CN110084137A - Data processing method, device and computer equipment based on Driving Scene - Google Patents
Data processing method, device and computer equipment based on Driving Scene Download PDFInfo
- Publication number
- CN110084137A CN110084137A CN201910269776.6A CN201910269776A CN110084137A CN 110084137 A CN110084137 A CN 110084137A CN 201910269776 A CN201910269776 A CN 201910269776A CN 110084137 A CN110084137 A CN 110084137A
- Authority
- CN
- China
- Prior art keywords
- information
- detection
- detection information
- characterization
- driving scene
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The present invention proposes a kind of data processing method based on Driving Scene, device and computer equipment, wherein, method includes: the first detection information for obtaining and driving, the semantic model obtained according to training, first detection information is calculated, obtain the first characterization information of Driving Scene, wherein, semantic model be used to indicate each Driving Scene characterization information and corresponding Driving Scene under corresponding relationship between the detection information that detects, from each second characterization information of storage, determine object representation information similar with the first characterization information, using corresponding second detection information of object representation information as object detection information, realize the detection information for quickly finding from database and belonging to similar Driving Scene with target Driving Scene, solves the technical problem that artificial enquiry efficiency is lower in the prior art.
Description
Technical field
The present invention relates to automatic Pilot technical field more particularly to a kind of data processing methods based on Driving Scene, dress
It sets and computer equipment.
Background technique
In unmanned technology, possess a large amount of parameter, and parameter and Driving Scene have corresponding relationship, need to pass through
Method based on data searches number corresponding with given scenario from the database for being stored with a large amount of unmanned drive test datas
According to the test of function being carried out, with the drive parameter under the corresponding scene of determination.
It in the prior art, is by manually carrying out manual queries, search efficiency is very low.
Summary of the invention
The present invention is directed to solve at least some of the technical problems in related technologies.
For this purpose, the first purpose of this invention is to propose a kind of data processing method based on Driving Scene, pass through instruction
The semantic model got is calculated the first characterization information and the second characterization information of Driving Scene, and believes from the second characterization
Object representation information similar with the first characterization information is determined in breath, using the corresponding detection information of object representation information as mesh
Detection information is marked, quickly finds the inspection for belonging to similar Driving Scene with target Driving Scene from the database of storage to realize
Measurement information improves search efficiency, solves and needs manually to search from a large amount of driving data in the prior art, efficiency
Lower technical problem.
Second object of the present invention is to propose a kind of data processing equipment based on Driving Scene.
Third object of the present invention is to propose a kind of computer equipment.
Fourth object of the present invention is to propose a kind of non-transitorycomputer readable storage medium.
In order to achieve the above object, first aspect present invention embodiment proposes a kind of data processing side based on Driving Scene
Method, comprising:
Obtain the first detection information driven;
According to the semantic model that training obtains, first detection information is calculated, the first of Driving Scene is obtained
Characterization information;Wherein, the semantic model be used to indicate each Driving Scene characterization information and corresponding Driving Scene under detect
Detection information between corresponding relationship;
From each second characterization information of storage, object representation information similar with first characterization information is determined;
Using corresponding second detection information of the object representation information as object detection information.
In order to achieve the above object, second aspect of the present invention embodiment proposes a kind of data processing dress based on Driving Scene
It sets, described device includes:
First obtains module, for obtaining the first detection information driven;
First determining module, the semantic model for being obtained according to training, calculates first detection information, obtains
To the first characterization information of Driving Scene;Wherein, the semantic model is used to indicate the characterization information of each Driving Scene and corresponding
The corresponding relationship between detection information detected under Driving Scene;
Second determining module, for from each second characterization information of storage, determination to be similar to first characterization information
Object representation information;
Processing module, for using corresponding second detection information of the object representation information as object detection information.
In order to achieve the above object, third aspect present invention embodiment proposes a kind of computer equipment, including memory, processing
Device and storage on a memory and the computer program that can run on a processor, when the processor executes described program, reality
The now data processing method based on Driving Scene as described in aforementioned first aspect.
In order to achieve the above object, fourth aspect present invention embodiment proposes a kind of non-transitory computer-readable storage medium
Matter is stored thereon with computer program, when which is executed by processor realize as described in aforementioned first aspect based on driving
The data processing method of scene.
The beneficial effect of the embodiment of the present invention may include it is following the utility model has the advantages that
The first detection information driven is obtained, according to the semantic model that training obtains, the first detection information is calculated,
Obtain the first characterization information of Driving Scene, wherein semantic model is used to indicate the characterization information of each Driving Scene and accordingly drives
The corresponding relationship between the detection information detected under scene is sailed, from each second characterization information of storage, determining and the first table
Object representation information as reference manner of breathing, it is real using corresponding second detection information of object representation information as object detection information
The detection information for quickly finding from the database of storage and belonging to similar Driving Scene with target Driving Scene is showed, has improved
Search efficiency.
The additional aspect of the present invention and advantage will be set forth in part in the description, and will partially become from the following description
Obviously, or practice through the invention is recognized.
Detailed description of the invention
Above-mentioned and/or additional aspect and advantage of the invention will become from the following description of the accompanying drawings of embodiments
Obviously and it is readily appreciated that, in which:
Fig. 1 is a kind of flow diagram of the data processing method based on Driving Scene provided by the embodiment of the present invention;
Fig. 2 is the process signal of data processing method of the another kind based on Driving Scene provided by the embodiment of the present invention
Figure;
Fig. 3 is a kind of flow diagram of the training method of semantic model provided by the embodiment of the present invention;
Fig. 4 is a kind of structural schematic diagram of semantic model provided by the embodiment of the present invention;
Fig. 5 is a kind of structural schematic diagram of the data processing equipment based on Driving Scene provided in an embodiment of the present invention;With
And
Fig. 6 shows the block diagram for being suitable for the exemplary computer device for being used to realize the application embodiment.
