Movatterモバイル変換


[0]ホーム

URL:


CN118982915B - Traffic accident early warning and triangular warning board deployment system based on AI - Google Patents

Traffic accident early warning and triangular warning board deployment system based on AI
Download PDF

Info

Publication number
CN118982915B
CN118982915BCN202411049938.2ACN202411049938ACN118982915BCN 118982915 BCN118982915 BCN 118982915BCN 202411049938 ACN202411049938 ACN 202411049938ACN 118982915 BCN118982915 BCN 118982915B
Authority
CN
China
Prior art keywords
data
historical
influence value
static
node sequence
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.)
Active
Application number
CN202411049938.2A
Other languages
Chinese (zh)
Other versions
CN118982915A (en
Inventor
于莹
邓翔
丁自勋
王鹏坤
钟文浩
徐定邦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ningbo Yongjia Auto Parts Co ltd
Original Assignee
Ningbo Yongjia Auto Parts Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Ningbo Yongjia Auto Parts Co ltdfiledCriticalNingbo Yongjia Auto Parts Co ltd
Priority to CN202411049938.2ApriorityCriticalpatent/CN118982915B/en
Publication of CN118982915ApublicationCriticalpatent/CN118982915A/en
Application grantedgrantedCritical
Publication of CN118982915BpublicationCriticalpatent/CN118982915B/en
Activelegal-statusCriticalCurrent
Anticipated expirationlegal-statusCritical

Links

Classifications

Landscapes

Abstract

The invention relates to the technical field of intelligent traffic, and discloses a traffic accident early warning and triangle warning board deployment system based on Al, which comprises calculating static influence value based on static feature data, calculating historical influence value based on historical feature data, determining an initial node sequence according to the static influence value and the historical influence value, inputting an event map and a G-layer feature map into a pre-constructed dynamic prediction model to obtain dynamic adjustment factors, finally, the initial node sequence is adjusted according to the dynamic adjustment factors to obtain a target node sequence, diversified data are adopted, static information and dynamic information are distinguished, the initial node sequence is generated according to static characteristic data and historical characteristic data, the dynamic adjustment factors are determined according to the dynamic characteristic data, and finally the initial node sequence is adjusted according to the dynamic adjustment factors to obtain the target node sequence, so that accuracy of a prediction result is further improved.

