CN110866538B - Vehicle trajectory data acquisition method and device, computer equipment and storage medium - Google Patents
Vehicle trajectory data acquisition method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN110866538B CN110866538B CN201910953306.1A CN201910953306A CN110866538B CN 110866538 B CN110866538 B CN 110866538B CN 201910953306 A CN201910953306 A CN 201910953306A CN 110866538 B CN110866538 B CN 110866538B
- Authority
- CN
- China
- Prior art keywords
- regression
- classification
- gps
- data
- samples
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/251—Fusion techniques of input or preprocessed data
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Traffic Control Systems (AREA)
- Navigation (AREA)
Abstract
The application relates to a vehicle track data acquisition method, a vehicle track data acquisition device, computer equipment and a storage medium, wherein the method comprises the following steps: the method comprises the steps of obtaining original regression samples carrying multiple sections of vehicle track data, carrying a GPS and OBD fusion data set when a GPS is effective and an OBD data set when the GPS is invalid in the original regression samples, converting the original regression samples into classification samples through an incremental regression frame, selecting support vector classification as a basic classifier to conduct repeated iterative training on the classification samples, screening samples harmful to a current scene through a transfer learning method in the iterative training process, reducing the weight of the samples harmful to the current scene, classifying and integrating support vectors generated by iterative training according to a weight voting mechanism, obtaining a regression model based on integrated support vector regression through a common point of the support vector classification and the support vector regression, constructing a nonlinear track interpolation model based on a general kernel function, and accurately obtaining vehicle track interpolation data.
Description
Technical Field
The present application relates to the field of data processing technologies, and in particular, to a method and an apparatus for acquiring vehicle trajectory data, a computer device, and a storage medium.
Background
With the development of economic level, the vehicle inventory is rapidly increased year by year at present, and the huge number of vehicles which are gradually increased year by year attracts countless scholars to carry out research on the vehicles, wherein the research mainly comprises the research on the social attributes of vehicle owners, the research on driving behaviors, the influence of vehicles on urban traffic and the like.
Various vehicle-based studies are primarily based on large-scale trajectory data generated when a private car travels on a road. The acquisition of vehicle trajectory data, which is most often also commonly used, relies on various types of motion sensors, for example, by installing a GPS (Global Positioning System) device on a private car.
At present, most of the GPS loaded on the vehicle is low-cost GPS, and the GPS is easy to have the defect of GPS signal interruption. When the GPS fails, a large amount of trajectory information is lost, which makes vehicle trajectory mining research difficult.
Disclosure of Invention
In view of the above, it is necessary to provide a vehicle trajectory data acquisition method, apparatus, computer device, and storage medium capable of acquiring continuous trajectory data in the event of a GPS signal interruption, in view of the above technical problems.
A vehicle trajectory data acquisition method, the method comprising:
acquiring an original regression sample carrying a plurality of sections of vehicle track data, wherein the original regression sample carries a GPS and OBD (On-Board Diagnostic) fusion data set when the GPS is effective and an OBD data set when the GPS is invalid;
converting the raw regression samples into classification samples by an incremental regression (R2C model) framework;
selecting Support Vector Classification (SVC) as a basic classifier to carry out multiple iterative training on the classified samples, and screening samples harmful to the current scene by a transfer learning method (TrAdaboost) in the iterative training process to reduce the weight of the samples harmful to the current scene;
according to a weight voting mechanism, classifying and integrating support vectors generated by iterative training, and obtaining a regression model based on integrated support vector regression through a common point of support vector classification and support vector regression;
and constructing a nonlinear track interpolation model based on the general kernel function and according to the regression model to obtain vehicle track interpolation data.
In one embodiment, the obtaining raw regression samples carrying multiple pieces of vehicle trajectory data includes:
acquiring original GPS data and OBD data of a vehicle, and fusing the original GPS data and the OBD data of the vehicle to obtain a GPS and OBD fused data set when a GPS is effective and an OBD data set when the GPS is ineffective;
and connecting the GPS and OBD fusion data set when the GPS is effective and the OBD data set when the GPS is invalid and carrying out blocking processing to obtain an original regression sample carrying multiple sections of vehicle track data.
In one embodiment, the converting the raw regression samples into classification samples by an incremental regression framework comprises:
identifying response variables in the raw regression samples;
moving the response variable up and down by a preset amount
And constructing a classification sample.
In one embodiment, after the converting the original regression samples into classification samples by the incremental regression framework, the method further includes:
respectively constructing classification type loss functions on two data sets carried by the original regression sample to obtain two classification type loss functions;
the selecting the support vector classification as a basic classifier to perform multiple iterative training on the classification sample comprises:
and obtaining a series of support vector classification classifiers by iteration according to the two classification type loss functions, wherein the weight of each support vector classification classifier is correspondingly changed along with the increase of the iteration times.
In one embodiment, in the iterative training process, samples harmful to the current scene are screened through a transfer learning method, and reducing the weight of the samples harmful to the current scene includes:
acquiring a preset sample correlation factor beta, wherein the preset sample correlation factor beta is defined on a GPS and OBD fusion data set when the GPS is effective through a transfer learning method, and the preset sample correlation factor beta represents the relationship strength between a certain sample and a target sample;
and adjusting the weight vector of each support vector classification in the iterative training process through the preset sample correlation factor beta.
In one embodiment, the integrating the support vector classifications generated by the iterative training according to the weight voting mechanism, and obtaining the regression model based on the integrated support vector regression through the common point of the support vector classification and the support vector regression includes:
according to a weight voting mechanism, carrying out classified integration on the support vectors generated by iterative training to obtain a weighted integration model which is most suitable for a current sample;
and solving the corresponding equation of the weighted integration model to obtain a linear regression model based on an incremental regression and transfer learning method.