Specific embodiment
The embodiment of the present invention is described below in detail, examples of the embodiments are shown in the accompanying drawings, wherein from beginning to end
Same or similar label indicates same or similar element or element with the same or similar functions.Below with reference to attached
The embodiment of figure description is exemplary, it is intended to is used to explain the present invention, and is not considered as limiting the invention.
Below with reference to the accompanying drawings the data processing method based on Driving Scene, device and the computer of the embodiment of the present invention are described
Equipment.
Fig. 1 is a kind of flow diagram of the data processing method based on Driving Scene provided by the embodiment of the present invention.
As shown in Figure 1, method includes the following steps:
Step 101, the first detection information of driving is obtained.
Wherein, the first detection information can be what the detection under known Driving Scene obtained.
Specifically, under known Driving Scene, the environmental data obtained to environment measuring is obtained and to travel condition of vehicle
The vehicle data detected, wherein to the environmental data that environment measuring obtains, adopted by the acquisition device in environment
What collection obtained, for example, passing through the collected data of devices such as camera, radar and global position system GPS.Vehicle data is
What the operation data of acquisition vehicle itself obtained, for example, wheel speed meter and the collected vehicle operation data of inertial sensor IMU.
In turn, it by the environmental data and vehicle data under collected known scene, is aligned according to detection time, obtains known drive
The first detection data under scene is sailed, and will test data progress vectorization expression and pass through as a kind of possible implementation
One-hot coding, is expressed as vector for the first detection data.
It should be noted that the vehicle data detected under Driving Scene is detected according to preset detection duration
, a data slot of the first detection information a length of detection duration when being.
Step 102, the semantic model obtained according to training, calculates the first detection information, obtains Driving Scene
First characterization information.
Wherein, semantic model be used to indicate each Driving Scene characterization information and corresponding Driving Scene under the detection that detects
Corresponding relationship between information.
Wherein, the characterization information of Driving Scene refers to the information that can indicate the feature of Driving Scene, is used to indicate the driving
The classification of scene, for example, being the Driving Scene of the Driving Scene or expressway under overpass or the driver training ground of fork in the road
Scape etc. refers to energy.
In the embodiment of the present invention, semantic model includes hidden layer and output layer, hidden layer, for true according to the detection information of input
The characterization information of fixed corresponding Driving Scene, output layer, the characterization information for being exported according to hidden layer predict the detection inputted
The context of information, and context is detection letter detected within the proximity detection period for the detection information for detecting input
Breath, that is to say, that when the detection period of the context of the detection information of the input predicted is the detection with the detection information of input
Duan Xianglin's, the parameter of hidden layer, i.e. the first weight matrix can be determined by being trained to semantic model.Wherein, semantic mould
The training method of type will be described in detail in following embodiments.
Specifically, the first weight matrix of hidden layer being trained to semantic model is obtained, by the of vector form
One detection information is multiplied with the first weight matrix of hidden layer, obtains the first characterization information, is by the first detection information specifically
It is multiplied respectively with multiple column vectors of the first weight matrix, obtains the first characterization letter of corresponding first detection information of each column vector
Breath.
Step 103, from each second characterization information of storage, object table reference similar with the first characterization information is determined
Breath.
For the generating mode of each second characterization information of storage, as a kind of possible implementation, the second characterization letter
Breath is when semantic model is trained using the second detection information as training sample, when training is completed according to training sample
It is calculated, and is stored.As alternatively possible implementation, the second characterization information is obtained according to training
First weight matrix of semantic model calculates determination to the second detection information of acquisition, and is stored.Wherein, above-mentioned
Second detection information refers to that under unknown Driving Scene, detection obtains.
Specifically, after acquiring the second characterization information, the similarity of the second characterization information and the first characterization information is calculated,
As a kind of possible implementation, either pressed from both sides by calculating the distance between the second characterization information and the first characterization information
Angle, to determine the similarity between the second characterization information and the first characterization information, for example, determining different tables by Euclidean distance
Similarity between reference breath, similarity is compared with preset threshold similarity, and determination is similar with the first characterization information
Object representation information from a large amount of history driving information, is efficiently quickly found and known driver training ground so as to realize
Object representation information in the similar unknown Driving Scene of the characterization information of scape, improves the efficiency of lookup.
Step 104, using corresponding second detection information of object representation information as object detection information.
Wherein, object detection information be to the first detection information detected in similar Driving Scene, meanwhile, target inspection
Measurement information and the first detection information can also be under identical Driving Scene that detection obtains.
Specifically, using corresponding second detection information of object representation information as object detection information, target detection letter
Breath can be determined as to the first detection information being to detect to obtain under similar Driving Scene, realize from the second detection information
Finding out with the first detection information is the object detection information detected in similar or identical Driving Scene, that is,
Saying realizes automatically from the detection data for finding in historical data with detecting under known Driving Scene, belongs to similar or phase
The detection data of same Driving Scene, while search efficiency is also higher.
In the data processing method based on Driving Scene of the present embodiment, the first detection information of driving is obtained, according to instruction
The semantic model got calculates the first detection information, obtains the first characterization information of Driving Scene, wherein semantic
Model be used to indicate each Driving Scene characterization information and corresponding Driving Scene under corresponding pass between the detection information that detects
System determines object representation information similar with the first characterization information, by object table reference from each second characterization information of storage
Corresponding second detection information is ceased as object detection information, is realized and is quickly found from database and target Driving Scene
Belong to the detection information of similar Driving Scene, search efficiency is higher.