Description

Traffic accident early warning and triangular warning board deployment system based on AI
Technical Field
The invention relates to the technical field of intelligent traffic, in particular to an AI-based traffic accident early warning and triangular warning board deployment system.
Background
With the continuous improvement of the living standard of people, the automobile conservation amount in China is increased year by year, traffic accidents and vehicle faults occur, in general, warning signs are required to be placed on roads where accidents occur frequently, so that drivers are reminded of paying attention to road safety, but because the dynamic characteristics (such as traffic flow) of the roads are continuously changed, an artificial intelligence technology is required for traffic accident early warning, the positions of the traffic warning signs are reasonably selected and deployed, big data, intelligent cities, data mining technologies and the like are gradually introduced into road traffic safety judgment, the application of the new technologies can be used for directionally analyzing the road traffic ecology, and in the prior art, although the traffic accidents are predicted through the big data and other technologies, certain problems still exist.
For example, chinese patent with publication number CN118230556a provides a traffic parameter prediction method, apparatus and device, according to static information and real-time dynamic information, a space-time diagram model of a road network structure is constructed, space feature extraction is performed on the space-time diagram model to obtain real-time space features of the road network structure, time feature extraction is performed on the space-time diagram model to obtain real-time features of the road network structure, and traffic parameter prediction is performed on the road network structure according to the real-time space features and the real-time features to obtain traffic parameters of the road network structure at preset future time.
The above patent predicts the traffic parameters, but the data used is not diversified enough, and the static information is not strong in real-time performance, the dynamic information is strong in real-time performance, and the difference between the dynamic information and the static information is large, so that the above patent predicts the static information and the dynamic information about traffic together, and the prediction result is definitely inaccurate.
In view of the above, the present invention proposes an AI-based traffic accident warning and triangle warning board deployment system to solve the above-mentioned problems.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention provides an AI-based traffic accident early warning and triangle warning board deployment system.
In order to achieve the above purpose, the present invention provides the following technical solutions:
Traffic accident early warning and triangle warning board deployment system based on AI includes:
the data acquisition module is used for acquiring road network data of the area to be monitored, determining M nodes in the area to be monitored based on the road network data, and respectively acquiring static characteristic data, dynamic characteristic data and historical characteristic data corresponding to the M nodes, wherein the dynamic characteristic data at least comprises descriptive text data and real-time image data, and M is an integer larger than 0;
The first processing module is used for calculating a static influence value based on the static characteristic data, calculating a historical influence value based on the historical characteristic data, and determining an initial node sequence according to the static influence value and the historical influence value;
the second processing module is used for constructing an event map based on the descriptive text data, acquiring a G layer feature map based on the real-time image data, inputting the event map and the G layer feature map into a pre-constructed dynamic prediction model, and acquiring a dynamic adjustment factor, wherein G is an integer larger than 1:
And the adjusting module is used for adjusting the initial node sequence according to the dynamic adjusting factor to obtain a target node sequence, and determining the corresponding warning sign position according to the target node sequence.
Further, the method for acquiring the G-layer feature map based on the real-time image data comprises the following steps:
and carrying out feature extraction processing on the real-time image data according to the preset G extraction depths and the feature pyramid network to obtain a G-layer feature map corresponding to the real-time image data, wherein the G extraction depths are in one-to-one correspondence with the G-layer feature map.
Further, the method for constructing the event map based on the descriptive text data comprises the following steps:
And carrying out keyword recognition on the descriptive text data to obtain H keywords, taking the H keywords as objects, determining the corresponding relation among the H objects, taking the H objects as nodes and the corresponding relation among the H objects as edges, and thus constructing the event graph.
Further, the method for calculating the static influence degree value based on the static characteristic data comprises the steps of:
Wherein SIV is a static influence value, WTP is a passing waiting time, NDE is a node distance, NRI is a road crossing number, SPD is a passing standard speed, log2 [. Cndot ] is a logarithmic function based on 2, cosh (& gtis a hyperbolic cosine function, sin-1 (&) is an arcsine function, exp (&) is an exponential function based on e, and e and k are constants greater than 0.
Further, the historical characteristic data comprises accident occurrence times, peak time times, accident response time and time duration, and the method for calculating the historical influence degree value based on the historical characteristic data comprises the following steps:
Where HIV is the historical impact value, NAS is the accident occurrence frequency, NPS is the peak time, DUP is the time duration, ART is the accident response time, tan-1 (. Cndot.) is the arctangent function, cot-1 (. Cndot.) is the anticontrol function, and cos-1 (. Cndot.) is the anticcosine function.
Further, the method for determining the initial node sequence according to the static influence value and the historical influence value comprises the following steps:
And calculating the sum of the static influence value and the historical influence value according to a preset proportion, taking the sum as an initial influence value, and carrying out descending arrangement according to the initial influence values corresponding to the M nodes to obtain an initial node sequence.
Further, the construction method of the dynamic prediction model comprises the following steps:
The method comprises the steps of obtaining a sample data set, wherein the sample data set comprises a historical event graph, a historical feature graph and a historical dynamic adjustment factor, dividing the sample data set into a sample training set and a sample testing set, constructing a regression network, taking the historical event graph and the historical feature graph in the sample training set as input data of the regression network, taking the historical dynamic adjustment factor in the sample training set as output data of the regression network, training the regression network to obtain an initial regression network for predicting the real-time dynamic adjustment factor, testing the initial regression network by utilizing the sample testing set, and outputting the initial regression network meeting the requirement of less than a preset error value as a dynamic prediction model.
Further, the method for obtaining the target node sequence by adjusting the initial node sequence according to the dynamic adjustment factor comprises the following steps:
Multiplying the initial influence value corresponding to each node in the initial node sequence by a dynamic adjustment factor to obtain a target influence value, and then performing descending arrangement according to the target influence value to obtain a target node sequence.
The traffic accident early warning and triangle warning board deployment method based on the AI comprises the following steps:
Acquiring road network data of an area to be monitored, determining M nodes in the area to be monitored based on the road network data, and respectively acquiring static characteristic data, dynamic characteristic data and historical characteristic data corresponding to the M nodes, wherein the dynamic characteristic data at least comprises descriptive text data and real-time image data, and M is an integer greater than 0;
Calculating a static influence value based on the static feature data, calculating a historical influence value based on the historical feature data, and determining an initial node sequence according to the static influence value and the historical influence value;
Constructing an event map based on descriptive text data, acquiring a G layer feature map based on real-time image data, inputting the event map and the G layer feature map into a pre-constructed dynamic prediction model, and acquiring a dynamic adjustment factor, wherein G is an integer greater than 1;
And adjusting the initial node sequence according to the dynamic adjustment factor to obtain a target node sequence, and determining the corresponding warning board position according to the target node sequence.
A computer readable storage medium, on which a computer program is stored, which when executed implements the AI-based traffic accident warning and triangle warning board deployment method described above.
Compared with the prior art, the invention has the beneficial effects that:
According to the method, a static influence value is calculated based on static feature data, a historical influence value is calculated based on historical feature data, an initial node sequence is determined according to the static influence value and the historical influence value, an event diagram is constructed based on descriptive text data, a G-layer feature diagram is obtained based on real-time image data, the event diagram and the G-layer feature diagram are input into a pre-constructed dynamic prediction model, a dynamic adjustment factor is obtained, finally the initial node sequence is adjusted according to the dynamic adjustment factor to obtain a target node sequence, and a corresponding warning board position is determined according to the target node sequence.
Drawings
FIG. 1 is a schematic diagram of a traffic accident early warning and triangle warning board deployment system based on AI in the invention;
FIG. 2 is a flow chart of the AI-based traffic accident warning and triangle warning board deployment method of the present invention;
FIG. 3 is a schematic diagram of an event diagram according to the present invention;
FIG. 4 is a schematic diagram of a computer readable storage medium according to the present invention.
Reference numerals:
11. nodes and 12 edges.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1
Referring to fig. 1, the disclosure of the present embodiment provides an AI-based traffic accident warning and triangle warning board deployment system, including:
the data acquisition module is used for acquiring road network data of the area to be monitored, determining M nodes in the area to be monitored based on the road network data, and respectively acquiring static characteristic data, dynamic characteristic data and historical characteristic data corresponding to the M nodes, wherein the dynamic characteristic data at least comprises descriptive text data and real-time image data, and M is an integer larger than 0;
In this embodiment, the road network data includes detailed road network map and geographic information data obtained from a map system, and further includes remote sensing images captured from an unmanned aerial vehicle, where the nodes include intersections and turning points of roads, and determining M nodes in the area to be monitored based on the road network data may be directly obtained by using an analysis tool in the existing map system (such as OpenStreetMap and hundred-degree maps), which is not described in detail in this embodiment.
It should be noted that, the above static feature data includes, but is not limited to, a node distance, a road intersection number, a passing standard speed and a passing waiting time, where the node distance refers to a distance between a node and other nodes in a preset range, and may be a total distance between the node and other nodes or an average distance between the node and other nodes, and by taking a target node as a center, a circular area with a diameter R is exemplified as a preset range, all nodes falling into the circular area need to calculate a distance between the node and the target node, and take a sum of the distances as a node distance corresponding to the target node, and the passing standard speed refers to an average speed passing through the node, for example, taking the node as a turning point of a road as an example, a corresponding sensor may be set, and an average speed passing through the turning point within a certain time is calculated as the passing standard speed.
In the above description, the road crossing number refers to the number of roads passing through the node, for example, an intersection formed by crossing three roads, the corresponding road crossing number is 3, and when the node is a turning point of a road, the road crossing number is generally 1, the passing waiting time refers to the maximum time for limiting the passing of the node, and the node is taken as an intersection example, so that the corresponding passing waiting time is red light waiting time, and it can be understood that if the node is a turning point of a road, the corresponding passing waiting time can be zero.
In this embodiment, the dynamic feature data at least includes descriptive text data and real-time image data, where the descriptive text data represents text data describing a real-time traffic state of the node, for example, "the intersection B is in a peak period, the traffic flow is large, and the vehicle queuing length exceeds 200 meters", and the real-time image data refers to a real-time image set acquired by a camera disposed at the node, where the descriptive text data may be acquired in real time by an existing navigation system.
The historical characteristic data comprises but is not limited to accident occurrence times, peak time, accident response time and time duration, wherein the accident occurrence times can be the average accident occurrence times of the node per month or the accident occurrence times of a plurality of months, the peak time times can be the peak times of each day or the peak times of a plurality of days, the accident response time is characterized by the time required for rescue to arrive after the accident, it is understood that the greater the accident response time is, the greater the influence degree of the accident on the node is, the probability of the subsequent occurrence of traffic accidents on the node is, the time duration refers to the time duration of the peak time, and the historical characteristic data is continuously recorded through a log and stored in a system.
The first processing module is used for calculating a static influence value based on the static characteristic data, calculating a historical influence value based on the historical characteristic data, and determining an initial node sequence according to the static influence value and the historical influence value;
the method for calculating the static influence value based on the static characteristic data comprises the following steps:
Wherein SIV is a static influence value, WTP is a passing waiting time, NDE is a node distance, NRI is a road crossing number, SPD is a passing standard speed, log2 [. Cndot ] is a logarithmic function based on 2, cosh (& gtis a hyperbolic cosine function, sin-1 (&) is an arcsine function, exp (&) is an exponential function based on e, and e and k are constants greater than 0.
In this embodiment, by taking the waiting time and the number of road intersections as an example, the longer the waiting time, the more serious the congestion degree of the corresponding node, the greater the probability of occurrence of an accident, the more complex the number of road intersections, the greater the probability of occurrence of an accident, and the static influence value is positively correlated with the probability of occurrence of an accident of the node.
The method for calculating the historical influence value based on the historical characteristic data comprises the following steps:
Where HIV is the historical impact value, NAS is the accident occurrence frequency, NPS is the peak time, DUP is the time duration, ART is the accident response time, tan-1 (. Cndot.) is the arctangent function, cot-1 (. Cndot.) is the anticontrol function, and cos-1 (. Cndot.) is the anticcosine function.
In this embodiment, the number of occurrences of the accident and the duration of the time period are taken as examples, the more the number of occurrences of the accident is, the greater the probability of occurrence of the accident for the corresponding node is, and similarly, the longer the duration of the time period is, the more serious the congestion degree of the corresponding node is, the greater the probability of occurrence of the accident is, and the history influence value is positively correlated with the probability of occurrence of the accident for the node.
The method for determining the initial node sequence according to the static influence value and the historical influence value comprises the following steps:
And calculating the sum of the static influence value and the historical influence value according to a preset proportion, taking the sum as an initial influence value, and carrying out descending arrangement according to the initial influence values corresponding to the M nodes to obtain an initial node sequence.
It should be noted that, the preset proportion refers to the weight between the static influence value and the historical influence value, the preset proportion can be determined by experience of an expert, the descending order is only exemplary, and the initial node sequence represents the arrangement of the accident probability of the nodes.
The second processing module is used for constructing an event map based on the descriptive text data, acquiring a G layer feature map based on the real-time image data, inputting the event map and the G layer feature map into a pre-constructed dynamic prediction model, and acquiring a dynamic adjustment factor, wherein G is an integer larger than 1:
the method for constructing the event map based on the descriptive text data comprises the following steps:
And carrying out keyword recognition on the descriptive text data to obtain H keywords, taking the H keywords as objects, determining the corresponding relation among the H objects, taking the H objects as nodes and the corresponding relation among the H objects as edges, and thus constructing the event graph.
It can be appreciated that in the prior art, there are various technologies specifically used for keyword recognition, for example, word2Vec may capture the semantic similarity of terms, extract keywords by clustering or other methods, YAKE is an unsupervised keyword extraction method, focus on single document keyword extraction, and may also use the existing pre-training language model to perform migration learning to generate a required keyword extraction model, which is not repeated in this embodiment.
It should be noted that the description text data characterizes text data describing the real-time traffic state of the node, and the description is specifically described below.
The node is taken as an intersection for illustration, the description text comprises three description sub-texts, and the three description sub-texts are respectively:
the first descriptive text is that the traffic is smooth before the peak of the intersection B, and vehicles can pass through quickly.
And the second descriptive text is that the intersection B is at the initial stage of the peak period, the traffic flow is increased, and the vehicles start to wait in line.
And the third descriptive text is that' the intersection B is in the middle and later period of the peak period, the traffic jam is serious, the vehicle queuing length exceeds 200 meters, and the intersection B needs to wait for a long time.
In the above description, the description text includes three description sub-texts, the keyword extraction is performed on the three description sub-texts, the keyword extracted from the first description sub-text is "intersection B", "before peak" and "clear", the keyword extracted from the second description sub-text is "intersection B", "early peak", "queuing", "waiting", the keyword extracted from the third description sub-text is "intersection B", "late peak middle", "congestion" and "long waiting", then the corresponding correspondence between the "before peak" and "clear", the correspondence between the "early peak" and the "queuing", "waiting" exists, the correspondence between the "late peak middle" and the "congestion" and the "long waiting" exists, it can be understood that the common keyword "intersection B" in the three description sub-texts is taken as a starting point, an event diagram is constructed, as shown in fig. 3, the node 11 and the edge 12 exist, it can be understood that the node 11 represents the keyword in the above content and the edge 12 represents the keyword.
The method for acquiring the G-layer feature map based on the real-time image data comprises the following steps:
and carrying out feature extraction processing on the real-time image data according to the preset G extraction depths and the feature pyramid network to obtain a G-layer feature map corresponding to the real-time image data, wherein the G extraction depths are in one-to-one correspondence with the G-layer feature map.
It should be noted that the feature pyramid network is a deep learning model structure for computer vision tasks, and is mainly characterized in that the feature pyramid network utilizes a basic convolution neural network (such as ResNet) to extract features from shallow layers to deep layers respectively by constructing a multi-scale feature map pyramid from high resolution to low resolution, so that the detection capability of the model on objects with different scales is enhanced, and the feature maps have different resolutions and can capture information with different scales.
Taking G as 3 and taking real-time image data as an intersection image as an example, three layers of feature images can be generated, wherein the first layer of feature images capture low-level edge and texture information, including low-level detail information such as lane marks, pedestrians, vehicle outlines and the like, the second layer of feature images capture medium-level shape and partial object information, including medium-level structure information such as more complete vehicles, pedestrians, traffic lights and the like, and the third layer of feature images capture high-level semantic information, including high-level semantic information such as layout, traffic flow modes and the like of the whole intersection.
The construction method of the dynamic prediction model comprises the following steps:
The method comprises the steps of obtaining a sample data set, wherein the sample data set comprises a historical event graph, a historical feature graph and a historical dynamic adjustment factor, dividing the sample data set into a sample training set and a sample testing set, constructing a regression network, taking the historical event graph and the historical feature graph in the sample training set as input data of the regression network, taking the historical dynamic adjustment factor in the sample training set as output data of the regression network, training the regression network to obtain an initial regression network for predicting the real-time dynamic adjustment factor, testing the initial regression network by utilizing the sample testing set, outputting the initial regression network meeting a preset error value as a dynamic prediction model, and optimizing the initial regression network to be a deep neural network model.
It can be understood that the above-mentioned historical event map is constructed by using historical description text data, and similarly, the above-mentioned historical feature map is extracted from historical image data, and the historical dynamic adjustment factors are obtained by grading the congestion degree of the nodes according to actual conditions by expert knowledge, so that the corresponding dynamic adjustment factors are set according to the congestion degree, and in this embodiment, the more serious the congestion degree of the nodes, the larger the value of the corresponding dynamic adjustment factors.
And the adjusting module is used for adjusting the initial node sequence according to the dynamic adjusting factor to obtain a target node sequence, and determining the corresponding warning sign position according to the target node sequence.
The method for obtaining the target node sequence by adjusting the initial node sequence according to the dynamic adjustment factor comprises the following steps:
Multiplying the initial influence value corresponding to each node in the initial node sequence by a dynamic adjustment factor to obtain a target influence value, and then performing descending arrangement according to the target influence value to obtain a target node sequence.
Specifically, each node has a corresponding initial influence value and a corresponding dynamic adjustment factor, so that a target influence value is obtained, and then the target node sequences are obtained by descending order.
In this embodiment, a corresponding warning board is provided on the path before each node, and the warning board can be automatically lifted or lightened to prompt the driver that the node in front is jammed and needs to be carefully driven, because the larger the target influence value in this embodiment is, the probability of accident of the node is larger, so that a corresponding influence threshold can be set, and when the target influence value corresponding to the node is larger than the influence threshold, the corresponding warning board is lifted or electrified, so that the warning board is more striking.
In this embodiment, M nodes in a region to be monitored are determined based on road network data, static feature data, dynamic feature data and historical feature data corresponding to the M nodes are respectively obtained, a static influence value is calculated based on the static feature data, a historical influence value is calculated based on the historical feature data, an initial node sequence is determined according to the static influence value and the historical influence value, an event map is built based on descriptive text data, a G-layer feature map is obtained based on real-time image data, the event map and the G-layer feature map are input into a pre-built dynamic prediction model, a dynamic adjustment factor is obtained, finally the initial node sequence is adjusted according to the dynamic adjustment factor to obtain a target node sequence, corresponding warning sign positions are determined according to the target node sequence, the static information and the dynamic information are distinguished, the initial node sequence is generated according to the static feature data and the historical feature data, the dynamic adjustment factor is determined according to the dynamic feature data, and finally the initial node sequence is adjusted according to the dynamic adjustment factor to obtain the target node sequence, and the accuracy of a prediction result is further improved.
Example 2
As shown in fig. 2, the present embodiment provides, based on embodiment 1, an AI-based traffic accident warning and triangle warning board deployment method, including:
S10, acquiring road network data of an area to be monitored, determining M nodes in the area to be monitored based on the road network data, and respectively acquiring static characteristic data, dynamic characteristic data and historical characteristic data corresponding to the M nodes, wherein the dynamic characteristic data at least comprises descriptive text data and real-time image data, and M is an integer larger than 0;
it should be noted that, the above static feature data includes, but is not limited to, a node distance, a road intersection number, a passing standard speed and a passing waiting time, where the node distance refers to a distance between a node and other nodes in a preset range, and may be a total distance between the node and other nodes or an average distance between the node and other nodes, and by taking a target node as a center, a circular area with a diameter R is exemplified as a preset range, all nodes falling into the circular area need to calculate a distance between the node and the target node, and take a sum of the distances as a node distance corresponding to the target node, and the passing standard speed refers to an average speed passing through the node, for example, taking the node as a turning point of a road as an example, a corresponding sensor may be set, and an average speed passing through the turning point within a certain time is calculated as the passing standard speed.
In the above description, the road crossing number refers to the number of roads passing through the node, for example, an intersection formed by crossing three roads, the corresponding road crossing number is 3, and when the node is a turning point of a road, the road crossing number is generally 1, the passing waiting time refers to the maximum time for limiting the passing of the node, and the node is taken as an intersection example, so that the corresponding passing waiting time is red light waiting time, and it can be understood that if the node is a turning point of a road, the corresponding passing waiting time can be zero.
In this embodiment, the dynamic feature data at least includes descriptive text data and real-time image data, where the descriptive text data represents text data describing a real-time traffic state of the node, for example, "the intersection B is in a peak period, the traffic flow is large, and the vehicle queuing length exceeds 200 meters", and the real-time image data refers to a real-time image set acquired by a camera disposed at the node, where the descriptive text data may be acquired in real time by an existing navigation system.
The historical characteristic data comprises but is not limited to accident occurrence times, peak time, accident response time and time duration, wherein the accident occurrence times can be the average accident occurrence times of the node per month or the accident occurrence times of a plurality of months, the peak time times can be the peak times of each day or the peak times of a plurality of days, the accident response time is characterized by the time required for rescue to arrive after the accident, and it is understood that the greater the accident response time is, the greater the influence degree of the accident on the node is, the probability of the subsequent occurrence of traffic accidents of the node is, the time duration refers to the time duration of the peak time, and the historical characteristic data is continuously recorded through a log and stored in a system.
S20, calculating a static influence value based on the static characteristic data, calculating a historical influence value based on the historical characteristic data, and determining an initial node sequence according to the static influence value and the historical influence value;
the method for calculating the static influence value based on the static characteristic data comprises the following steps:
Wherein SIV is a static influence value, WTP is a passing waiting time, NDE is a node distance, NRI is a road crossing number, SPD is a passing standard speed, log2 [. Cndot ] is a logarithmic function based on 2, cosh (& gtis a hyperbolic cosine function, sin-1 (&) is an arcsine function, exp (&) is an exponential function based on e, and e and k are constants greater than 0.
The method for calculating the historical influence value based on the historical characteristic data comprises the following steps:
Where HIV is the historical impact value, NAS is the accident occurrence frequency, NPS is the peak time, DUP is the time duration, ART is the accident response time, tan-1 (. Cndot.) is the arctangent function, cot-1 (. Cndot.) is the anticontrol function, and cos-1 (. Cndot.) is the anticcosine function.
The method for determining the initial node sequence according to the static influence value and the historical influence value comprises the following steps:
And calculating the sum of the static influence value and the historical influence value according to a preset proportion, taking the sum as an initial influence value, and carrying out descending arrangement according to the initial influence values corresponding to the M nodes to obtain an initial node sequence.
It should be noted that, the preset proportion refers to the weight between the static influence value and the historical influence value, the preset proportion can be determined by experience of an expert, the descending order is only exemplary, and the initial node sequence represents the arrangement of the accident probability of the nodes.
S30, constructing an event map based on descriptive text data, acquiring a G layer feature map based on real-time image data, inputting the event map and the G layer feature map into a pre-constructed dynamic prediction model, and acquiring a dynamic adjustment factor, wherein G is an integer larger than 1;
the method for constructing the event map based on the descriptive text data comprises the following steps:
And carrying out keyword recognition on the descriptive text data to obtain H keywords, taking the H keywords as objects, determining the corresponding relation among the H objects, taking the H objects as nodes and the corresponding relation among the H objects as edges, and thus constructing the event graph.