In one embodiment, the constructing a nonlinear trajectory interpolation model based on the general kernel function and according to the regression model to obtain vehicle trajectory interpolation data includes:
acquiring a preset general kernel function;
carrying out support vector classification training on the general kernel function to obtain a decision function;
obtaining an integrated classification plane according to the regression model and the decision function;
converting the ensemble classification plane to a regression classification plane;
solving the regression classification plane, and constructing a nonlinear track interpolation model;
optimizing parameters in the nonlinear track interpolation model through a Bayesian optimization algorithm;
and obtaining vehicle track interpolation data according to the nonlinear track interpolation model after parameter optimization.
A vehicle trajectory data acquisition device, the device comprising:
the system comprises a data set acquisition module, a data processing module and a data processing module, wherein the data set acquisition module is used for acquiring an original regression sample carrying a plurality of sections of vehicle track data, and the original regression sample carries a GPS and OBD fusion data set when a GPS is effective and an OBD data set when the GPS is invalid;
a conversion module for converting the original regression sample into a classification sample through an incremental regression frame;
the iterative training module is used for selecting the support vector classification as a basic classifier to carry out multiple iterative training on the classified samples, screening the samples harmful to the current scene through a transfer learning method in the iterative training process, and reducing the weight of the samples harmful to the current scene;
the regression model building module is used for classifying and integrating the support vectors generated by iterative training according to a weight voting mechanism and obtaining a regression model based on integrated support vector regression through a common point of support vector classification and support vector regression;
and the track data acquisition module is used for constructing a nonlinear track interpolation model based on the general kernel function and according to the regression model to obtain vehicle track interpolation data.
A computer device comprising a memory storing a computer program and a processor implementing the steps of the method as described above when executing the computer program.
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method as described above.
The vehicle track data acquisition method, the vehicle track data acquisition device, the computer equipment and the storage medium acquire an original regression sample carrying a plurality of sections of vehicle track data, the original regression sample carries a GPS and OBD fusion data set when the GPS is effective and an OBD data set when the GPS is invalid, the original regression sample is converted into a classification sample through an incremental regression frame, the regression problem is converted into a classification problem, a support vector classification is selected as a basic classifier to carry out repeated iterative training on the classification sample, a sample harmful to the current scene is screened through a transfer learning method in the iterative training process, the weight of the sample harmful to the current scene is reduced, the support vector generated by the iterative training is classified and integrated according to a weight voting mechanism, and a regression model based on integrated support vector regression is obtained through a common point of the support vector classification and the support vector regression, and constructing a nonlinear track interpolation model based on the general kernel function, and accurately obtaining vehicle track interpolation data based on a transfer learning method and an incremental regression frame.
Drawings
FIG. 1 is a diagram of an exemplary vehicle trajectory data acquisition method;
FIG. 2 is a schematic flow chart diagram of a vehicle trajectory data acquisition method in one embodiment;
FIG. 3 is a schematic flow chart diagram illustrating a vehicle trajectory data acquisition method according to another embodiment;
FIG. 4 is a block diagram of an overall flowchart of a vehicle trajectory data acquisition method according to an embodiment;
FIG. 5 is a schematic illustration of a road segment with GPS dead time and corresponding missing track in one embodiment;
FIG. 6 is a diagram illustrating various scenarios requiring trajectory interpolation;
FIG. 7 is a plot of the accumulated error in longitude for a trace interpolation comparison experiment;
FIG. 8 is a plot of the cumulative error at latitude for a trace interpolation contrast experiment;
FIG. 9 is a schematic diagram showing the configuration of a vehicle trajectory data acquisition device according to an embodiment;
FIG. 10 is a diagram showing an internal structure of a computer device according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The vehicle track data acquisition method provided by the application can be applied to the application environment shown in fig. 1. The terminal 102 communicates with the server 104 through a network, and the terminal 102 may be connected to the vehicle-mounted GPS and the vehicle-mounted OBD, respectively, receive data of the vehicle-mounted GPS and the vehicle-mounted OBD, and upload the data to the server 104. The server 104 obtains an original regression sample carrying a plurality of pieces of vehicle track data, the original regression sample carries a GPS and OBD fusion data set when the GPS is effective and an OBD data set when the GPS is ineffective, converting the original regression sample into a classification sample through an incremental regression frame, converting the regression problem into a classification problem, selecting support vector classification as a basic classifier to carry out multiple iterative training on the classification sample, and in the iterative training process, samples harmful to the current scene are screened by a transfer learning method, the weight of the samples harmful to the current scene is reduced, and (3) classifying and integrating the support vectors generated by iterative training according to a weight voting mechanism, obtaining a regression model based on integrated support vector regression through a common point of support vector classification and support vector regression, and constructing a nonlinear track interpolation model based on a general kernel function to obtain vehicle track interpolation data. The terminal 102 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices, and the server 104 may be implemented by an independent server or a server cluster formed by a plurality of servers.
In one embodiment, as shown in fig. 2, a vehicle trajectory data acquisition method is provided, which is described by taking the method as an example applied to the server in fig. 1, and includes the following steps:
s100: and acquiring an original regression sample carrying a plurality of sections of vehicle track data, wherein the original regression sample carries a GPS and OBD fusion data set when the GPS is effective and an OBD data set when the GPS is invalid.