Based on a upper embodiment, the embodiment of the invention provides a kind of possibility of data processing method based on Driving Scene
Implementation, Fig. 2 is that the process of data processing method of the another kind based on Driving Scene provided by the embodiment of the present invention is shown
It is intended to.
As shown in Fig. 2, returning method comprises the steps of: before step 103
Step 201, the second detection information is obtained.
Wherein, the second detection information refers to that under unknown Driving Scene, detection obtains.
Specifically, it obtains the environmental data obtained in history driving procedure to environment measuring and travel condition of vehicle is carried out
Detect obtained vehicle data, wherein to the environmental data that environment measuring obtains, acquired by the acquisition device in environment
It arrives, for example, passing through the collected data of devices such as camera, radar and global position system GPS.Vehicle data is acquisition
What the operation data of vehicle itself obtained, for example, wheel speed meter and the collected vehicle operation data of inertial sensor IMU.In turn,
It is aligned according to detection time, obtains history driving information, history driving information is counted according to corresponding detection duration
According to fragment, each second detection information by the arrangement of detection period is obtained, for example, it is 20 seconds a length of when detection, by history driving information
It is divided according to 20 seconds durations, obtains each second detection information according to long array when detection, that is to say, that each second detection
The data slot of a length of detection duration when information is, wherein the detection duration of the second detection information and the first detection information is
It is identical, in order to the comparison between corresponding second characterization information and the first characterization information.In turn, the second detection information is led to
It crosses one-hot coding mode to be encoded, the second detection information of obtained vector form.
Step 202, according to the second detection information, corresponding second characterization information is calculated.
For the generating mode of the second characterization information, as a kind of possible implementation, the vector form acquired
The second detection information semantic model is trained for generating training sample pair so that training complete semantic model
Acquire corresponding second characterization information of the second detection information, wherein using the second detection information as training sample pair, utilize
The method that training sample is trained semantic model will be described in detail in next embodiment.
As alternatively possible implementation, the second characterization information is the first power of the semantic model obtained according to training
Weight matrix carries out the second detection information of acquisition to calculate determination.
In the data processing method based on Driving Scene of the embodiment of the present invention, examined according to the environment under the unknown scene of history
Measured data and vehicle operation data, are aligned according to detection time, and according to detection duration carry out data fragmentation, obtain according to
Each second detection information for detecting period arrangement, and determines the second characterization information according to the second detection information, in turn, can by pair
Similarity identification between first characterization information and the second characterization information, determining and the first characterization information from the second characterization information
The higher object representation information of similarity, it is real using corresponding second detection data of object representation information as target detection data
Showed the detection information for quickly finding from database and belonging to similar Driving Scene with target Driving Scene, search efficiency compared with
It is high.
Analysis through the foregoing embodiment is it is found that the semantic model that training obtains is used to indicate the characterization letter of each Driving Scene
The corresponding relationship between detection information detected under breath and corresponding Driving Scene, for this purpose, present embodiments providing a kind of semanteme
The possible implementation of the training method of model, Fig. 3 are a kind of training side of semantic model provided by the embodiment of the present invention
The flow diagram of method.
As shown in figure 3, this method may comprise steps of:
Step 301, training sample pair is determined from the second detection information.
The training effectiveness that semantic model is improved in the embodiment of the present invention, the second detection information that will acquire is as training
Sample, specifically from the second detection information obtain training sample pair, training sample to comprising effective sample to it is invalid
Sample pair, effective sample to and invalid sample to being obtained from the second detection information, wherein effective sample is to including inspection
Survey period adjacent two group of second detection information, invalid sample pair, including detection period non-conterminous two group of second detection information.
Specifically, the history driving information that can be will acquire carries out data fragmentation according to corresponding detection duration, obtains according to detection
When long array each second detection information, for example, detection when it is 20 seconds a length of, that is to say, that when each second detection information is 20 seconds
Long detection information.In turn, each second detection information can use one-hot coding mode, by each detection information be converted into
The detection information of amount form.
For example, each second detection information are as follows: w0, w1..., wt, then effective sample is to can for example be expressed as (wt,ct),
That is effective sample is to being to detect period adjacent two groups of detection information ctAnd wt, non-conterminous two group second of the period of others detection
Detection information is known as invalid sample pair, can be expressed as
Step 302, it is trained using training sample to semantic model, to adjust the first weight matrix value, and
The second weight matrix value of output layer is adjusted, so that the context of output layer output minimizes the error.
Wherein, semantic model, such as can be the negative sampling model of Skip-Gram, semantic model includes hidden layer and output
Layer, as shown in Figure 4, wherein the first weight matrix is the parameters weighting matrix of hidden layer in semantic model, and the second weight matrix is language
The parameters weighting matrix of output layer in adopted model.Hidden layer is used for the second detection information according to input, determines corresponding Driving Scene
The second characterization information, output layer is used for the second characterization information of Driving Scene exported according to hidden layer, exports the second of input
The context for the second detection information that the probability distribution of the corresponding context of detection information, i.e. prediction are inputted, wherein up and down
Text is detected detection information within the proximity detection period for the second detection information for detecting input.
It should be noted that the output of the semantic model in Fig. 4 is only to illustrate, do not limit the invention, exports
The second detection information context probability distribution, the window namely defeated of output can be adjusted according to the demand of practical application
Out in result the probability distribution of the context of the second detection information quantity.
In the embodiment of the present invention, through the above steps obtained in effective sample to and invalid sample pair, to semantic model
It is trained, to adjust the value of the first weight matrix and the second weight matrix, so that the context error of output layer output is most
Small, so that the training of semantic model is completed, the first weight matrix of hidden layer is determined, that is to say, that the purpose master of semantic model training
If determining the first weight matrix of hidden layer.
Specifically, a stochastic variable Y can be introduced, for characterizing whether one group of detection information is effective sample pair, with
The value of machine variable Y is 0 or 1, and whether one group of detection information is that the probability of effective sample pair is expressed as: and P (Y | wt,ct)。
When being trained to semantic model, sample centering effective sample is to quantity and invalid sample to the configuration side of quantity
Formula can flexibly be set according to the scale of sample set, for each effective sample pair, if the number of invalid sample pair is
K, then the sample for K+1 is trained structure to semantic model with sample, the loss of the semantic model constructed
Function can indicate are as follows:When hands-on, for
Each effective sample pair, for example, 5 invalid samples pair can be configured, with sample to the composition of sample for 6 to semanteme
Model is trained, and the iteration speed of model parameter can be improved, and improves the efficiency of model training.
As a kind of possible implementation, it is iterated solution using gradient descent method, it is continuous to adjust the first weight
Matrix value, i.e. the parameters weighting matrix of hidden layer, and the second weight matrix value of continuous adjustment, i.e. the parameter power of output layer
Weight matrix, until the context of output layer output minimizes the error, i.e. semantic model training is completed, the first weight matrix of hidden layer
The second characterization information of corresponding Driving Scene can be determined according to the second detection information of input, that is to say, that semantic model study
The corresponding relationship between the detection information for characterizing and accordingly detecting under Driving Scene of each Driving Scene is obtained.
It should be noted that the first weight matrix multiple column vectors that include determined after the completion of training, each arrange to
Corresponding second characterization information of one the second detection information of amount instruction, that is to say, that each column vector of the first weight matrix point
It Wei not corresponding second characterization information of each second detection information.In turn, the first weight of the hidden layer for the semantic model that training is completed
Matrix, after calculating the first detection information under known scene, i.e., by the first weight of the first detection information and hidden layer
The first characterization information is calculated in matrix column multiplication of vectors, that is, realizes and find in the second unknown characterization information and the
The similar object representation information of one characterization information, it can realize the data search of automation, and efficiency is higher.
In the training method of the semantic model of the embodiment of the present invention, using the history driving information obtained in advance, instruction is generated
Practice sample pair, training sample centering include effective sample to and invalid sample pair, based on effective sample to and invalid sample construct
Loss function, by gradient decline alternative manner, with the parameters weighting matrix of the hidden layer of semantic model and the ginseng of output layer
Number weight matrix is adjusted, and in the context error minimum of output layer output, semantic model then complete by training, and training is completed
Semantic model study to the characterization of each Driving Scene and corresponding Driving Scene under it is corresponding between the detection information that detects
Relationship can quickly find the detection under Driving Scene similar with known Driving Scene by the semantic model that training obtains
Data improve the efficiency of lookup.
In order to more clearly illustrate the data processing method based on Driving Scene of above-described embodiment, under overpass
For automatic Pilot scene, it is further illustrated:
The algorithm of automatic Pilot shows poor under overpass, needs to obtain the inspection of the Driving Scene under a large amount of overpass
Measurement information improves the accuracy of automatic Pilot under overpass by the adjustment of parameter for being tested.Specifically, according to
History detection data carries out the division of alignment of data and data slot, obtains the detect under unknown Driving Scene second detection
Data determine training sample pair from the second detection data, are trained using training sample to semantic model that training is completed
Afterwards, the first weight matrix of semantic model determines, obtained the first weight matrix of training includes multiple column vectors, each arrange to
Amount one the second characterization information of instruction, that is to say, that each column vector of the first weight matrix is respectively each second detection information
Corresponding second characterization information.In turn, according to the first detection data of automatic Pilot under the overpass of acquisition, by the first testing number
According to the first weight matrix of the semantic model completed using training carry out that the first characterization information is calculated.By the first characterization information
With the second characterization information by similarity algorithm, determine in the second characterization information under unknown Driving Scene and under overpass from
The similarity degree of first characterization information of dynamic Driving Scene is higher than the object representation information of preset threshold, and then by object table reference
Cease corresponding second detection information as the detection information under the Driving Scene under overpass, thus, realize automation from
In the data of a large amount of history Driving Scene, quickly finds and detect testing number under the automatic Pilot scene under overpass
According to similar detection data, wherein search the scene of determining detection data and the automatic Pilot of overpass from historic scenery
Scene can be identical or similar scene, so that the detection data under similar or same scene is found out, into
The test of row scene can be used for improving the accuracy of the automatic Pilot of the scene, look into automatically to not only realize
It looks for, search speed is fast, and it is high-efficient, the accuracy of automatic Pilot under the scene can also be improved.
In order to realize above-described embodiment, the present invention also proposes a kind of data processing equipment based on Driving Scene.
Fig. 5 is a kind of structural schematic diagram of the data processing equipment based on Driving Scene provided in an embodiment of the present invention.
As shown in figure 5, the device includes: the first acquisition module 51, the first determining module 52, the second determining module 53 and place
Manage module 54.
First obtains module 51, for for obtaining the first detection information driven.
First determining module 52 carries out first detection information for the semantic model for being obtained according to training
It calculates, obtains the first characterization information of Driving Scene;Wherein, the semantic model is used to indicate the characterization information of each Driving Scene
With the corresponding relationship between the detection information that is detected under corresponding Driving Scene.
Second determining module 53, for from each second characterization information of storage, determining and the first characterization information phase
As object representation information.
Processing module 54, for using corresponding second detection information of the object representation information as object detection information.
Further, in a kind of possible implementation of the embodiment of the present invention, described device further include:
Second obtains module, for obtaining the environmental data obtained in history driving procedure to environment measuring and transporting to vehicle
The vehicle data that row state is detected;It is aligned according to detection time, obtains history driving information, by the history
Driving information carries out data fragmentation according to corresponding detection duration, obtains each second detection information by the arrangement of detection period, root
According to second detection information, corresponding second characterization information is calculated.
As a kind of possible implementation, described above first obtains module 51, is specifically used for:
The data that environment and travel condition of vehicle will be detected under the known Driving Scene, when according to detection
Between be aligned, obtain first detection information.
As a kind of possible implementation, above-mentioned first determining module 52, comprising:
Acquiring unit, for obtaining the first weight matrix of the hidden layer being trained to the semantic model;It is described
Semantic model includes the hidden layer and output layer;The hidden layer determines corresponding Driving Scene for the detection information according to input
Characterization information;Output layer, the characterization information for being exported according to the hidden layer, prediction obtain the detection information of the input
Context, the context are detection letters detected within the proximity detection period for detecting the detection information of the input
Breath;
Computing unit is obtained for will be multiplied with the first detection information with first weight matrix described in vector form
To first characterization information.
As a kind of possible implementation, the semantic model is using training sample to being trained, to adjust the
One weight matrix value, and the second weight matrix value of the adjustment output layer, so that the context of output layer output misses
Difference minimizes;Wherein, the training sample to include effective sample to and invalid sample pair;The effective sample pair, including inspection
Survey period adjacent two group of second detection information;The invalid sample pair, including non-conterminous two group of second detection of detection period
Information.
As a kind of possible implementation, first weight matrix, including multiple column vectors, each column vector refer to
Show second characterization information.
It should be noted that the aforementioned data for being also applied for the embodiment to the explanation of data processing method embodiment
Processing unit, principle is identical, and details are not described herein again.
In the data processing equipment based on Driving Scene of the present embodiment, according to the environment measuring number under the unknown scene of history
According to and vehicle operation data, be aligned according to detection time, and according to detection duration carry out data fragmentation, obtain according to detection
Each second detection information of period arrangement, the second detection information can be generated training sample pair, be trained to semantic model, into
Row calculates the first detection information of acquisition and the second detection information using the semantic model that training obtains, is corresponded to
The first characterization information and the second characterization information, by between the first characterization information and the second characterization information similarity know
Not, the determining and higher object representation information of the first characterization information similarity from the second characterization information, by object representation information
Corresponding second detection data is realized and is quickly found from database and target Driving Scene category as target detection data
In the detection information of similar Driving Scene, search efficiency is higher.
In order to realize above-described embodiment, the embodiment of the present invention also proposed a kind of computer equipment, including memory, processing
Device and storage on a memory and the computer program that can run on a processor, when the processor executes described program, reality
The now data processing method based on Driving Scene as described in preceding method embodiment.
Fig. 6 shows the block diagram for being suitable for the exemplary computer device for being used to realize the application embodiment.What Fig. 5 was shown
Computer equipment 12 is only an example, should not function to the embodiment of the present application and use scope bring any restrictions.
As shown in fig. 6, computer equipment 12 is showed in the form of universal computing device.The component of computer equipment 12 can be with
Including but not limited to: one or more processor or processing unit 16, system storage 28 connect different system components
The bus 18 of (including system storage 28 and processing unit 16).
Bus 18 indicates one of a few class bus structures or a variety of, including memory bus or Memory Controller,
Peripheral bus, graphics acceleration port, processor or the local bus using any bus structures in a variety of bus structures.It lifts
For example, these architectures include but is not limited to industry standard architecture (Industry Standard
Architecture;Hereinafter referred to as: ISA) bus, microchannel architecture (Micro Channel Architecture;Below
Referred to as: MAC) bus, enhanced isa bus, Video Electronics Standards Association (Video Electronics Standards
Association;Hereinafter referred to as: VESA) local bus and peripheral component interconnection (Peripheral Component
Interconnection;Hereinafter referred to as: PCI) bus.
Computer equipment 12 typically comprises a variety of computer system readable media.These media can be it is any can be by
The usable medium that computer equipment 12 accesses, including volatile and non-volatile media, moveable and immovable medium.
Memory 28 may include the computer system readable media of form of volatile memory, such as random access memory
Device (Random Access Memory;Hereinafter referred to as: RAM) 30 and/or cache memory 32.Computer equipment 12 can be with
It further comprise other removable/nonremovable, volatile/non-volatile computer system storage mediums.Only as an example,
Storage system 34 can be used for reading and writing immovable, non-volatile magnetic media, and (Fig. 6 do not show, commonly referred to as " hard drive
Device ").Although being not shown in Fig. 6, the disk for reading and writing to removable non-volatile magnetic disk (such as " floppy disk ") can be provided and driven
Dynamic device, and to removable anonvolatile optical disk (such as: compact disc read-only memory (Compact Disc Read Only
Memory;Hereinafter referred to as: CD-ROM), digital multi CD-ROM (Digital Video Disc Read Only
Memory;Hereinafter referred to as: DVD-ROM) or other optical mediums) read-write CD drive.In these cases, each driving
Device can be connected by one or more data media interfaces with bus 18.Memory 28 may include that at least one program produces
Product, the program product have one group of (for example, at least one) program module, and it is each that these program modules are configured to perform the application
The function of embodiment.
Program/utility 40 with one group of (at least one) program module 42 can store in such as memory 28
In, such program module 42 include but is not limited to operating system, one or more application program, other program modules and
It may include the realization of network environment in program data, each of these examples or certain combination.Program module 42 is usual
Execute the function and/or method in embodiments described herein.
Computer equipment 12 can also be with one or more external equipments 14 (such as keyboard, sensing equipment, display 24
Deng) communication, can also be enabled a user to one or more equipment interact with the computer equipment 12 communicate, and/or with make
The computer equipment 12 any equipment (such as network interface card, the modulatedemodulate that can be communicated with one or more of the other calculating equipment
Adjust device etc.) communication.This communication can be carried out by input/output (I/O) interface 22.Also, computer equipment 12 may be used also
To pass through network adapter 20 and one or more network (such as local area network (Local Area Network;Hereinafter referred to as:
LAN), wide area network (Wide Area Network;Hereinafter referred to as: WAN) and/or public network, for example, internet) communication.Such as figure
Shown, network adapter 20 is communicated by bus 18 with other modules of computer equipment 12.It should be understood that although not showing in figure
Out, other hardware and/or software module can be used in conjunction with computer equipment 12, including but not limited to: microcode, device drives
Device, redundant processing unit, external disk drive array, RAID system, tape drive and data backup storage system etc..
Processing unit 16 by the program that is stored in system storage 28 of operation, thereby executing various function application and
Data processing, such as realize the method referred in previous embodiment.
In order to realize above-described embodiment, the embodiment of the present invention also proposed a kind of non-transitory computer-readable storage medium
Matter is stored thereon with computer program, realized when which is executed by processor as described in preceding method embodiment based on driving
Sail the data processing method of scene.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or spy described in conjunction with this embodiment or example
Point is included at least one embodiment or example of the invention.In the present specification, schematic expression of the above terms are not
It must be directed to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be in office
It can be combined in any suitable manner in one or more embodiment or examples.In addition, without conflicting with each other, the skill of this field
Art personnel can tie the feature of different embodiments or examples described in this specification and different embodiments or examples
It closes and combines.
In addition, term " first ", " second " are used for descriptive purposes only and cannot be understood as indicating or suggesting relative importance
Or implicitly indicate the quantity of indicated technical characteristic.Define " first " as a result, the feature of " second " can be expressed or
Implicitly include at least one this feature.In the description of the present invention, the meaning of " plurality " is at least two, such as two, three
It is a etc., unless otherwise specifically defined.
Any process described otherwise above or method description are construed as in flow chart or herein, and expression includes
It is one or more for realizing custom logic function or process the step of executable instruction code module, segment or portion
Point, and the range of the preferred embodiment of the present invention includes other realization, wherein can not press shown or discussed suitable
Sequence, including according to related function by it is basic simultaneously in the way of or in the opposite order, Lai Zhihang function, this should be of the invention
Embodiment person of ordinary skill in the field understood.
Expression or logic and/or step described otherwise above herein in flow charts, for example, being considered use
In the order list for the executable instruction for realizing logic function, may be embodied in any computer-readable medium, for
Instruction execution system, device or equipment (such as computer based system, including the system of processor or other can be held from instruction
The instruction fetch of row system, device or equipment and the system executed instruction) it uses, or combine these instruction execution systems, device or set
It is standby and use.For the purpose of this specification, " computer-readable medium ", which can be, any may include, stores, communicates, propagates or pass
Defeated program is for instruction execution system, device or equipment or the dress used in conjunction with these instruction execution systems, device or equipment
It sets.The more specific example (non-exhaustive list) of computer-readable medium include the following: there is the electricity of one or more wirings
Interconnecting piece (electronic device), portable computer diskette box (magnetic device), random access memory (RAM), read-only memory
(ROM), erasable edit read-only storage (EPROM or flash memory), fiber device and portable optic disk is read-only deposits
Reservoir (CDROM).In addition, computer-readable medium can even is that the paper that can print described program on it or other are suitable
Medium, because can then be edited, be interpreted or when necessary with it for example by carrying out optical scanner to paper or other media
His suitable method is handled electronically to obtain described program, is then stored in computer storage.
It should be appreciated that each section of the invention can be realized with hardware, software, firmware or their combination.Above-mentioned
In embodiment, software that multiple steps or method can be executed in memory and by suitable instruction execution system with storage
Or firmware is realized.Such as, if realized with hardware in another embodiment, following skill well known in the art can be used
Any one of art or their combination are realized: have for data-signal is realized the logic gates of logic function from
Logic circuit is dissipated, the specific integrated circuit with suitable combinational logic gate circuit, programmable gate array (PGA), scene can compile
Journey gate array (FPGA) etc..
Those skilled in the art are understood that realize all or part of step that above-described embodiment method carries
It suddenly is that relevant hardware can be instructed to complete by program, the program can store in a kind of computer-readable storage medium
In matter, which when being executed, includes the steps that one or a combination set of embodiment of the method.
It, can also be in addition, each functional unit in each embodiment of the present invention can integrate in a processing module
It is that each unit physically exists alone, can also be integrated in two or more units in a module.Above-mentioned integrated mould
Block both can take the form of hardware realization, can also be realized in the form of software function module.The integrated module is such as
Fruit is realized and when sold or used as an independent product in the form of software function module, also can store in a computer
In read/write memory medium.
Storage medium mentioned above can be read-only memory, disk or CD etc..Although having been shown and retouching above
The embodiment of the present invention is stated, it is to be understood that above-described embodiment is exemplary, and should not be understood as to limit of the invention
System, those skilled in the art can be changed above-described embodiment, modify, replace and become within the scope of the invention
Type.
Claims (14)
1. a kind of data processing method based on Driving Scene, which is characterized in that the described method comprises the following steps:
Obtain the first detection information driven;
According to the semantic model that training obtains, first detection information is calculated, obtains the first characterization of Driving Scene
Information;Wherein, the semantic model be used to indicate each Driving Scene characterization information and corresponding Driving Scene under the inspection that detects
Corresponding relationship between measurement information;
From each second characterization information of storage, object representation information similar with first characterization information is determined;
Using corresponding second detection information of the object representation information as object detection information.
2. data processing method according to claim 1, which is characterized in that the first detection information of the acquisition, comprising:
The data that environment and travel condition of vehicle will be detected under the known Driving Scene, according to detection time into
Row alignment, obtains first detection information.
3. data processing method according to claim 1, which is characterized in that each second characterization information from storage
In, before determining object representation information similar with first characterization information, further includes:
It obtains the environmental data obtained in history driving procedure to environment measuring and travel condition of vehicle is detected
Vehicle data;
It is aligned according to detection time, obtains history driving information;
The history driving information is subjected to data fragmentation according to corresponding detection duration, obtains each the by the arrangement of detection period
Two detection informations;
According to second detection information, corresponding second characterization information is calculated.
4. data processing method according to claim 1-3, which is characterized in that the language obtained according to training
Adopted model calculates first detection information, obtains the first characterization information of Driving Scene, comprising:
Obtain the first weight matrix of the hidden layer being trained to the semantic model;The semantic model includes described hidden
Layer and output layer;The hidden layer determines the characterization information of corresponding Driving Scene for the detection information according to input;Output layer,
Characterization information for being exported according to the hidden layer, prediction obtain the context of the detection information of the input, the context
It is detected detection information within the proximity detection period for detecting the detection information of the input;
First detection information of vector form is multiplied with the first weight matrix of the hidden layer, obtains first characterization
Information.
5. data processing method according to claim 4, which is characterized in that
The semantic model is using training sample to being trained, to adjust the first weight matrix value, and described in adjustment
Second weight matrix value of output layer, so that the context of output layer output minimizes the error;Wherein, the training sample pair
Including effective sample to and invalid sample pair;The effective sample pair, including two group of second detection information that the detection period is adjacent;
The invalid sample pair, including detection period non-conterminous two group of second detection information.
6. data processing method according to claim 5, which is characterized in that
First weight matrix, including multiple column vectors, each column vector indicate second characterization information.
7. a kind of data processing equipment based on Driving Scene, which is characterized in that described device includes:
First obtains module, for obtaining the first detection information driven;
First determining module, the semantic model for being obtained according to training, calculates first detection information, is driven
Sail the first characterization information of scene;Wherein, the semantic model is used to indicate the characterization information and corresponding driving of each Driving Scene
The corresponding relationship between detection information detected under scene;
Second determining module, for determining mesh similar with first characterization information from each second characterization information of storage
Mark characterization information;
Processing module, for using corresponding second detection information of the object representation information as object detection information.
8. data processing equipment according to claim 7, which is characterized in that described first obtains module, is specifically used for: will
The data detected under the known Driving Scene to environment and travel condition of vehicle carry out pair according to detection time
Together, first detection information is obtained.
9. data processing equipment according to claim 7, which is characterized in that described device, further includes:
Second obtains module, for obtaining the environmental data obtained in history driving procedure to environment measuring and running shape to vehicle
The vehicle data that state is detected;It is aligned according to detection time, obtains history driving information;The history is driven
Information carries out data fragmentation according to corresponding detection duration, obtains each second detection information by the arrangement of detection period;According to institute
The second detection information is stated, corresponding second characterization information is calculated.
10. according to the described in any item data processing equipments of claim 7-9, which is characterized in that first determining module, packet
It includes:
Acquiring unit, for obtaining the first weight matrix of the hidden layer being trained to the semantic model;The semanteme
Model includes the hidden layer and output layer;The hidden layer determines the table of corresponding Driving Scene for the detection information according to input
Reference breath;Output layer, the characterization information for being exported according to the hidden layer, prediction obtain the upper and lower of the detection information of the input
Text, the context are detected detection informations within the proximity detection period for detecting the detection information of the input;
Computing unit obtains institute for will be multiplied with the first detection information with first weight matrix described in vector form
State the first characterization information.
11. data processing equipment according to claim 10, which is characterized in that
The semantic model is using training sample to being trained, to adjust the first weight matrix value, and described in adjustment
Second weight matrix value of output layer, so that the context of output layer output minimizes the error;Wherein, the training sample pair
Including effective sample to and invalid sample pair;The effective sample pair, including two group of second detection information that the detection period is adjacent;
The invalid sample pair, including detection period non-conterminous two group of second detection information.
12. data processing method according to claim 11, which is characterized in that
First weight matrix, including multiple column vectors, each column vector are respectively the second characterization letter of each second detection information
Breath.
13. a kind of computer equipment, which is characterized in that including memory, processor and store on a memory and can handle
The computer program run on device when the processor executes described program, realizes such as base as claimed in any one of claims 1 to 6
In the data processing method of Driving Scene.
14. a kind of non-transitorycomputer readable storage medium, is stored thereon with computer program, which is characterized in that the program
Such as the data processing method as claimed in any one of claims 1 to 6 based on Driving Scene is realized when being executed by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910269776.6A CN110084137A (en) | 2019-04-04 | 2019-04-04 | Data processing method, device and computer equipment based on Driving Scene |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910269776.6A CN110084137A (en) | 2019-04-04 | 2019-04-04 | Data processing method, device and computer equipment based on Driving Scene |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110084137A true CN110084137A (en) | 2019-08-02 |
Family
ID=67414369
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910269776.6A Pending CN110084137A (en) | 2019-04-04 | 2019-04-04 | Data processing method, device and computer equipment based on Driving Scene |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110084137A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110705483A (en) * | 2019-10-08 | 2020-01-17 | Oppo广东移动通信有限公司 | Driving reminding method, device, terminal and storage medium |
CN111652153A (en) * | 2020-06-04 | 2020-09-11 | 北京百度网讯科技有限公司 | Scene automatic identification method and device, unmanned vehicle and storage medium |
CN113673403A (en) * | 2021-08-12 | 2021-11-19 | 深圳普捷利科技有限公司 | Driving environment detection method, system, device, computer equipment, computer readable storage medium and automobile |
CN114245303A (en) * | 2021-12-22 | 2022-03-25 | 诺博汽车系统有限公司 | Data acquisition method and device, readable storage medium and vehicle |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150317337A1 (en) * | 2014-05-05 | 2015-11-05 | General Electric Company | Systems and Methods for Identifying and Driving Actionable Insights from Data |
CN105389379A (en) * | 2015-11-20 | 2016-03-09 | 重庆邮电大学 | Rubbish article classification method based on distributed feature representation of text |
CN106445988A (en) * | 2016-06-01 | 2017-02-22 | 上海坤士合生信息科技有限公司 | Intelligent big data processing method and system |
CN108228757A (en) * | 2017-12-21 | 2018-06-29 | 北京市商汤科技开发有限公司 | Image search method and device, electronic equipment, storage medium, program |
CN108595636A (en) * | 2018-04-25 | 2018-09-28 | 复旦大学 | The image search method of cartographical sketching based on depth cross-module state correlation study |
CN108995657A (en) * | 2017-06-06 | 2018-12-14 | 百度(美国)有限责任公司 | Operate the method and data processing system of automatic driving vehicle |
-
2019
- 2019-04-04 CN CN201910269776.6A patent/CN110084137A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20150317337A1 (en) * | 2014-05-05 | 2015-11-05 | General Electric Company | Systems and Methods for Identifying and Driving Actionable Insights from Data |
CN105389379A (en) * | 2015-11-20 | 2016-03-09 | 重庆邮电大学 | Rubbish article classification method based on distributed feature representation of text |
CN106445988A (en) * | 2016-06-01 | 2017-02-22 | 上海坤士合生信息科技有限公司 | Intelligent big data processing method and system |
CN108995657A (en) * | 2017-06-06 | 2018-12-14 | 百度(美国)有限责任公司 | Operate the method and data processing system of automatic driving vehicle |
CN108228757A (en) * | 2017-12-21 | 2018-06-29 | 北京市商汤科技开发有限公司 | Image search method and device, electronic equipment, storage medium, program |
CN108595636A (en) * | 2018-04-25 | 2018-09-28 | 复旦大学 | The image search method of cartographical sketching based on depth cross-module state correlation study |
Non-Patent Citations (2)
Title |
---|
YUSUKE FUCHIDA等: "Driving word2vec:Distributed Semantic Vector Representation for Symbolized Naturalistic Driving Data", 《2016 IEEE INTELLIGENT VEHICLES SYMPOSIUM》 * |
施晋等: "时空依赖的城市道路旅行时间预测", 《软件学报》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110705483A (en) * | 2019-10-08 | 2020-01-17 | Oppo广东移动通信有限公司 | Driving reminding method, device, terminal and storage medium |
CN110705483B (en) * | 2019-10-08 | 2022-11-18 | Oppo广东移动通信有限公司 | Driving reminding method, device, terminal and storage medium |
CN111652153A (en) * | 2020-06-04 | 2020-09-11 | 北京百度网讯科技有限公司 | Scene automatic identification method and device, unmanned vehicle and storage medium |
CN111652153B (en) * | 2020-06-04 | 2023-12-22 | 北京百度网讯科技有限公司 | Scene automatic identification method and device, unmanned vehicle and storage medium |
CN113673403A (en) * | 2021-08-12 | 2021-11-19 | 深圳普捷利科技有限公司 | Driving environment detection method, system, device, computer equipment, computer readable storage medium and automobile |
CN114245303A (en) * | 2021-12-22 | 2022-03-25 | 诺博汽车系统有限公司 | Data acquisition method and device, readable storage medium and vehicle |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110084137A (en) | Data processing method, device and computer equipment based on Driving Scene | |
CN106294344B (en) | Video retrieval method and device | |
CN108470159A (en) | Lane line data processing method, device, computer equipment and storage medium | |
CN109871909A (en) | Image-recognizing method and device | |
CN108230353A (en) | Method for tracking target, system and electronic equipment | |
Ege et al. | Image-based food calorie estimation using recipe information | |
US20210090266A1 (en) | Method and device for labeling point of interest | |
CN109961433A (en) | Product defects detection method, device and computer equipment | |
CN109100142B (en) | A kind of semi-supervised method for diagnosing faults of bearing based on graph theory | |
CN110045734A (en) | Method, apparatus and computer equipment are determined based on the parameters weighting of path planning | |
CN106227884B (en) | A kind of recommended method of calling a taxi online based on collaborative filtering | |
CN109711285A (en) | Training, test method and the device of identification model | |
CN109710705A (en) | Map point of interest treating method and apparatus | |
CN109977826A (en) | The classification recognition methods of object and device | |
CN109977322A (en) | Trip mode recommended method, device, computer equipment and readable storage medium storing program for executing | |
CN110097121A (en) | A kind of classification method of driving trace, device, electronic equipment and storage medium | |
CN115131580B (en) | Space target small sample identification method based on attention mechanism | |
CN110084230A (en) | Vehicle body direction detection method and device based on image | |
CN110147837A (en) | The intensive object detection method of any direction, system and the equipment focused based on feature | |
CN109740632A (en) | Similarity model training method and device based on the more measurands of multisensor | |
WO2023159527A1 (en) | Detector training method and apparatus, and storage medium | |
CN105203327B (en) | A kind of gas circuit measurement parameter selection method applied to engine air passage analysis | |
CN104361600A (en) | Motion recognition method and system | |
CN109388781B (en) | Method and device for processing measurement data | |
Zhang et al. | Trajectory similarity assessment on road networks via embedding learning |
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 |