A road division module, configured to divide the current road into a plurality of data collection points based on a preset distance interval according to the basic path of the current road; The optimization module is used to select the optimal collection point set of the current road based on several data collection points of the current road and historical related data based on the improved optimization algorithm; The multi-dimensional data acquisition module is also used to select vehicle driving data information whose vehicle position information is the same as the position information of each data collection point in the optimal collection point set from the acquired vehicle driving data information according to the selected optimal collection point set of the current road, and input it into the data preprocessing module.

6. A multi-dimensional traffic safety abnormality analysis and early warning system according to claim 5, characterized in that: The optimization module includes: The initial screening module is used to randomly generate N initial populations according to several data collection points of the current road. The individuals of the initial population are the sets formed by the data collection points. Based on the preset restriction conditions, the individuals of the initial population are screened and judged. If the conditions are met, the corresponding initial population is a feasible solution. Otherwise, it is an infeasible solution and the corresponding initial population is eliminated. The restrictions are: Where M is the total number of data collection bags, M max is the maximum allowed quantity, ρ k For a specific position P k The density of surrounding data collection points, ρ mix is the minimum permissible density; A fitness calculation module, used to calculate the first fitness and the second fitness of the selected population; The first fitness is calculated as follows: In the formula, F1 is the first fitness, D1 is the total value of the data collection points corresponding to the population, and x i is the vehicle passing rate of the i-th data collection point in the population, c i is the number of historical vehicles passing through the i-th data collection point in the population, C is the total number of historical vehicles passing through, w1, w2 are the corresponding weighting coefficients, B is a random number between 0 and 1, b y is a random threshold; The second fitness is calculated as follows: In the formula, F2 is the second fitness, D2 is the distance difference between data collection points in the population, and d pq is the distance between the qth data collection point and the pth data collection point; A selection module is used to select a population whose first fitness is less than or equal to a preset first fitness threshold within a first number of iterations; When the first number of iterations is exceeded, selecting a population whose second fitness is less than or equal to a preset second fitness threshold; The crossover mutation module is used to cross and mutate the selected population to obtain the offspring population; A loop module is used to continue executing the fitness calculation module after obtaining the offspring population until a preset number of iterations is met; The output module is used to output the offspring population as the optimal solution set for multi-objective optimization.

7. A multi-dimensional traffic safety abnormality analysis and early warning system according to claim 6, characterized in that: The server also includes: The precise module is used to determine the data collection point corresponding to the position information based on the position information corresponding to the position where the abnormality exists on the current road identified when the determined risk level is the second level, the third level or the fourth level, and select the vehicle driving data information within the preset range where the corresponding data collection point is located from the vehicle driving data information obtained by the multi-dimensional data acquisition module, and determine the position corresponding to the initial point position where the corresponding data collection point has an abnormality based on the vehicle driving data information within the range where the corresponding data collection point is located, and upload it to the server, wherein the initial point position is the initial position corresponding to the abnormality of the data collection point.

8. A multi-dimensional traffic safety abnormality analysis and early warning method, characterized by: A multi-dimensional traffic safety anomaly analysis and warning system using any one of claims 1 to 7 above.

9. A multi-dimensional traffic safety anomaly analysis and early warning program product, comprising a computer program, characterized in that: When the computer program is executed, a multi-dimensional traffic safety anomaly analysis and early warning system according to any one of claims 1 to 7 is implemented.

Images