The OBD is a vehicle-mounted diagnosis system which monitors the running condition of an engine and the working state of an exhaust aftertreatment system at any time, and immediately sends out a warning once the condition that the emission exceeds the standard is found, when the system breaks down, a fault lamp or a warning lamp for checking the engine is turned on, meanwhile, the OBD system stores fault information into a memory, relevant information can be read in the form of fault codes through a standard diagnosis instrument and a diagnosis interface, and maintenance personnel can quickly and accurately determine the nature and the position of the fault according to the prompt of the fault codes. Specifically, the server can receive vehicle GPS data and OBD data which are collected by the external terminal, and the data are screened and classified into GPS and OBD fusion data when the GPS is effective and OBD data when the GPS is invalid. In this embodiment, the purpose of acquiring both GPS and OBD types of data is to hopefully refine the vehicle trajectory data in the event of GPS failure with the OBD data.
As shown in fig. 3, in one embodiment, step S100 includes:
s120: and acquiring original GPS data and OBD data of the vehicle, and fusing the original GPS data and the OBD data of the vehicle to obtain a GPS and OBD fusion data set when the GPS is effective and an OBD data set when the GPS is invalid.
S140: and connecting the GPS and OBD fusion data set when the GPS is effective and the OBD data set when the GPS is invalid and carrying out blocking processing to obtain an original regression sample carrying a plurality of sections of vehicle track data.
The data are divided into two blocks by fusing the GPS data and the OBD data, wherein one block is the GPS data and the OBD data when the GPS is effective, and the other block is the only OBD number when the GPS is invalid; and then connecting the two pieces of data, and performing blocking operation to obtain multiple pieces of track data. Specifically, the original regression sample data consists of two pieces of data: GPS/OBD dataset with GPS enabled time
And the only OBD dataset when GPS fails
The series of regression data obtained after the blocking operation is as follows:
SR＝{(Xi,yi)|Xi∈Rn+m,yi∈R,i＝1,2,...,n+m}
s200: the raw regression samples are converted to classification samples by an incremental regression framework.
And converting the regression problem corresponding to the original regression sample into a classification problem corresponding to the classification sample through an incremental regression frame. Specifically, the conversion process is as follows: identifying response variables in the original regression samples; moving the response variable up and down by a preset amount
And constructing a classification sample. The regression problem can be converted to a classification problem by constructing a categorical loss function on the specified data set after constructing the classification sample. The specified data set refers to a GPS and OBD fusion data set when the GPS is effective and an OBD data set when the GPS is invalid.
The process of converting the regression sample into the classification sample in the above step S200 will be described in detail based on the mathematical formula.
In the step of S200, the response variable y of the original regression sample is moved up and down
The following classification samples were constructed:
the relationship between the original regression sample and the new classification label at this time can be expressed as:
the classification type loss at this time is designed to be only at
The calculated error rate of (c) is as follows:
where p is a probability distribution,
is a classification model F
The above-mentioned predicted value is obtained,
representing the true value.
S300: and selecting support vector classification as a basic classifier to carry out repeated iterative training on the classified samples, and screening the samples harmful to the current scene by a transfer learning method in the iterative training process to reduce the weight of the samples harmful to the current scene.
Processing in step S300 desirably migrates the OBD data onto the GPS data through iterative training and migration learning algorithms to ultimately enable accurate vehicle trajectory interpolation data. The harmful sample is sample data which is irrelevant to the current scene and interferes with the final finding of the target sample, and the sample data can be understood as noise data. In practical application, the weights corresponding to all classified samples in the iterative training process can be continuously adjusted through a transfer learning method, the weights of harmful samples in the samples are reduced, and the weights of related samples are improved, so that the purpose of transferring the OBD information to the GPS information is achieved.
S400: and (3) according to a weight voting mechanism, classifying and integrating the support vectors generated by iterative training, and obtaining a regression model based on integrated support vector regression through the common point of support vector classification and support vector regression.
And integrating a series of support vector classifications obtained by iterative training in the step S300 according to a weight voting mechanism, and then obtaining a regression model based on integrated support vector regression through a common point of the support vector classifications and the support vector regression.
Specifically, in step S400, there is a normal vector
The following classification models were constructed:
where b is the deviation. According to the weight voting mechanism, a series of support vector classifications obtained after step S300 is completed are integrated to obtain a weighted integration model most suitable for the current sample, as follows:
wherein W is the classifier weight vector obtained after updating the sample weight, PXIs input with a sampleProbability distribution of feature vector X, pyIs the probability of the response vector in the original regression sample. At this time, the information of the response variable y of the original regression that we need is also included, and solving this equation can obtain a new linear regression model based on R2C and the transfer learning method:
s500: and constructing a nonlinear track interpolation model based on the general kernel function and according to the regression model to obtain vehicle track interpolation data.
For a non-linear regression model, a kernel function needs to be introduced. However, since we change the structure of the original regression model, the commonly used kernel function cannot be directly applied in the framework, and the corresponding change should be made. Common kernel functions include linear kernel functions (RBF) kernel functions, and Sigmoid kernel functions. And obtaining an integrated classification plane by the regression model and the general kernel function, converting the integrated classification plane into a regression classification plane, solving an equation of the regression classification plane, and constructing a nonlinear track interpolation model to obtain vehicle track interpolation data.
The vehicle track data acquisition method comprises the steps of acquiring an original regression sample carrying a plurality of sections of vehicle track data, carrying a GPS and OBD fusion data set when the GPS is effective and an OBD data set when the GPS is invalid in the original regression sample, converting the original regression sample into a classification sample through an incremental regression frame, converting the regression problem into a classification problem, selecting support vector classification as a basic classifier to carry out repeated iterative training on the classification sample, screening a sample harmful to the current scene through a transfer learning method in the iterative training process, reducing the weight of the sample harmful to the current scene, integrating the support vector generated by the iterative training according to a weight voting mechanism, obtaining a regression model based on integrated support vector regression through a common point of the support vector classification and the support vector regression, and constructing a nonlinear track interpolation model based on a universal kernel function, and accurately obtaining vehicle track interpolation data based on a transfer learning method and an incremental regression frame.
