CN109255956A - A kind of charge station's magnitude of traffic flow method for detecting abnormality - Google Patents
A kind of charge station's magnitude of traffic flow method for detecting abnormality Download PDFInfo
- Publication number
- CN109255956A CN109255956A CN201811341385.2A CN201811341385A CN109255956A CN 109255956 A CN109255956 A CN 109255956A CN 201811341385 A CN201811341385 A CN 201811341385A CN 109255956 A CN109255956 A CN 109255956A
- Authority
- CN
- China
- Prior art keywords
- traffic flow
- data
- magnitude
- charge station
- time series
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0125—Traffic data processing
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0137—Measuring and analyzing of parameters relative to traffic conditions for specific applications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/065—Traffic control systems for road vehicles by counting the vehicles in a section of the road or in a parking area, i.e. comparing incoming count with outgoing count
Abstract
The invention discloses a kind of charge station's magnitude of traffic flow method for detecting abnormality, the data of the disengaging highway by charge station's vehicle is cleaned, are simplified, are sorted and data on flows timesharing accounting；It generates magnitude of traffic flow time series data and it is handled, using the time series data of generation as input, the magnitude of traffic flow of next period is output, is predicted using DBN model；By the relative error for calculating the traffic flow of prediction and the actual traffic stream of sliding-model control, judge whether the relative error is more than abnormal threshold value, if the relative error is more than threshold value, then there is exception in the current magnitude of traffic flow of the charge station, otherwise, which is under normal circumstances；The prediction to the magnitude of traffic flow is realized by deepness belief network college charge station magnitude of traffic flow feature using charge station's traffic throat floater judgment models based on deepness belief network, solves the problems, such as that the subjectivity in charge station's magnitude of traffic flow exception deterministic process is random.
Description
Technical field
The invention belongs to apply algorithm field, and in particular to a kind of charge station's magnitude of traffic flow method for detecting abnormality.
Background technique
Traffic throat floater judges the distribution of the work to charge station and management, the positioning of freeway network accident, highway
The assurance of road network operation situation has great significance.Correct charge station's traffic throat floater judgement is highway intelligent transportation system
The important prerequisite of the real charging channel control of system, traffic guidance and distribution, self-navigation, accident detection etc..Current toll gate
Magnitude of traffic flow judgement extremely mainly by toll collector combine the period, whether the information such as festivals or holidays, current location, and rule of thumb
Judge whether present flow rate is abnormal.The subjectivity of this judgment mode is larger, and needs a large amount of experience auxiliary for many years.
Summary of the invention
In order to solve problems of the prior art, the present invention discloses a kind of charge station's magnitude of traffic flow abnormality detection side
Method, using charge station's traffic throat floater judgment models based on deepness belief network.Pass through deepness belief network college charge station
Magnitude of traffic flow feature is realized the prediction to the magnitude of traffic flow, by TMRE (traffic mean relative error) and is combined
Outlier threshold is arranged in artificial experience, realizes the judgement of charge station's magnitude of traffic flow exception, solves charge station's magnitude of traffic flow and sentence extremely
The random problem of subjectivity during disconnected.
To achieve the goals above, the technical solution adopted by the present invention is that, a kind of charge station's magnitude of traffic flow abnormality detection side
Method, comprising the following steps:
Step 1, the data of each disengaging high speed by charge station's vehicle are handled；
Step 101 current data original to the vehicle of charge station pre-process, and carry out data cleansing, data compaction, number
According to sequence and data on flows timesharing accounting；
Step 102 generates magnitude of traffic flow time series data and handles it, first original logical from the vehicle of charge station
Row data generate magnitude of traffic flow time series data；Again to magnitude of traffic flow time series data be normalized and standardization at
Reason；
Step 2, based on the forecasting traffic flow of DBN: next using step 1 time series data generated as input
The magnitude of traffic flow of period is output, is predicted using DBN model；
Step 3, by calculating the relative error of the traffic flow of prediction and the actual traffic stream of sliding-model control, described in judgement
Whether relative error is more than abnormal threshold value, if the relative error is more than threshold value, the current magnitude of traffic flow of the charge station
There is exception, if the relative error is no more than threshold value, which is under normal circumstances.