It can be appreciated that in the prior art, there are various technologies specifically used for keyword recognition, for example, word2Vec may capture the semantic similarity of terms, extract keywords by clustering or other methods, YAKE is an unsupervised keyword extraction method, focus on single document keyword extraction, and may also use the existing pre-training language model to perform migration learning to generate a required keyword extraction model, which is not repeated in this embodiment.
The method for acquiring the G-layer feature map based on the real-time image data comprises the following steps:
and carrying out feature extraction processing on the real-time image data according to the preset G extraction depths and the feature pyramid network to obtain a G-layer feature map corresponding to the real-time image data, wherein the G extraction depths are in one-to-one correspondence with the G-layer feature map.
The construction method of the dynamic prediction model comprises the following steps:
The method comprises the steps of obtaining a sample data set, wherein the sample data set comprises a historical event graph, a historical feature graph and a historical dynamic adjustment factor, dividing the sample data set into a sample training set and a sample testing set, constructing a regression network, taking the historical event graph and the historical feature graph in the sample training set as input data of the regression network, taking the historical dynamic adjustment factor in the sample training set as output data of the regression network, training the regression network to obtain an initial regression network for predicting the real-time dynamic adjustment factor, testing the initial regression network by utilizing the sample testing set, outputting the initial regression network meeting a preset error value as a dynamic prediction model, and optimizing the initial regression network to be a deep neural network model.
And S40, adjusting the initial node sequence according to the dynamic adjustment factor to obtain a target node sequence, and determining the corresponding warning board position according to the target node sequence.
The method for obtaining the target node sequence by adjusting the initial node sequence according to the dynamic adjustment factor comprises the following steps:
Multiplying the initial influence value corresponding to each node in the initial node sequence by a dynamic adjustment factor to obtain a target influence value, and then performing descending arrangement according to the target influence value to obtain a target node sequence.
Specifically, each node has a corresponding initial influence value and a corresponding dynamic adjustment factor, so that a target influence value is obtained, and then the target node sequences are obtained by descending order.
In this embodiment, a corresponding warning board is provided on the path before each node, and the warning board can be automatically lifted or lightened to prompt the driver that the node in front is jammed and needs to be carefully driven, because the larger the target influence value in this embodiment is, the probability of accident of the node is larger, so that a corresponding influence threshold can be set, and when the target influence value corresponding to the node is larger than the influence threshold, the corresponding warning board is lifted or electrified, so that the warning board is more striking.
Example 3
The embodiment discloses an electronic device, which comprises a power supply, an interface, a keyboard, a memory, a central processing unit and a computer program stored on the memory and capable of running on the central processing unit, wherein the central processing unit realizes the AI-based traffic accident early warning and triangle warning board deployment method provided by the methods when executing the computer program, the interface comprises a network interface and a data interface, the network interface comprises a wired or wireless interface, and the data interface comprises an input or output interface.
Since the electronic device described in this embodiment is an electronic device used to implement the AI-based traffic accident warning and triangle warning board deployment method in this embodiment, those skilled in the art can understand the specific implementation manner of the electronic device and various variations thereof, so that the method of implementing the embodiment of the present application will not be described in detail herein. As long as the person skilled in the art implements the electronic equipment adopted by the traffic accident early warning and triangle warning board deployment method based on AI in the embodiment of the application, the electronic equipment belongs to the scope of the application to be protected.
Example 4
As shown in fig. 4, the disclosure of the present embodiment provides a computer readable storage medium, on which a computer program is stored, which when executed implements the above-mentioned AI-based traffic accident warning and triangle warning board deployment method.
The above formulas are all formulas with dimensionality removed and numerical value calculated, the formulas are formulas with the latest real situation obtained by software simulation by collecting a large amount of data, and preset parameters, weights and threshold selection in the formulas are set by those skilled in the art according to the actual situation.
The above embodiments may be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above-described embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the processes or functions described in accordance with embodiments of the present invention are produced in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from one website site, computer, server, or data center to another website site, computer, server, or data center over a wired network or a wireless network. The computer readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that contains one or more sets of available media. The usable medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium may be a solid state disk.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, and are not repeated herein.
In the several embodiments provided in the present invention, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely one, and there may be additional divisions in actual implementation, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other forms.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.
The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Finally, the foregoing description of the preferred embodiment of the invention is provided for the purpose of illustration only, and is not intended to limit the invention to the particular form disclosed, but on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention.

Claims (5)

Obtaining a sample data set, wherein the sample data set comprises a historical event graph, a historical feature graph and a historical dynamic adjustment factor, dividing the sample data set into a sample training set and a sample testing set, constructing a regression network, taking the historical event graph and the historical feature graph in the sample training set as input data of the regression network, taking the historical dynamic adjustment factor in the sample training set as output data of the regression network, training the regression network to obtain an initial regression network for predicting the real-time dynamic adjustment factor, testing the initial regression network by using the sample testing set, and outputting the initial regression network meeting the requirement of less than a preset error value as a dynamic prediction model;
CN202411049938.2A2024-07-312024-07-31Traffic accident early warning and triangular warning board deployment system based on AIActiveCN118982915B (en)

Priority Applications (1)

Application NumberPriority DateFiling DateTitle
CN202411049938.2ACN118982915B (en)2024-07-312024-07-31Traffic accident early warning and triangular warning board deployment system based on AI

Applications Claiming Priority (1)

Application NumberPriority DateFiling DateTitle
CN202411049938.2ACN118982915B (en)2024-07-312024-07-31Traffic accident early warning and triangular warning board deployment system based on AI

Publications (2)

Publication NumberPublication Date
CN118982915A CN118982915A (en)2024-11-19
CN118982915Btrue CN118982915B (en)2025-03-28

Family

ID=93448684

Family Applications (1)

Application NumberTitlePriority DateFiling Date
CN202411049938.2AActiveCN118982915B (en)2024-07-312024-07-31Traffic accident early warning and triangular warning board deployment system based on AI

Country Status (1)

CountryLink
CN (1)CN118982915B (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
CN115691132A (en)*2022-10-282023-02-03北京东土正创科技有限公司 Method, device, equipment and medium for determining traffic signal information
CN116312015A (en)*2023-03-242023-06-23吴嘉鹏Traffic infrastructure real-time monitoring and early warning system based on big dipper data cloud platform

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
CN112307840A (en)*2019-07-312021-02-02浙江商汤科技开发有限公司Indicator light detection method, device, equipment and computer readable storage medium
CN115272656A (en)*2022-07-292022-11-01京东方科技集团股份有限公司Environment detection alarm method and device, computer equipment and storage medium
CN117037486A (en)*2023-08-092023-11-10智慧互通科技股份有限公司Intelligent scheduling method and system based on multi-target fusion
CN118038682B (en)*2024-04-122024-07-16北京数字光芯集成电路设计有限公司Driving guiding method, device, equipment and medium based on digital projection lamp
CN118395130B (en)*2024-06-272024-08-16山东物慧信息科技有限公司Motion trail analysis and early warning method based on edge calculation

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
CN115691132A (en)*2022-10-282023-02-03北京东土正创科技有限公司 Method, device, equipment and medium for determining traffic signal information
CN116312015A (en)*2023-03-242023-06-23吴嘉鹏Traffic infrastructure real-time monitoring and early warning system based on big dipper data cloud platform

Also Published As

Publication numberPublication date
CN118982915A (en)2024-11-19

Similar Documents

PublicationPublication DateTitle
US9355063B2 (en)Parking lot detection using probe data
CN110378483A (en)The system and method for training machine learning model being deployed on analog platform
WO2022110611A1 (en)Pedestrian road-crossing behavior prediction method for plane intersection
JP7399891B2 (en) Providing additional instructions for difficult maneuvers during navigation
US20160098861A1 (en)Safety system for augmenting roadway objects on a heads-up display
US11694426B2 (en)Determining traffic control features based on telemetry patterns within digital image representations of vehicle telemetry data
WO2021082464A1 (en)Method and device for predicting destination of vehicle
CN112418696B (en)Construction method and device of urban traffic dynamic knowledge graph
CN110782120A (en)Method, system, equipment and medium for evaluating traffic flow model
CN112101187B (en)Road identification method and device and electronic equipment
CN119360555B (en) A vehicle-mounted geological disaster early warning system and method based on environmental perception
CN117033162A (en)Automatic driving virtual simulation test method and system based on scene generation
JP2024530216A (en) SYSTEM AND METHOD FOR MONITORING A VEHICLE - Patent application
CN116046008A (en)Situation awareness-based route planning method, system and efficiency evaluation device
Anagnostopoulos et al.Predicting roundabout lane capacity using artificial neural networks
CN111325811A (en)Processing method and processing device for lane line data
Nissimagoudar et al.Detection of potholes and speed breaker for autonomous vehicles
EP4328888A1 (en)Method and processing system for processing probe data and probe
CN118982915B (en)Traffic accident early warning and triangular warning board deployment system based on AI
CN114139984A (en) Risk prediction method of urban traffic accident based on collaborative perception of traffic flow and accident
CN116363530B (en)Method and device for positioning expressway pavement diseases
MuhammadDetection parking Spaces by using the ResNet50 Algorithm
Saif et al.Szr5: A modern technique to detect potholes using convolution neural network
Hayawi et al.Detection parking Spaces by using the AlexNet Algorithm
KR102383000B1 (en)Apparatus for system for operating safe driving system and operation method thereof

Legal Events

DateCodeTitleDescription
PB01Publication
PB01Publication
SE01Entry into force of request for substantive examination
SE01Entry into force of request for substantive examination
GR01Patent grant
GR01Patent grant

[8]ページ先頭

©2009-2025 Movatter.jp