In one embodiment, after converting the original regression samples into classification samples through the incremental regression framework, the method further includes: respectively constructing classification type loss functions on two data sets carried by an original regression sample to obtain two classification type loss functions; selecting support vector classification as a basic classifier to perform multiple iterative training on classification samples comprises the following steps: and obtaining a series of support vector classification classifiers by iteration according to the two classification type loss functions, wherein the weight of each support vector classification classifier is correspondingly changed along with the increase of the iteration times.
And respectively constructing classification type loss functions on a GPS and OBD fusion data set when the GPS is effective and an OBD data set when the GPS is invalid to obtain two classification type loss functions. Through iteration, a series of support vector classification classifiers are obtained, and the weight of each support vector classification changes correspondingly with the increase of the iteration times.
In one embodiment, in the iterative training process, samples harmful to the current scene are screened through a transfer learning method, and reducing the weight of the samples harmful to the current scene includes:
acquiring a preset sample correlation factor beta, wherein the preset sample correlation factor beta is obtained by defining on a GPS and OBD fusion data set when the GPS is effective through a transfer learning method, and the preset sample correlation factor beta represents the relationship strength between a certain sample and a target sample; and adjusting the weight vector of each support vector classification in the iterative training process by presetting a sample correlation factor beta.
Using TrAdboost in
In the above, a sample correlation factor β is defined, which is expressed as the strength of the relationship between a certain sample and the target sample. Therefore, in each iteration, the effect of irrelevant samples is weakened, and samples harmful to the current application scene are filtered out. The weight update at this time can be described as:
wherein ω isSIs a weight vector that represents the weights of the series of support vector classifications being integrated. The samples obtained after the updating process all contain information which is strongly related to the current application scene.
In one embodiment, the step S500 specifically includes:
acquiring a preset general kernel function; carrying out support vector classification training on the general kernel function to obtain a decision function; obtaining an integrated classification plane according to the regression model and the decision function; converting the integrated classification plane into a regression classification plane; solving a regression classification plane, and constructing a nonlinear track interpolation model; optimizing parameters in the nonlinear track interpolation model through a Bayesian optimization algorithm; and obtaining vehicle track interpolation data according to the nonlinear track interpolation model after parameter optimization.
Usually, a kernel function is introduced, and after training of support vector classification, the following decision function is obtained:
wherein K represents a kernel function, C is a compensation factor, and
is a support vector trained by support vector classification. Correspondingly, the integrated classification plane can be expressed as:
in the present framework, with linear kernel function
For example, one can obtain:
it is noted that all information obtained from training is implicit in the support vector
Therefore, by solving this equation, we can get the final nonlinear regression model. The regression kernel space obtained after the solution is not the same as the kernel space of the linear classification, so the original kernel function cannot be directly applied to the regression framework. The resulting nonlinear integrated regression model is shown below:
through matrix calculation, the following equation is obtained after the above formula is transformed:
with this equation, it can be seen that the output response variables of the regression model are also included, so the final integrated model can be obtained by solving:
similarly, the final model can be derived using two other commonly used kernel functions. In addition, in order to automatically optimize and adjust the related parameters, Bayesian optimization is adopted for parameter selection. The optimization model is assumed to be:
x＝argmaxx∈Xf(x)
where X is the candidate set of parameter combinations and f (X) represents the accuracy of the model. The goal of the model is to choose an X from X such that the value of f (X) is maximized, which can be done in two steps. First, a prior model, usually a Gaussian process, is selected; the posterior distribution is then obtained by calculation. Before we can do this we must assume that f (x) is gaussian-distributed and can be described as:
f(x)～GP((μ(x),k(x,x′))
where μ (x) is the mean function and k (x, x') represents the covariance function. Then, the Acquisition function (Acquisition function) is used to find the next suitable sample point. We select the current mainstream's acquired Improvement (EI) function, which belongs to the expectation function, and takes into account the relationship of f (x) and f (x). Where f (X) represents the objective function value of X that has been optimized. The optimization model at this time can be expressed as:
x＝argmaxx∈XE(max0,fk+i(x)-f(x*)|Sk)
where k is the number of samples, SkThe first k samples are represented. In combination with the prior assumption of high-dimensional normal distribution, we can obtain the acquisition model as follows:
where Φ and Φ are the PDF and CDF of a normal distribution, respectively, μ (x) and σ (x) are expressed as the mean and variance of the objective function by GP, respectively, and Z is equal to the ratio of (μ (x) -f (x)) to σ (x). By the Bayes optimization model, the hyper-parameters related by the application can be automatically optimized, adjusted and acquired. The hyper-parameters involved are: window size (block size), compensation factor C in SVM, and parameters in various types of kernel functions, such as γ, d, and r in polynomial kernel functions.
As shown in fig. 4, in one application example, the vehicle trajectory data acquisition method of the present application may specifically be a scheme for acquiring large-scale private car trajectory interpolation data based on transfer learning and incremental regression, and the scheme is mainly divided into four blocks: the data fusion part fuses the extracted GPS data and the OBD data, and divides the input data into two sources, namely the GPS data and the OBD data (the GPS is not failed) and the OBD data only (the GPS is failed). The former only serves as training data, and the latter serves as both training data and test data. While determining the window size (block size) this data is blocked. Secondly, the data is sent to an R2C module for training. After the number of iterations is determined, the trained Base learning machine (Base leaner) will be updated many times. And then solving the weight value of the corresponding basic learning machine through the error rate, and integrating a series of basic learning machines by combining a weight voting mechanism. Thirdly, in the iteration process, TrAdaboost is used for acting on samples with poor classification results. The quality of the classification result is related to the correlation between the sample and the target, and samples with worse classification results are less correlated with the samples required by the application scene, so that the information does not need to be migrated, and the samples are filtered. And fourthly, automatically selecting and adjusting parameters by using Bayesian optimization, and then converting the classification hyperplane obtained after R2C and TrAdaboost training into a regression hyperplane to obtain a track interpolation model, and finally obtaining a complete track data set.