The data include information time, the charge station for passing in and out highway and toll amount, in step 101, are removed different
Constant value data, missing Value Data and obvious problem data.
In step 101, select 4 fields from the database of charge station: time of entering the station, outbound time, enter the station station name with
And outbound station name is used to according to the time period accurately statistics charge station's magnitude of traffic flow.
In step 102, by through removing exceptional value, obvious problem data and missing values data in chronological sequence sequence into
Row sequence；The number of vehicles passed through in each characteristic time interval is counted, characteristic time interval is 1 hour.
In step 102, from the behaviour of step 101 treated charge station's Raw Data Generation magnitude of traffic flow time series data
It is as follows to make process:
Model parameter is first set, and the model parameter includes test set quantity, characteristic and the step-length of input model；
Regeneration training set is output and input, and is inputted as m*n matrix, m is number of samples, and n is characterized number, is exported as m*
S matrix, m are number of samples, and s is step-length.
In step 102, by the same Feature Mapping of the training set different data of magnitude of traffic flow time series data to [0,1]
Between, it is normalized using min-max, operating method is as follows:
Wherein xmin=Min (x1,x2,…,xm), xmax=Max (x1,x2,…,xm)。
In step 102, the feature different to the training set data of magnitude of traffic flow time series data carries out stretching behaviour
Make, so that the feature between homometric(al) is not comparable,
Wherein x1For xjIn first value.
In step 2, DBN model is divided into two training module, prediction module parts, and training module instructs given data
To practice, the feature of learning time sequence data generates the weight parameter of model and saves in the form of a file, later, prediction model
The next hour magnitude of traffic flow of charge station is predicted by stress model and using time series data.
In step 2, the running frequency of training module is set as in 1 time/January, pressing the present invention first in the operation of prediction module
Middle time series generation method, is the time series data in past 24 hours by the processing of original charge data, and prediction model uses
Past 24 hours time series datas predict lower traffic flow in 1 hour, and the running frequency of prediction model is 1 time/1 hour.
In step 3, discretization actual traffic data on flows, calculation formula is as follows:
Wherein, FpIt is the traffic flow of prediction, FmIt is actual traffic flow, n is the time interval for counting flow, that is, is pressed n minutes
The statistics magnitude of traffic flow simultaneously calculates relative error.
Compared with prior art, the present invention at least has the advantages that
Different from the neural network of traditional discrimination model, deepness belief network is a generative probabilistic model, generates mould
Type is the Joint Distribution established between an observation data and label, to P (Observation | Label) and P (Label |
Observation it) all assesses, by its interneuronal weight of training, entire neural network can be made according to most probably
Rate generates training data, and after the automatic refinement of bottom-layer network, the feature of refinement is applied on other networks can compare hand
The dynamic result for extracting feature is more preferable；The invention proposes a kind of abnormality judgment methods of discretization, by calculating predicting traffic flow
With the relative error (TMRE, traffic mean relative error) of actual traffic stream, sentence to whether flow is made extremely
It is disconnected, traffic flow is predicted using DBN model, without extracting the feature of data by experience, effectively artificial experience is avoided to judge
The randomness of Traffic Anomaly improves the accuracy judged extremely, by the time for greatly reducing personnel's empirical learning and training.
Further, the validity that abnormal Value Data, missing Value Data and obvious problem data guarantee charge data is removed.
Detailed description of the invention
Fig. 1 is time series data generating process schematic diagram；
Fig. 2 is the structure chart of DBN；
Fig. 3 is loss decline figure；
Fig. 4 is the magnitude of traffic flow result for predicting 30 days；
Fig. 5 is the magnitude of traffic flow result for predicting 5 days；
Fig. 6 is the mre scatter plot of 5 days predicted values.