Fig. 5 is a data set of a trajectory involved in the interpolation algorithm for verification according to the present invention (where the white trajectory with end points is a GPS failure trajectory), where the data set is a complex section of sand including a straight line, an overpass, a high city speed, a low city speed, a curve in a city, and a continuous right-angled curve in a city. The GPS failure and the missing track of the road sections are forced to occur, and the specific details are shown in FIG. 6, wherein a straight line is data which is missing for 36 seconds and passes through an entrance where a traffic flow enters and an intersection where the traffic flow exits; the viaduct is the most complex scene in the verification test, is set to be a track which is missing for 68 seconds, and is extremely complex due to the fact that the viaduct has continuous ascending and continuous descending and a plurality of entrances and exits; the urban highway section selects the missing track of 38 seconds; the city selects school road segments at low speed, and 40 seconds of missing track occurs. Controllable sidewalk traffic lights for students to freely select are arranged at the front and the back of the road section, so that vehicles often step on and stop in the road section, and the track interpolation has the trouble of time delay. The urban curve selects a curve section with a higher speed, and the missing track time is 44 seconds. Since the speed of the road segment is high and the direction angle of the road segment changes, a large deviation may occur in the track interpolation task. And finally, a continuous right-angle bend, wherein the failure time is 58 seconds. The road section is also an internal road section of a school, college buildings and trees are shielded at two sides of the road section, so that the collected GPS has some deviation, and real difficulty is brought to a track interpolation task.
The effect of the trajectory interpolation method proposed by the present invention is shown in fig. 7 and fig. 8, respectively. The comparison method comprises the following steps: a traditional machine learning method (1), a neural network method (2), an ensemble learning method (3) and an ETR method (4), and an incremental learning method (5) and PAR (6). The experimental evaluation index is the average absolute error. The abscissa of fig. 6 and 7 is the different application scenarios and the ordinate is the value of the mean absolute error. The viaduct and the continuous right-angle curve are scenes with the largest experimental deviation, but the method provided by the invention tends to be stable in performance and still has relatively good effect. In addition, it can also be seen from the comparison of the rightmost averages that the method proposed by the present invention performs overall better than other comparative methods.
It should be understood that although the various steps in the flow charts of fig. 2-3 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least some of the steps in fig. 2-3 may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performance of the sub-steps or stages is not necessarily sequential, but may be performed in turn or alternating with other steps or at least some of the sub-steps or stages of other steps.
In addition, as shown in fig. 9, the present application also provides a vehicle trajectory data acquisition device including:
the data set acquisition module 100 is configured to acquire an original regression sample carrying multiple pieces of vehicle trajectory data, where the original regression sample carries a GPS and OBD fusion data set when the GPS is valid and an OBD data set when the GPS is invalid;
a conversion module 200, configured to convert the original regression sample into a classification sample through an incremental regression frame;
the iterative training module 300 is configured to select support vector classification as a basic classifier to perform multiple iterative training on classified samples, and filter samples harmful to the current scene through a transfer learning method in an iterative training process to reduce weights of the samples harmful to the current scene;
the regression model building module 400 is used for classifying and integrating the support vectors generated by iterative training according to a weight voting mechanism, and obtaining a regression model based on integrated support vector regression through a common point of support vector classification and support vector regression;
and the track data acquisition module 500 is used for constructing a nonlinear track interpolation model based on the general kernel function and according to the regression model to obtain vehicle track interpolation data.
The vehicle track data acquisition device acquires an original regression sample carrying a plurality of sections of vehicle track data, the original regression sample carries a GPS and OBD fusion data set when the GPS is effective and an OBD data set when the GPS is invalid, the original regression sample is converted into a classification sample through an incremental regression frame, the regression problem is converted into a classification problem, a support vector classification is selected as a basic classifier to carry out repeated iterative training on the classification sample, a sample harmful to the current scene is screened through a transfer learning method in the iterative training process, the weight of the sample harmful to the current scene is reduced, the support vector generated by the iterative training is classified and integrated according to a weight voting mechanism, a regression model based on integrated support vector regression is obtained through a common point of the support vector classification and the support vector regression, and a nonlinear track interpolation model is constructed based on a universal kernel function, and accurately obtaining vehicle track interpolation data based on a transfer learning method and an incremental regression frame.
In one embodiment, the data set acquiring module 100 is further configured to acquire vehicle original GPS data and OBD data, and fuse the vehicle original GPS data and the OBD data to obtain a GPS and OBD fused data set when the GPS is valid and an OBD data set when the GPS is invalid; and connecting the GPS and OBD fusion data set when the GPS is effective and the OBD data set when the GPS is invalid and carrying out blocking processing to obtain an original regression sample carrying a plurality of sections of vehicle track data.
In one embodiment, the conversion module 200 is further configured to identify response variables in the original regression samples; moving the response variable up and down by a preset amount
And constructing a classification sample.
In one embodiment, the iterative training module 300 is further configured to respectively construct a classification loss function on two data sets carried by the original regression sample, so as to obtain two classification loss functions; and obtaining a series of support vector classification classifiers by iteration according to the two classification type loss functions, wherein the weight of each support vector classification classifier is correspondingly changed along with the increase of the iteration times.
In one embodiment, the iterative training module 300 is further configured to obtain a preset sample correlation factor β, where the preset sample correlation factor β is defined on a GPS and OBD fusion data set when the GPS is valid by a transfer learning method, and the preset sample correlation factor β represents a relationship strength between a certain sample and a target sample; and adjusting the weight vector of each support vector classification in the iterative training process by presetting a sample correlation factor beta.
In one embodiment, the regression model building module 400 is further configured to classify and integrate the support vectors generated by the iterative training according to a weight voting mechanism to obtain a weighted integration model of the most suitable current sample; and solving a corresponding equation of the weighted integration model to obtain a linear regression model based on the incremental regression and the transfer learning method.
In one embodiment, the trajectory data obtaining module 500 is further configured to obtain a preset generic kernel function; carrying out support vector classification training on the general kernel function to obtain a decision function; obtaining an integrated classification plane according to the regression model and the decision function; converting the integrated classification plane into a regression classification plane; solving a regression classification plane, and constructing a nonlinear track interpolation model; optimizing parameters in the nonlinear track interpolation model through a Bayesian optimization algorithm; and obtaining vehicle track interpolation data according to the nonlinear track interpolation model after parameter optimization.
For specific definition of the vehicle trajectory data acquisition device, reference may be made to the above definition of the vehicle trajectory data acquisition method, which is not described herein again. The modules in the vehicle trajectory data acquisition device can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 10. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used to store historical GPS data and OBD data. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a vehicle trajectory data acquisition method.
Those skilled in the art will appreciate that the architecture shown in fig. 10 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program:
acquiring an original regression sample carrying a plurality of sections of vehicle track data, wherein the original regression sample carries a GPS and OBD fusion data set when a GPS is effective and an OBD data set when the GPS is invalid;
converting the original regression sample into a classification sample through an incremental regression frame;
selecting support vector classification as a basic classifier to carry out repeated iterative training on classified samples, and screening samples harmful to the current scene by a transfer learning method in the iterative training process to reduce the weight of the samples harmful to the current scene;
according to a weight voting mechanism, classifying and integrating support vectors generated by iterative training, and obtaining a regression model based on integrated support vector regression through a common point of support vector classification and support vector regression;
and constructing a nonlinear track interpolation model based on the general kernel function and according to the regression model to obtain vehicle track interpolation data.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
acquiring original GPS data and OBD data of a vehicle, and fusing the original GPS data and the OBD data of the vehicle to obtain a GPS and OBD fused data set when a GPS is effective and an OBD data set when the GPS is ineffective; and connecting the GPS and OBD fusion data set when the GPS is effective and the OBD data set when the GPS is invalid and carrying out blocking processing to obtain an original regression sample carrying a plurality of sections of vehicle track data.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
identifying response variables in the original regression samples; moving the response variable up and down by a preset amount
And constructing a classification sample.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
respectively constructing classification type loss functions on two data sets carried by an original regression sample to obtain two classification type loss functions; and obtaining a series of support vector classification classifiers by iteration according to the two classification type loss functions, wherein the weight of each support vector classification classifier is correspondingly changed along with the increase of the iteration times.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
acquiring a preset sample correlation factor beta, wherein the preset sample correlation factor beta is obtained by defining on a GPS and OBD fusion data set when the GPS is effective through a transfer learning method, and the preset sample correlation factor beta represents the relationship strength between a certain sample and a target sample; and adjusting the weight vector of each support vector classification in the iterative training process by presetting a sample correlation factor beta.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
according to a weight voting mechanism, carrying out classified integration on the support vectors generated by iterative training to obtain a weighted integration model which is most suitable for a current sample; and solving a corresponding equation of the weighted integration model to obtain a linear regression model based on the incremental regression and the transfer learning method.
In one embodiment, the processor, when executing the computer program, further performs the steps of:
acquiring a preset general kernel function; carrying out support vector classification training on the general kernel function to obtain a decision function; obtaining an integrated classification plane according to the regression model and the decision function; converting the integrated classification plane into a regression classification plane; solving a regression classification plane, and constructing a nonlinear track interpolation model; optimizing parameters in the nonlinear track interpolation model through a Bayesian optimization algorithm; and obtaining vehicle track interpolation data according to the nonlinear track interpolation model after parameter optimization.
In one embodiment, a computer-readable storage medium is provided, having a computer program stored thereon, which when executed by a processor, performs the steps of:
acquiring an original regression sample carrying a plurality of sections of vehicle track data, wherein the original regression sample carries a GPS and OBD fusion data set when a GPS is effective and an OBD data set when the GPS is invalid;
converting the original regression sample into a classification sample through an incremental regression frame;
selecting support vector classification as a basic classifier to carry out repeated iterative training on classified samples, and screening samples harmful to the current scene by a transfer learning method in the iterative training process to reduce the weight of the samples harmful to the current scene;
according to a weight voting mechanism, classifying and integrating support vectors generated by iterative training, and obtaining a regression model based on integrated support vector regression through a common point of support vector classification and support vector regression;
and constructing a nonlinear track interpolation model based on the general kernel function and according to the regression model to obtain vehicle track interpolation data.
In one embodiment, the computer program when executed by the processor further performs the steps of:
acquiring original GPS data and OBD data of a vehicle, and fusing the original GPS data and the OBD data of the vehicle to obtain a GPS and OBD fused data set when a GPS is effective and an OBD data set when the GPS is ineffective; and connecting the GPS and OBD fusion data set when the GPS is effective and the OBD data set when the GPS is invalid and carrying out blocking processing to obtain an original regression sample carrying a plurality of sections of vehicle track data.
In one embodiment, the computer program when executed by the processor further performs the steps of:
identifying response variables in the original regression samples; moving the response variable up and down by a preset amount
And constructing a classification sample.