Specific embodiment
The following describes the present invention in detail with reference to the accompanying drawings and specific embodiments.
The present invention learns the time that 3 months entrance flows are gone over by a certain charge station using deepness belief network (DBN)
Sequence data is realized to the following 1 hour forecasting traffic flow of charge station's entrance, on this basis, after calculating discretization
Currently practical flow and the relative error (TMRE) of predicted flow rate then determine after TMRE reaches the threshold of sensitivity of setting
Present flow rate is abnormal；The detection method of i.e. a kind of charge station's magnitude of traffic flow exception based on deepness belief network (DBN) study,
Specifically includes the following steps:
Step 1, data processing: each passing vehicle disengaging highway time of charge station's original data record, into
The charge station of highway and toll amount information out need to carry out certain to reduce the accuracy of later period algorithm and realizing difficulty
Data processing；This process is divided into two parts again, and data prediction and time series data generate.
Step 101, data prediction:
In process of data preprocessing, data cleansing, data compaction, data sorting and data on flows timesharing system are carried out
Meter；
1) abnormal Value Data and missing Value Data are removed, abnormal data includes that the number of axle is abnormal, toll amount is abnormal and logical
Row time anomaly lacks Value Data other than comprising absent field, further includes the incomplete situation of data out of the station.Pass through removing
Abnormal data, missing Value Data, guarantee the validity of charge data.
2) since charge data occupies larger, moving operation inconvenience and increase data processing duration, therefore data essence need to be carried out
Letter selects 4 fields according to the actual demand of later period algorithm from database: time of entering the station, outbound time, enter the station station name with
And outbound station name.
3) data sorting is carried out, data are in chronological sequence sequentially ranked up, is established for the generation of time series data
Basis.
4) trip information of one vehicle of a charging data record, data on flows timesharing accounting process are exactly to count each spy
The number of vehicles passed through in sign time interval.Preferred feature time interval of the present invention is 1 hour.
Step 102 time series data generates and processing
A) time series data generates
There are more close relationship in freeway traffic flow amount data and time, and in one day, 8 points to 22 points are flows
Peak period, in one week, workaday flow is obviously on the high side compared to weekend；Therefore, the present invention uses magnitude of traffic flow time series
Data are inputted as the feature of model；The operating process for generating magnitude of traffic flow time series data from charge station's data is as follows:
Parameter is first set, the parameter include training set and test set size, characteristic (sequence data window size) with
And step-length；When determining magnitude of traffic flow outlier threshold, the ratio between training set and test set are 3, in actual production, test set 1；
Characteristic is set as 12 or 24；Step-length depends primarily on the duration to be predicted, and preferably, step-length is set as 1 to the present invention, that is, predicts
The magnitude of traffic flow of next hour period；
The input and output data for generating neural network input as m*n matrix, and m is number of samples, and n is characterized number, defeated
It is out m*s matrix, m is number of samples, s step-length, as shown in Figure 1, for 1 step-length, illustrating time series data with 4 features
Generating process selects preceding 4 data as first input sample, and the 5th data are as its corresponding output sample.Select 2-
5 data as second input sample, the 6th data as the corresponding output sample of second input sample, and so on,
Until data can not generate output, stopping generates data.
B) time series data is handled
To reduce neural network learning difficulty, the convergence rate of lift scheme, the precision of prediction of lift scheme；Comparison with
After being predicted without using normalization with standardization or using single treatment, the present invention makes full use of normalization and standardization clock synchronization
Between data handled so that training speed is effectively promoted with training effect.
When normalized, the same Feature Mapping of the training set different data of magnitude of traffic flow time series data is arrived
It between [0,1], is normalized using min-max, operating method is as follows:
Wherein xmin=Min (x1,x2,…,xm), xmax=Max (x1,x2,…,xm)。
Standardization can carry out stretching operation to the different features in certain sample, so that the not feature between homometric(al)
It is comparable, standardization is also a kind of preliminary feature extraction, when can simplify the learning process of neural network to the magnitude of traffic flow
Between the different feature of training set data of sequence data carry out stretching operation so that have can not for the feature between homometric(al)
Than property,
Wherein x1For xjIn first value.
Step 2, based on the forecasting traffic flow of DBN
The structure chart of the position Fig. 2 DBN；Deepness belief network (Deep Belief Network) is a generative probabilistic model,
Opposite with the neural network of traditional discrimination model, generating model is the joint point established between an observation data and label
Cloth is all assessed P (Observation | Label) and P (Label | Observation), by between its neuron of training
Weight, entire neural network can be allowed to generate training data according to maximum probability, and use DBN model, without relying on
Experience extracts the feature of data, but after the automatic refinement of bottom-layer network, the feature of refinement is applied on other networks
It can be more preferable than the result of manual extraction feature.
3 months a certain charge station's traffic flow time series datas, Neural Network Data are defeated in the past using DBN model study by the present invention
The time series data entered are as follows: time interval 1 hour, number of features 24 (when 1 day a length of), step-length 1.Data output is described
Traffic flow in next hour of charge station.
The model is divided into two parts of training module and prediction module in the application, and training module instructs given data
Practice, the feature of learning time sequence data generates the weight parameter of model and saves in the form of a file；The operation of training module
Set of frequency is 1 time/January；Firstly the need of time series generation method in the present invention is pressed in the operation of prediction module, by original receipts
Take the time series data that data processing is 24 hours in the past, later, prediction model is by stress model and uses time series number
According to the magnitude of traffic flow for going out next hour as input prediction, the running frequency of prediction model is 1 time/1 hour.
Step 3, by calculate prediction traffic flow and sliding-model control actual traffic stream relative error (TMRE,
Traffic mean relative error), judge whether relative error is more than abnormal threshold value, since predicted value is discrete
1 hour flow, and actual flow is continuous flow, therefore needs discretization actual flow, and calculation formula is as follows:
Wherein, FpIt is the traffic flow of prediction, FmIt is actual traffic flow, n is the time interval for counting flow, that is, is pressed n minutes
The statistics magnitude of traffic flow simultaneously calculates relative error (n >=10).In this calculation formula, FpIt is that prediction model provides, it is only necessary to fixed
When acquire FmAnd the value of n is set, you can get it exports the judging result of traffic throat floater by n minutes intervals.
Embodiment
Using DBN carry out charge station's magnitude of traffic flow time series data study be convergence and it is achievable, training
Cheng Zhong, loss value constantly reduce, and gradually tend to optimal solution, and the tendency chart that loss value changes over time is as shown in Figure 3；In addition,
Most reasonable training parameter can be also selected between training time and training effect by Fig. 3, guaranteed while quick training
Training effect.
Fig. 4 and Fig. 5 is respectively shown 3 months time series datas of a certain charge station are learnt using DBN model after, carry out 1 small
When predict, the result of the magnitude of traffic flow of continuous prediction 30 days and 5 days；Model prediction trend and accuracy are believed as can be seen from Figure 4
It ceases, the prediction case of detailed comparisons' traffic flow is shown in Fig. 5, the method for the invention is obtained by the result that Fig. 4 and Fig. 5 are shown
Prediction effect is good, has effectively tracked the variation of traffic flow, in the case where traffic throat floater, will be obvious that biggish mistake
Difference, to judge that traffic throat floater is laid a good foundation.
Choose 5 days prediction data and check TMRE distribution map, as shown in fig. 6, most TMRE all fall in ± 0.4 it
Between, therefore the threshold value of Traffic Anomaly may be configured as 0.4；That is the case where TMRE > 0.4 is traffic throat floater, in Fig. 6 above threshold line and
The data of lower section are abnormal numerical value.
Claims (10)
1. a kind of charge station's magnitude of traffic flow method for detecting abnormality, which comprises the following steps:
Step 1, the data of each disengaging highway by charge station's vehicle are handled；
Step 101 current data original to the vehicle of charge station pre-process, and carry out data cleansing, data compaction, data row
Sequence and data on flows timesharing accounting；
Step 102 generates magnitude of traffic flow time series data and handles it, first from the original current number of vehicle of charge station
According to generation magnitude of traffic flow time series data；Magnitude of traffic flow time series data is normalized again and standardization；
Step 2, based on the forecasting traffic flow of DBN: using step 1 time series data generated as input, next time
The magnitude of traffic flow of section is output, is predicted using DBN model；
Step 3, by calculating the relative error of the traffic flow of prediction and the actual traffic stream of sliding-model control, judge described opposite
Whether error is more than abnormal threshold value, if the relative error is more than threshold value, the current magnitude of traffic flow of the charge station occurs
Exception, if the relative error is no more than threshold value, which is under normal circumstances.
2. charge station's magnitude of traffic flow method for detecting abnormality according to claim 1, which is characterized in that the data include letter
Cease time, the charge station for passing in and out highway and toll amount, in step 101, remove abnormal Value Data, missing Value Data with it is bright
Aobvious problem data.
3. charge station's magnitude of traffic flow method for detecting abnormality according to claim 2, which is characterized in that in step 101, from receipts
Take in the database at station and select 4 fields: time of entering the station, outbound time enter the station station name and outbound station name is used to according to the time period
Accurate statistics charge station's magnitude of traffic flow.
4. charge station's magnitude of traffic flow method for detecting abnormality according to claim 2, which is characterized in that, will be through in step 102
The data for crossing removing exceptional value, obvious problem data and missing values are in chronological sequence sequentially ranked up；When counting each feature
Between the number of vehicles that passes through in interval, characteristic time interval is 1 hour.
5. charge station's magnitude of traffic flow method for detecting abnormality according to claim 1, which is characterized in that in step 102, from step
The operating process of rapid 101 treated charge station's Raw Data Generation magnitude of traffic flow time series datas is as follows:
Model parameter is first set, and the model parameter includes test set quantity, characteristic and the step-length of input model；
Regeneration training set is output and input, and is inputted as m*n matrix, m is number of samples, and n is characterized number, is exported as m*s square
Battle array, m is number of samples, and s is step-length.
6. charge station's magnitude of traffic flow method for detecting abnormality according to claim 5, which is characterized in that in step 102, will hand over
The same Feature Mapping of the training set different data of through-current capacity time series data is between [0,1], using min-max normalizing
Change, operating method is as follows:
Wherein xmin=Min (x1,x2,…,xm), xmax=Max (x1,x2,…,xm)。
7. charge station's magnitude of traffic flow method for detecting abnormality according to claim 6, which is characterized in that in step 102, to friendship
The different feature of the training set data of through-current capacity time series data carries out stretching operation, so that the not spy between homometric(al)
Sign is comparable,
Wherein x1For xjIn first value.
8. charge station's magnitude of traffic flow method for detecting abnormality according to claim 1, which is characterized in that in step 2, DBN mould
Type is divided into two training module, prediction module parts, and training module is trained given data, learning time sequence data
Feature generates the weight parameter of model and saves in the form of a file, and later, prediction model is by stress model and uses time sequence
Column data predicts the next hour magnitude of traffic flow of charge station.
9. charge station's magnitude of traffic flow method for detecting abnormality according to claim 1, which is characterized in that in step 2, training mould
The running frequency of block is set as in 1 time/January, pressing time series generation method in the present invention first in the operation of prediction module, will be former
The time series data that the processing of beginning charge data is 24 hours in the past, prediction model use 24 hours in the past time series datas
Predict lower traffic flow in 1 hour, the running frequency of prediction model is 1 time/1 hour.
10. charge station's magnitude of traffic flow method for detecting abnormality according to claim 1, which is characterized in that discrete in step 3
Change actual traffic data on flows, calculation formula is as follows:
Wherein, FpIt is the traffic flow of prediction, FmIt is actual traffic flow, n is the time interval for counting flow, i.e., counted by n minutes
The magnitude of traffic flow simultaneously calculates relative error.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811341385.2A CN109255956A (en) | 2018-11-12 | 2018-11-12 | A kind of charge station's magnitude of traffic flow method for detecting abnormality |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811341385.2A CN109255956A (en) | 2018-11-12 | 2018-11-12 | A kind of charge station's magnitude of traffic flow method for detecting abnormality |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109255956A true CN109255956A (en) | 2019-01-22 |
Family
ID=65043294
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811341385.2A Pending CN109255956A (en) | 2018-11-12 | 2018-11-12 | A kind of charge station's magnitude of traffic flow method for detecting abnormality |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109255956A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111882858A (en) * | 2020-06-01 | 2020-11-03 | 重庆大学 | Method for predicting queuing length of expressway abnormal events based on multi-source data |
| CN114495498A (en) * | 2022-01-20 | 2022-05-13 | 青岛海信网络科技股份有限公司 | Traffic data distribution effectiveness judging method and device |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3007019B2 (en) * | 1995-04-07 | 2000-02-07 | 三菱電機株式会社 | Traffic flow measurement device |
| CN101488284A (en) * | 2008-01-16 | 2009-07-22 | 闵万里 | Intelligent management system for road traffic condition instant prediction |
| CN102129776A (en) * | 2011-04-28 | 2011-07-20 | 北京市劳动保护科学研究所 | Automatic detection method and system of abnormal pedestrian traffic state |
| CN102819956A (en) * | 2012-06-05 | 2012-12-12 | 浙江大学 | Detecting method for road traffic accident on basis of single-section annular coil detector |
| CN103021176A (en) * | 2012-11-29 | 2013-04-03 | 浙江大学 | Discriminating method based on section detector for urban traffic state |
| KR20140028801A (en) * | 2012-08-30 | 2014-03-10 | 경희대학교 산학협력단 | Prediction of urban congestion using its based data |
| CN105702029A (en) * | 2016-02-22 | 2016-06-22 | 北京航空航天大学 | Express way traffic state prediction method taking spatial-temporal correlation into account at different times |
| CN105894808A (en) * | 2014-12-03 | 2016-08-24 | 北京旺德瑞通科技发展有限公司 | Method and device for detecting traffic incident |
| CN106816008A (en) * | 2017-02-22 | 2017-06-09 | 银江股份有限公司 | A kind of congestion in road early warning and congestion form time forecasting methods |
| JP2018142334A (en) * | 2018-04-02 | 2018-09-13 | 株式会社デンソー | Vehicle information processing system, on-vehicle device and information processing device |
-
2018
- 2018-11-12 CN CN201811341385.2A patent/CN109255956A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3007019B2 (en) * | 1995-04-07 | 2000-02-07 | 三菱電機株式会社 | Traffic flow measurement device |
| CN101488284A (en) * | 2008-01-16 | 2009-07-22 | 闵万里 | Intelligent management system for road traffic condition instant prediction |
| CN102129776A (en) * | 2011-04-28 | 2011-07-20 | 北京市劳动保护科学研究所 | Automatic detection method and system of abnormal pedestrian traffic state |
| CN102819956A (en) * | 2012-06-05 | 2012-12-12 | 浙江大学 | Detecting method for road traffic accident on basis of single-section annular coil detector |
| KR20140028801A (en) * | 2012-08-30 | 2014-03-10 | 경희대학교 산학협력단 | Prediction of urban congestion using its based data |
| CN103021176A (en) * | 2012-11-29 | 2013-04-03 | 浙江大学 | Discriminating method based on section detector for urban traffic state |
| CN105894808A (en) * | 2014-12-03 | 2016-08-24 | 北京旺德瑞通科技发展有限公司 | Method and device for detecting traffic incident |
| CN105702029A (en) * | 2016-02-22 | 2016-06-22 | 北京航空航天大学 | Express way traffic state prediction method taking spatial-temporal correlation into account at different times |
| CN106816008A (en) * | 2017-02-22 | 2017-06-09 | 银江股份有限公司 | A kind of congestion in road early warning and congestion form time forecasting methods |
| JP2018142334A (en) * | 2018-04-02 | 2018-09-13 | 株式会社デンソー | Vehicle information processing system, on-vehicle device and information processing device |
Non-Patent Citations (3)
| Title |
|---|
| MENGJIAO QIN, ZHIHANG LI, ZHENHONG DU: "Red tide time series forecasting by combining ARIMA and deep belief network", 《KNOWLEDGE-BASED SYSTEMS》 * |
| 江德浩: "基于深度信念网络的短时交通流预测", 《中国优秀硕士学位论文全文数据库 工程科技II辑》 * |
| 韩坤林: "基于车检器及收费数据融合的高速公路异常状态识别研究", 《中国优秀硕士学位论文全文数据库 工程科技II辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111882858A (en) * | 2020-06-01 | 2020-11-03 | 重庆大学 | Method for predicting queuing length of expressway abnormal events based on multi-source data |
| CN114495498A (en) * | 2022-01-20 | 2022-05-13 | 青岛海信网络科技股份有限公司 | Traffic data distribution effectiveness judging method and device |
| CN114495498B (en) * | 2022-01-20 | 2023-01-10 | 青岛海信网络科技股份有限公司 | Traffic data distribution effectiveness judging method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107085943B (en) | Short-term prediction method and system for road travel time | |
| CN111144039B (en) | Train dynamic weighing system and weighing method based on deep learning | |
| CN109871876B (en) | Expressway road condition identification and prediction method based on floating car data | |
| CN104778837A (en) | Multi-time scale forecasting method for road traffic running situation | |
| CN115240431B (en) | Real-time online simulation system and method for traffic flow of highway toll station | |
| CN114049765B (en) | Urban road network traffic flow OD estimation method based on automatic vehicle number plate identification data | |
| CN110299011A (en) | A kind of traffic flow forecasting method of the highway arbitrary cross-section based on charge data | |
| CN111951553B (en) | Prediction method based on traffic big data platform and mesoscopic simulation model | |
| CN109345832B (en) | Urban road overtaking prediction method based on deep recurrent neural network | |
| CN111724589A (en) | Multi-source data-based highway section flow estimation method | |
| CN114596700B (en) | Real-time traffic estimation method for expressway section based on portal data | |
| CN107564290A (en) | A kind of urban road intersection saturation volume rate computational methods | |
| CN111063204B (en) | Expressway vehicle speed prediction model training method based on toll station flow | |
| CN113159374B (en) | Data-driven urban traffic flow rate mode identification and real-time prediction early warning method | |
| CN107025468A (en) | Highway congestion recognition methods based on PCA GA SVM algorithms | |
| Yao et al. | The effect of image recognition traffic prediction method under deep learning and naive Bayes algorithm on freeway traffic safety | |
| CN115762169B (en) | Unmanned intelligent control system and method for sanitation vehicle | |
| CN112767684A (en) | Highway traffic jam detection method based on charging data | |
| CN111724592B (en) | Highway traffic jam detection method based on charging data and checkpoint data | |
| CN115691120A (en) | Congestion identification method and system based on highway running water data | |
| CN109255956A (en) | A kind of charge station's magnitude of traffic flow method for detecting abnormality | |
| Wang et al. | Vehicle reidentification with self-adaptive time windows for real-time travel time estimation | |
| CN116631186A (en) | Expressway traffic accident risk assessment method and system based on dangerous driving event data | |
| CN112860782A (en) | Pure electric vehicle driving range estimation method based on big data analysis | |
| CN114882069A (en) | Taxi track abnormity detection method based on LSTM network and attention mechanism |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190122 |