In one embodiment, the computer program when executed by the processor further performs the steps of:
respectively constructing classification type loss functions on two data sets carried by an original regression sample to obtain two classification type loss functions; and obtaining a series of support vector classification classifiers by iteration according to the two classification type loss functions, wherein the weight of each support vector classification classifier is correspondingly changed along with the increase of the iteration times.
In one embodiment, the computer program when executed by the processor further performs the steps of:
acquiring a preset sample correlation factor beta, wherein the preset sample correlation factor beta is obtained by defining on a GPS and OBD fusion data set when the GPS is effective through a transfer learning method, and the preset sample correlation factor beta represents the relationship strength between a certain sample and a target sample; and adjusting the weight vector of each support vector classification in the iterative training process by presetting a sample correlation factor beta.
In one embodiment, the computer program when executed by the processor further performs the steps of:
according to a weight voting mechanism, carrying out classified integration on the support vectors generated by iterative training to obtain a weighted integration model which is most suitable for a current sample; and solving a corresponding equation of the weighted integration model to obtain a linear regression model based on the incremental regression and the transfer learning method.
In one embodiment, the computer program when executed by the processor further performs the steps of:
acquiring a preset general kernel function; carrying out support vector classification training on the general kernel function to obtain a decision function; obtaining an integrated classification plane according to the regression model and the decision function; converting the integrated classification plane into a regression classification plane; solving a regression classification plane, and constructing a nonlinear track interpolation model; optimizing parameters in the nonlinear track interpolation model through a Bayesian optimization algorithm; and obtaining vehicle track interpolation data according to the nonlinear track interpolation model after parameter optimization.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware related to instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above examples only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A vehicle trajectory data acquisition method, the method comprising:
acquiring an original regression sample carrying a plurality of sections of vehicle track data, wherein the original regression sample carries a GPS and OBD fusion data set when a GPS is effective and an OBD data set when the GPS is invalid;
converting the original regression samples into classification samples through an incremental regression framework;
selecting support vector classification as a basic classifier to carry out repeated iterative training on the classified samples, and screening samples harmful to the current scene by a transfer learning method in the iterative training process to reduce the weight of the samples harmful to the current scene;
according to a weight voting mechanism, classifying and integrating support vectors generated by iterative training, and obtaining a regression model based on integrated support vector regression through a common point of support vector classification and support vector regression;
and constructing a nonlinear track interpolation model based on the general kernel function and according to the regression model to obtain vehicle track interpolation data.
2. The method of claim 1, wherein obtaining raw regression samples carrying a plurality of pieces of vehicle trajectory data comprises:
acquiring original GPS data and OBD data of a vehicle, and fusing the original GPS data and the OBD data of the vehicle to obtain a GPS and OBD fused data set when a GPS is effective and an OBD data set when the GPS is ineffective;
and connecting the GPS and OBD fusion data set when the GPS is effective and the OBD data set when the GPS is invalid and carrying out blocking processing to obtain an original regression sample carrying multiple sections of vehicle track data.
3. The method of claim 1, wherein converting the raw regression samples to classification samples through an incremental regression framework comprises:
identifying response variables in the raw regression samples;
moving the response variable up and down by a preset amount
And constructing a classification sample.
4. The method of claim 1, wherein after converting the raw regression samples to classified samples by an incremental regression framework, further comprising:
respectively constructing classification type loss functions on two data sets carried by the original regression sample to obtain two classification type loss functions;
the selecting the support vector classification as a basic classifier to perform multiple iterative training on the classification sample comprises:
and obtaining a series of support vector classification classifiers by iteration according to the two classification type loss functions, wherein the weight of each support vector classification classifier is correspondingly changed along with the increase of the iteration times.
5. The method of claim 1, wherein samples harmful to the current scene are filtered through a transfer learning method in the iterative training process, and reducing the weight of the samples harmful to the current scene comprises:
acquiring a preset sample correlation factor beta, wherein the preset sample correlation factor beta is defined on a GPS and OBD fusion data set when the GPS is effective through a transfer learning method, and the preset sample correlation factor beta represents the relationship strength between a certain sample and a target sample;
and adjusting the weight vector of each support vector classification in the iterative training process through the preset sample correlation factor beta.
6. The method of claim 1, wherein the integrating the support vector classifications generated by iterative training according to a weight voting mechanism, and obtaining a regression model based on integrated support vector regression through a common point of support vector classification and support vector regression comprises:
according to a weight voting mechanism, carrying out classified integration on the support vectors generated by iterative training to obtain a weighted integration model which is most suitable for a current sample;
and solving the corresponding equation of the weighted integration model to obtain a linear regression model based on an incremental regression and transfer learning method.
7. The method of claim 1, wherein constructing a nonlinear trajectory interpolation model based on the generic kernel function and according to the regression model to obtain vehicle trajectory interpolation data comprises:
acquiring a preset general kernel function;
carrying out support vector classification training on the general kernel function to obtain a decision function;
obtaining an integrated classification plane according to the regression model and the decision function;
converting the ensemble classification plane to a regression classification plane;
solving the regression classification plane, and constructing a nonlinear track interpolation model;
optimizing parameters in the nonlinear track interpolation model through a Bayesian optimization algorithm;
and obtaining vehicle track interpolation data according to the nonlinear track interpolation model after parameter optimization.
8. A vehicle trajectory data acquisition apparatus, characterized by comprising:
the system comprises a data set acquisition module, a data processing module and a data processing module, wherein the data set acquisition module is used for acquiring an original regression sample carrying a plurality of sections of vehicle track data, and the original regression sample carries a GPS and OBD fusion data set when a GPS is effective and an OBD data set when the GPS is invalid;
a conversion module for converting the original regression sample into a classification sample through an incremental regression frame;
the iterative training module is used for selecting the support vector classification as a basic classifier to carry out multiple iterative training on the classified samples, screening the samples harmful to the current scene through a transfer learning method in the iterative training process, and reducing the weight of the samples harmful to the current scene;
the regression model building module is used for classifying and integrating the support vectors generated by iterative training according to a weight voting mechanism and obtaining a regression model based on integrated support vector regression through a common point of support vector classification and support vector regression;
and the track data acquisition module is used for constructing a nonlinear track interpolation model based on the general kernel function and according to the regression model to obtain vehicle track interpolation data.
9. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method of any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910953306.1A CN110866538B (en) | 2019-10-09 | 2019-10-09 | Vehicle trajectory data acquisition method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910953306.1A CN110866538B (en) | 2019-10-09 | 2019-10-09 | Vehicle trajectory data acquisition method and device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110866538A CN110866538A (en) | 2020-03-06 |
| CN110866538B true CN110866538B (en) | 2022-03-11 |
Family
ID=69652778
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910953306.1A Active CN110866538B (en) | 2019-10-09 | 2019-10-09 | Vehicle trajectory data acquisition method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110866538B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112749801A (en) * | 2021-01-22 | 2021-05-04 | 上海商汤智能科技有限公司 | Neural network training and image processing method and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9494694B1 (en) * | 2015-12-09 | 2016-11-15 | International Business Machines Corporation | Method and apparatus of road location inference for moving object |
| CN106314438A (en) * | 2016-08-15 | 2017-01-11 | 西北工业大学 | Method and system for detecting abnormal track in driver driving track |
| CN106370190A (en) * | 2015-07-20 | 2017-02-01 | 腾讯科技（深圳）有限公司 | Vehicle navigation method, position marking method, apparatus, and system |
-
2019
- 2019-10-09 CN CN201910953306.1A patent/CN110866538B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106370190A (en) * | 2015-07-20 | 2017-02-01 | 腾讯科技（深圳）有限公司 | Vehicle navigation method, position marking method, apparatus, and system |
| US9494694B1 (en) * | 2015-12-09 | 2016-11-15 | International Business Machines Corporation | Method and apparatus of road location inference for moving object |
| CN106314438A (en) * | 2016-08-15 | 2017-01-11 | 西北工业大学 | Method and system for detecting abnormal track in driver driving track |
Non-Patent Citations (5)
| Title |
|---|
| AdaBoost regression algorithm based on classification-type loss;Lin Gao 等;《2010 8th World Congress on Intelligent Control and Automation》;20100823;全文 * |
| Bayesian optimization of support vector machine for regression prediction of short-term traffic flow;Dong Wang 等;《Intelligent Data Analysis》;20190430;全文 * |
| Boosting for Transfer Learning;Wenyuan Dai 等;《Proceedings of the 24 th International Conference on Machine Learning》;20070124;全文 * |
| Integrating On-board Diagnostics Speed Data with Sparse GPS Measurements for Vehicle Trajectory Estimation;Sumeet Kumar 等;《The SICE Annual Conference 2013》;20130917;全文 * |
| Short-term traffic volume prediction by ensemble learning in concept drifting environments;Jianhua Xiao 等;《Knowledge-Based Systems》;20190115;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110866538A (en) | 2020-03-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Bai et al. | Stg2seq: Spatial-temporal graph to sequence model for multi-step passenger demand forecasting | |
| Guo et al. | Road damage detection algorithm for improved YOLOv5 | |
| US20210295162A1 (en) | Neural network model training method and apparatus, computer device, and storage medium | |
| CN113221905B (en) | Semantic segmentation unsupervised domain adaptation method, device and system based on uniform clustering and storage medium | |
| US20200094824A1 (en) | Learning to simulate | |
| CN114330868A (en) | Passenger flow prediction method based on self-attention personalized enhanced graph convolution network | |
| CN111414874B (en) | Driving risk prediction method, device and equipment based on human face and storage medium | |
| CN112699941B (en) | Plant disease severity image classification method, device, equipment and storage medium | |
| Rezapour et al. | Application of deep learning techniques in predicting motorcycle crash severity | |
| US10732694B2 (en) | Power state control of a mobile device | |
| CN112270355A (en) | Active safety prediction method based on big data technology and SAE-GRU | |
| KR20220144356A (en) | Systems and Methods for Automatic Model Generation | |
| US20230140289A1 (en) | Traffic accident prediction systems and methods | |
| CN110866538B (en) | Vehicle trajectory data acquisition method and device, computer equipment and storage medium | |
| Chelbi et al. | Worst-case scenarios identification approach for the evaluation of advanced driver assistance systems in intelligent/autonomous vehicles under multiple conditions | |
| Hsueh et al. | A short-term traffic speed prediction model based on LSTM networks | |
| Saisree et al. | Pothole detection using deep learning classification method | |
| US20210333764A1 (en) | Electronic control device and neural network update system | |
| CN112699858B (en) | Unmanned platform smoke fog sensing method and system, computer equipment and storage medium | |
| Li et al. | Short-term traffic prediction with deep neural networks and adaptive transfer learning | |
| CN112288279A (en) | Business risk assessment method and device based on natural language processing and linear regression | |
| Vadhwani et al. | Predictive analysis of injury severity of person across angle crashes using machine learning models | |
| US11562184B2 (en) | Image-based vehicle classification | |
| CN114548463A (en) | Line information prediction method, line information prediction device, computer equipment and storage medium | |
| CN110472656B (en) | Vehicle image classification method, device, computer equipment and storage medium |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |