CN110188586A - System and application method based on meteorological observation, road camera shooting visibility identification - Google Patents
System and application method based on meteorological observation, road camera shooting visibility identification Download PDFInfo
- Publication number
- CN110188586A CN110188586A CN201810329654.7A CN201810329654A CN110188586A CN 110188586 A CN110188586 A CN 110188586A CN 201810329654 A CN201810329654 A CN 201810329654A CN 110188586 A CN110188586 A CN 110188586A
- Authority
- CN
- China
- Prior art keywords
- visibility
- module
- camera
- data
- observation
- 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.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/94—Hardware or software architectures specially adapted for image or video understanding
- G06V10/95—Hardware or software architectures specially adapted for image or video understanding structured as a network, e.g. client-server architectures
-
- 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
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/54—Surveillance or monitoring of activities, e.g. for recognising suspicious objects of traffic, e.g. cars on the road, trains or boats
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Databases & Information Systems (AREA)
- Traffic Control Systems (AREA)
Abstract
System based on meteorological observation, road camera shooting visibility identification estimates inspection module, interface service module including user interactive module, freeway surveillance and control image data acquiring module, the location information extraction module of highway camera, camera and weather station observation data association module, visibility observation data acquisition module, freeway surveillance and control image and visibility big data analysis module, the visibility recognition training module based on depth convolutional Neural grid, real time job scheduler module, visibility.The present invention is using highway camera shooting data as object, pass through machine learning and artificial intelligence technology, through related software collective effect, road visibility information is automatically extracted from high speed monitoring image, it is able to achieve sudden mist of playing a game and realizes effectively monitoring, monitoring information is provided for road foggy weather passage, increases the utility value of monitoring camera, provides advantageous Information Assurance to reduce the traffic weather disaster that dense fog induces.
Description
Technical field
The present invention relates to data analysis systems and analysis method field, especially a kind of to be imaged based on meteorological observation, road
The system and application method of visibility identification.
Background technique
In meteorological field, dense fog refers to that a large amount of small water droplets swim in the air, and weather of the horizontal visibility less than 1.0km is existing
As.Dense fog (or thick fog) is relatively common one of diastrous weather.It is high with occurrence probability, occurrence scope is wide, endangers journey
Spend big feature.There is its trace in China throughout the year, can all observe and learn its harm by experience.It is rapid with China's highway
Development, influence of the dense fog to highway is more prominent, and the traffic accidents as caused by dense fog are commonplace, dense fog
Lead to On The Deterioration of Visibility Over, is easy the more vehicles caused head and the tail and bumps against severe traffic accidents.
In dense fog, due to being misled by subjective ambient enviroment, it may appear that " people is confused in mist " phenomenon, driver always recognize
The safe speed and spacing that think from dense fog there are also a distance, such subjectivity for oneself and safe speed actually required and
It is larger away from differing, and it is less than the spacing that is perfectly safe at this moment visual range just, as long as the deceleration of front truck is greater than certain value,
Rear-end collision can occur.Another reason for dense fog initiation traffic accidents is the region that some highways are passed through
Weather conditions are more complicated, changeable, and different section visual range differences are larger, and driver is difficult to regulate the speed in time and spacing
And rear-end collision occurs.The beginning generally occurred in rear-end collision, since front truck can travel a distance before complete stop, because
The braking distance of this rear car is longer, and the severity accordingly collided is not very big relatively.When the vehicle of collision is more, before this
The vehicle bumped against a bit has been essentially resting, trails the vehicle to come up below in this way and is not only easier to continue to bump against, and bumps against
When speed it is higher, consequence is more serious, this is why under dense fog weather conditions, once traffic thing occurs for highway
Therefore generally all more serious reason.
Although the traffic accidents that closing high-speed highway can prevent dense fog from causing, its cost is apparent
's.Therefore, some developed countries test installation dense fog on some important highways and alert automatically and speed control in recent years
System processed installs visibility detector every 400-1000m on the basis of original traffic surveillance and control system, while every 300-
500m installation variable speed-limit sign, the visibility distance that traffic control system is detected according to different sections of highway visibility detector, in real time
The reason of controlling the speed limit value of each speed(-)limit sign, while also showing speed limit, such system not only can remind driver to pay attention in time
Front dense fog can also recommend safe speed to driver.West Europe some country practice have shown that: this system for reduce dense fog
The traffic accidents of initiation are very effective, and the dense fog of 50m is not less than especially for visibility distance, and it is public to be still able to maintain high speed
The normal operation on road, and need not closing high-speed highway.Certainly, this full-automatic fog warning and speed control system to be installed
System, expense is fairly expensive.China can only be larger in the volume of traffic at present, considers that experiment is this on the multiple highway of dense fog
System.Based on above-mentioned, data are observed by the visibility meter of expressway weather observation station, realize the greasy weather to height for traffic department
Greasy weather vehicle pass-through scheme is then formulated in the effective monitoring of fast highway, and reducing traffic accident seems particularly necessary, but
It is that a expressway weather observation station can monitor visibility data only more than 1300 at present in China, it is difficult to meet under foggy weather
The current demand ensured of expressway safety.On the other hand, major part highway in China's is assembled with number and simulation at present
Mixed video monitoring system, the quantity of national highway camera monitoring system is 10,000 or more, and camera is in addition to having taken road
And the case where vehicle, weather phenomenon is also recorded or even reflects the information of atmospheric visibility, and can be in minute grade interval
Record, these data can be used to carry out Real Time Monitoring to expressway visibility.But existing application technology is to taking the photograph
Analysis and utilization can not be effectively carried out as head obtains data, thus actual needs far can not be reached.
Summary of the invention
In order to overcome existing freeway traffic weather station that can not carry out effective greasy weather monitoring and highway camera
The effect data of acquisition is bad, can not formulate the drawbacks of corresponding measure provides effective help for relevant department, the present invention provides
One kind is using highway camera shooting data as object, by machine learning and artificial intelligence technology, automatically from high speed monitoring image
Middle extraction road visibility information is able to achieve sudden mist of playing a game and realizes effectively monitoring, provides for road foggy weather passage
Monitoring information formulates related counter-measure for relevant department and provides reasonable data supporting, and thereby reduces traffic accident hair
The raw system and application method based on meteorological observation, road camera shooting visibility identification.
The technical solution adopted by the present invention to solve the technical problems is:
System based on meteorological observation, road camera shooting visibility identification, it is characterised in that including user interactive module, high speed
Highway monitoring image data acquiring module, the location information extraction module of highway camera, camera and weather station are observed
Data association module, visibility observation data acquisition module, freeway surveillance and control image and visibility big data analysis module, base
Inspection module is estimated in the visibility recognition training module, real time job scheduler module, visibility of depth convolutional Neural grid, is connect
Mouth service module.
The user interactive module, the main input and output for completing analysis platform and user, parameter setting, intelligent method are calculated
The interaction of library, threshold library etc..
The freeway surveillance and control image data acquiring module allows to import different video compressed encoding and different-format is deposited
The video file of storage extracts image data with the shooting interval of setting, identifies the shooting time of image;Allow to carry out not
With the conversion between picture format, the picture formats such as JPG, BMP, PNG, TIFF of mainstream are supported;It supports to carry out the size of image
Cutting and resolution adjustment;It supports to carry out color stretching and setting contrast to image.
The location information extraction module of the highway camera represents highway camera with road pile No.
Camera site is extracted the latitude and longitude information of each camera position, is compiled with highway according to the pile No. information of highway
Number, pile No., shooting date and time establish image Naming conventions, Uniform Name is carried out to all images by specification, by all figures
As being input in image data base, in addition to image data, there are also highway number, camera latitude and longitude information, date informations
Including year, month, day, hour, min.
Data association module is observed in the camera and weather station, and function is as follows: A, with the variable distance of setting, according to
Highway pile No. coordinate file generates buffer area;B will be located in buffer area automatically according to the geographical coordinate of weather station
Weather station extract;C calculates each camera apart from weather station according to the coordinate of the position of camera and weather station
Space length, take the space correlation visibility observation station apart from the smallest weather station as camera.
The visibility observes data acquisition module, obtains expressway weather observation station or periphery associated by camera
Visibility that meteorological station is observed, temperature, relative temperature, wind speed, wind direction information.Extract the longitude and latitude of each meteorological station
Information.Specification is named to the meteorological data file of observation with site number, observation date and time, by specification to owning
Meteorological measuring file in buffer area carries out Uniform Name, is input to all in database, in addition to meteorological element data,
There are also weather station latitude and longitude information, date information includes year, month, day, hour, min, Road Weather calamity source assessment models,
Weather station 30 years reorganization history meteorological measurings in periphery are handled, respectively statistics frost sleet, low visibility, strong wind and
The frequency that precipitation occurs carries out calamity source grade classification according to related grade scale.
The freeway surveillance and control image and visibility big data analysis module, mainly by distributed memory system, distribution
Formula scheduling of resource frame, big data Computational frame, scientific research sample generate, five part of sample preprocessing forms.
The visibility recognition training module based on depth convolutional Neural grid, it is real under the support in intelligent algorithm library
The now monitoring photo sky identification function based on depth convolutional Neural grid;The building of visibility identification model, parameter tune are provided
Whole and training;The update of training parameter file.
The real time job scheduler module provides historical data processing function, realizes photo and weather observation data in the past
Processing and storage;Visibility model training function is provided;Real time job Parameter File is set, realizes that backstage automatically processes in real time
Data reads monitoring picture in real time and provides visibility estimated value.
The visibility estimates inspection module, calculates the root-mean-square error and TS scoring of visibility estimation.
The interface service realizes the visibility estimation program of the adjustment highway camera shooting photo based on machine learning
Calling function provides interface service, inputs picture, returns to visibility estimated value.
System application method the following steps based on meteorological observation, road camera shooting visibility identification: 1, pass through high speed
The location information extraction module of highway camera, freeway surveillance and control image data acquiring module to freeway surveillance and control image and
Related data is acquired processing;2, it observes data by the location information extraction module of highway camera, visibility and adopts
Collection module is acquired expressway weather observation station data;3, data acquisition module is observed to meteorological station by visibility
Visibility data are handled;4, data association module is observed by camera and weather station and carries out camera and weather station
Spatial match;5, by freeway surveillance and control image and visibility big data analysis module, realizes and be based on depth convolutional Neural net
The sky detection of network (DCN);6, by the visibility recognition training module based on depth convolutional Neural grid, realizes and be based on depth
The visibility of convolutional neural networks (DCN) identifies.
The medicine have the advantages that the present invention passes through machine learning and artificial intelligence using highway camera shooting data as object
Energy technology, the location information through user interactive module, freeway surveillance and control image data acquiring module, highway camera mention
Modulus block, camera and weather station observation data association module, visibility observe data acquisition module, freeway surveillance and control image
It is dispatched with visibility big data analysis module, the visibility recognition training module based on depth convolutional Neural grid, real time job
Module, visibility estimate inspection module, interface service module collective effect, and road energy is automatically extracted from high speed monitoring image
Degree of opinion information, is able to achieve sudden mist of playing a game and realizes effectively monitoring, monitoring information is provided for road foggy weather passage, to have
Pass department formulates related counter-measure and provides valid data support.The following whole nation highway planning will realize system-wide and the whole network
Monitoring camera-shooting, technology, a camera are equivalent to a visibility visualizer, are equivalent between 2-10km according to the present invention
Every the national expressway visibility real-time monitoring of upper realization, it is possible to reduce the huge cost that visibility meter is laid, while increasing prison
The utility value of camera is controlled, more can provide advantageous Information Assurance to reduce the traffic weather disaster that dense fog induces.It is based on
It is above-mentioned, so the application prospect that the present invention has had.
Detailed description of the invention
The present invention will be further described with reference to the accompanying drawings and examples.
Fig. 1 is that the present invention is based on the systems of meteorological observation, road camera shooting visibility identification to constitute block diagram.
Fig. 2 is that the present invention is based on the system application method workflow block diagrams of meteorological observation, road camera shooting visibility identification.
Fig. 3 is application method operating process block diagram of the present invention.
Fig. 4 is the modular unit block architecture diagram that the present invention uses.
Fig. 5 is the depth convolutional Neural grid chart of sky identification
Fig. 6 is the depth convolutional Neural grid chart of visibility identification.
Specific embodiment
Shown in Fig. 1, visibility identifying system, including user interactive module, high speed public affairs are imaged based on meteorological observation, road
Number is observed in road monitoring image data acquisition module, the location information extraction module of highway camera, camera and weather station
According to relating module, visibility observation data acquisition module, freeway surveillance and control image and visibility big data analysis module, it is based on
Visibility recognition training module, real time job scheduler module, the visibility of depth convolutional Neural grid estimate inspection module, interface
Service module.
Shown in Fig. 1,2,3,4, user interactive module, main completion analysis platform and the input and output of user, parameter are set
It sets, intelligent method calculates the interaction of library, threshold library etc..Freeway surveillance and control image data acquiring module allows to import different video pressure
The video file for reducing the staff code and different-format storage, extracts image data with the shooting interval of setting, identifies image
Shooting time;Allow to carry out the conversion between different images format, supports the picture formats such as JPG, BMP, PNG, TIFF of mainstream;Branch
It holds and cutting and resolution adjustment is carried out to the size of image;It supports to carry out color stretching and setting contrast to image.High speed is public
The location information extraction module of road camera, the camera site of highway camera is represented with road pile No., according to high speed
The pile No. information of highway, extracts the latitude and longitude information of each camera position, with highway number, pile No., shooting date
Image Naming conventions are established with the time, Uniform Name is carried out to all images by specification, all images are input to image data
In library, in addition to image data, there are also highway number, camera latitude and longitude information, date information include year, month, day, when,
Point.
Shown in Fig. 1,2,3,4, data association module is observed in camera and weather station, and function is as follows: A, with setting can
The distance of change generates buffer area according to highway pile No. coordinate file;B automatically will according to the geographical coordinate of weather station
Weather station in buffer area extracts;C calculates each camera shooting according to the coordinate of the position of camera and weather station
Space length of the head apart from weather station, takes the space correlation visibility observation station apart from the smallest weather station as camera.Energy
Degree of opinion observes data acquisition module, obtains expressway weather observation station associated by camera or periphery meteorological station is observed
Visibility, temperature, relative temperature, wind speed, wind direction information, extract the latitude and longitude information of each meteorological station, compiled with website
Number, observation date and time specification is named to the meteorological data file of observation, by standardizing to the gas in all buffer areas
As observation data file progress Uniform Name, it is input to all in database, in addition to meteorological element data, there are also meteorological observations
It stands latitude and longitude information, date information includes year, month, day, hour, min, Road Weather calamity source assessment models, to periphery weather station
Reorganization history meteorological measuring is handled within 30 years, and statistics frost sleet, low visibility, strong wind and precipitation occur respectively
The frequency carries out calamity source grade classification according to related grade scale.
Shown in Fig. 1,2,3,4, freeway surveillance and control image and visibility big data analysis module are mainly deposited by distribution
Five parts such as storage system, distributed resource scheduling frame, big data Computational frame, the generation of scientific research sample, sample preprocessing composition.
1, distributed memory system: in order to solve the bottlenecks and availability, scalability etc. such as single machine storage existing capacity, performance
The problem of aspect, provides large capacity, height by the way that data dispersion is stored in more storage equipment for large-scale storage application
Performance, High Availabitity, favorable expandability storage service;It is to solve large volume file at the beginning of due to distributed memory system architecture design
Storage, while considering that meteorological data has the characteristics that quantity of documents is mostly big with file population product, it is therefore desirable at existing point
It is transformed and extends on the basis of cloth storage system (HDFS), design reasonable data store organisation to solve magnanimity meteorology
The storage problem of data.2, distributed resource scheduling frame: by using YARN as big data and machine learning Demonstration Platform
Distributed resource scheduling frame;YARN is next-generation MapReduce frame, and the frame is mainly from MRvl resource management framework
Decoupling comes out, and provides scheduling feature for each application component;YARN is mainly consisted of three parts: ResourceManager,
NodeManager,ApplicationMaster;ResourceManager is a global resource manager, is responsible for entire
The resource management and distribution of system, it is the most crucial module of YARN frame;It is mainly made of scheduler and application manager;
NodeManager is the resource and task management device on each node;On the one hand, it can periodically converge to ResourceManager
Report the resource service condition of this node and the operating status of each Container (abstract concept of resource allocation unit);It is another
Aspect, it can receive and process the various requests such as task start/stopping from ApplicationMaster;What user submitted
Each application program includes an AM, it is actually the JobTracker of a simplified version, major function include: with
ResourceManager scheduler is negotiated to obtain resource, be communicated with NodeManager to start/stopping task, monitoring institute
There is the operating status of task, and is again task application resource to restart task when task transports skilful failure;Utilize the collection of YARN
Group resource management function, can effectively improve the utilization rate of PC cluster resource, reduce task distribution, data sharing at
This.3, big data Computational frame: Spark is chosen as big data and the big data Computational frame of machine learning Demonstration Platform;
Spark is by the open source universal parallel cloud computing platform of UC Berkeley AMP development in laboratory, and Spark is based on
The distributed computing that MapReduce thought is realized, possesses advantage possessed by Hadoop MapReduce;But different places are fortune
Calculating intermediate output result can be stored in memory, to no longer need to read and write HDFS, therefore Spark can preferably operation data be dug
Pick and machine learning etc. need the MapReduce algorithm of iteration.4, scientific research sample quickly generates: being submitted according to scientific research personnel
Data processing request using the big data computing basic facility of first three layer building, and utilizes JNI technology, processing rapidly and efficiently
Involved original meteorological data in big data and machine learning Demonstration Platform, and the actual demand that is generated according to sample into
Row data, which are extracted, quickly generates work with combine the initial sample of completion.5, sample preprocessing: the sample submitted according to scientific research personnel
Present treatment request carries out feature selecting to sample and original sample is divided into trained sample using big data computing basic facility
Sheet and test sample, for model training, model parameter tuning and model selection.
Shown in Fig. 1,2,3,4, the visibility recognition training module based on depth convolutional Neural grid, in intelligent algorithm library
Support under, realize the monitoring photo sky identification function based on depth convolutional Neural grid;Visibility identification model is provided
It builds, parameter adjusts and training;The update of training parameter file.Real time job scheduler module provides historical data processing function,
Realize the processing and storage of photo and weather observation data in the past;Visibility model training function is provided;Real time job is arranged to join
Number file, realizes that backstage automatically processes real time data, reads monitoring picture in real time and provides visibility estimated value.Visibility estimation
Inspection module calculates the root-mean-square error and TS scoring of visibility estimation.Interface service module realizes the tune based on machine learning
The visibility estimation program calling function of whole highway camera shooting photo, provides interface service, inputs picture, return to visibility
Estimated value.
Shown in Fig. 2, the present embodiment will be seen by taking the identification of Anhui Province's highway camera visibility as an example based on meteorological
It surveys, the system application method of road camera shooting visibility identification illustrates.1, pass through the location information of highway camera
Extraction module, freeway surveillance and control image data acquiring module are acquired place to freeway surveillance and control image and related data
Reason;Anhui the whole province highway camera data file of collection is handled, according to 60 minutes time intervals from video flowing
Image data is extracted in file;Analysis extraction is carried out to the location information of highway camera, is represented with road pile No.
The camera site of highway camera;According to the pile No. information of highway, the longitude and latitude of each camera position is extracted
Information;Image Naming conventions are established with highway number, pile No., shooting date and time, all images are carried out by specification
All images are input in image data base by Uniform Name, and in addition to image data, there are also highway numbers, camera warp
Latitude information, date information include year, month, day, hour, min.2, pass through location information extraction module, the energy of highway camera
Degree of opinion observation data acquisition module is acquired expressway weather observation station data;It acquires meteorological on Anhui Province's highway
The visibility data of observation station observation, carry out analysis extraction to the location information of expressway weather observation station, with road pile No.
To represent expressway weather station location;According to the pile No. information of highway, each expressway weather observation station is extracted
Latitude and longitude information;The meteorological data file of observation is named with highway number, pile No., observation date and time
Specification carries out Uniform Name to all expressway weathers observation data file by specification;It is input in database, removes by all
Meteorological element data, there are also highway number, expressway weather observation station latitude and longitude information, date information include year,
The moon, day, when, point.3, data acquisition module is observed by visibility, meteorological station visibility is handled;It is public with distance high speed
Road 50km is buffer area, determines the number for being located at the meteorological station in buffer area having visibility to observe, obtains these meteorological observations
Visibility, the temperature, relative temperature, wind speed, wind direction information of interval observation in the station every 10 minutes;Extract each meteorological station
Latitude and longitude information;Specification is named to the meteorological data file of observation with site number, observation date and time, by standardizing
Uniform Name is carried out to the meteorological measuring file in all buffer areas;All data are input in database, in addition to gas
As factor data, there are also weather station latitude and longitude information, date information includes year, month, day, hour, min.4, by camera with
Weather station observes data association module and carries out camera and weather station spatial match;Utilize longitude and latitude where camera and gas
As the longitude and latitude of observation station, space length calculating is carried out, it is in place to represent camera institute away from the nearest meteorological station of camera
The meteorological observation set;It is replaced in the case where weather observation data missing using secondary close weather station;After completing matching, so that it may
Nearest weather station corresponding with shooting time to be observed according to its shooting date for each monitoring photo
Visibility data corresponded with photo, be X with picture data, with corresponding visibility observation data be Y.
Fig. 2, shown in 5, it is public to pass through high speed for the system application method 5 based on meteorological observation, road camera shooting visibility identification
Road monitoring image and visibility big data analysis module realize the sky detection for being based on depth convolutional neural networks (DCN);It can see
Degree is the bulk of optical feedback of atmospheric physics state, and only in the presence of having 1/4 or more sky in image, image is just suitable for identifying
Visibility builds training and classification that depth convolutional neural networks carry out sky by manually marking sky sample;Detailed process
It is as follows:
Step1: the random sampling of picture data;
Step2: manual identification has 1/4 or more sky to be then identified as sky photo in photo, is otherwise identified as non-sky and shines
Piece;
Step3: following depth convolutional neural networks are established:
Step4: setting objective function and searching algorithm;
Step5: it is iterated training;
Step6: being based on training parameter, carries out sky identification to all pictures;
Step7: removal does not identify the camera data of sky.
Fig. 2, shown in 6, the system application method 6 based on meteorological observation, road camera shooting visibility identification, by based on deep
The visibility recognition training module of convolutional Neural grid is spent, realizes that the visibility based on depth convolutional neural networks (DCN) identifies;
It is X with picture data, is Y with corresponding visibility observation data, the photo of high-speed capture is mostly color mode at present, photo
Red, green, blue channel can be equivalent to a three-dimensional data as the 3rd dimension, such photo, and the content of array is photo
The gray value of 0-255, by work is drilled in the presence of having 1/4 or more sky in image, image is left to see for identification
Degree, visibility is continuous physical quantity, identifies and can see directly from photo so building depth convolutional neural networks regression model
Degree value (being 3 layers of oriented cycles network in practical operation), detailed process is as follows:
Step1: the random sampling of picture data (X) and visibility observation (Y), for training, a part is used for a part
It examines;
Step2: depth convolutional neural networks have very huge demand to the performance of computer, need in practical work process
Resolution decreasing operation is carried out to picture data for existing computer resource, it is right using the resampling of image processing tool
Image data take out thick;
Step3: sample cleaning, the clear sky sample identification accuracy on the high side that will affect visibility are needed according to meteorological observation
Visibility, selected to greater than the sample of 1km;
Step4: establishing following depth convolutional neural networks, and with the tri-dimensional picture of highway shooting, (the 3rd dimension is r, g, b
Corresponding color channel) it is X, it is Y with visibility observation, it is finally defeated with full articulamentum by convolution, 3 circulations in pond
Visibility numerical value out.
Step5: setting objective function and searching algorithm;
Step6: it is iterated training;
Step7: being based on training parameter, carries out visibility identification to all pictures;
Step8: estimating visibility on test set, and carries out recognition accuracy according to estimated value and measured value and comment
Valence.
Shown in Fig. 2, operating process is as follows in training mode by the present invention: the acquisition of A highway camera data, by fixation
Time interval abstract image file, unloading are unified format and are named according to specification;It is geographical that B obtains highway camera
Position is converted by pile No. information, generates latitude and longitude coordinates information;C camera and weather station space correlation, according to
Camera and weather station geographical location calculate spherical distance, are associated by minimum range rule;The observation of D visibility
Data acquisition obtains the visibility instrument sight for beginning and ending time corresponding weather station acquisition of taking pictures with highway camera
Measured data;E freeway surveillance and control image and visibility big data analysis generate the sample for the training of depth convolutional neural networks
This simultaneously carries out sample preprocessing;Visibility recognition training of the F based on depth convolutional Neural grid saves training result;G visibility
Training effect is examined.
Shown in Fig. 2, the present invention is as follows in business model operating process: A highway camera data acquires and pre- in real time
Processing, extracts photo at regular intervals, formats and resolution adjustment, carries out image stretch;B calls backstage
Program carries out visibility estimation to real-time monitoring photo;C exports visibility estimation result.
Shown in Fig. 2, the present invention is as follows in interface service mode operating process: A receives service request, mentions in request message
Pictorial information and spatial positional information are supplied;B downloads image data, formats to image and resolution adjustment, carries out
Image stretch;C obtain picture location information, call the training parameter file of corresponding region, by pretreated picture as
Output runs visibility identification module;D carries out quality control and adjustment to visibility value, returns to visibility value.
Shown in Fig. 2, steps are as follows using computer software flow operations by the present invention: 1, start computer, and start and be
System;2, it opens highway and images picture collection module, mentioned from the data that highway camera is shot by 60 minutes intervals
Picture is taken out, JPG format is unified for;3, meteorological observation visibility acquisition module is opened, by expressway weather observation station and peace
The visibility data of emblem Meteorological Bureau of Shanxi Province meteorological observation station observation was extracted according to 60 minutes intervals;4, according to where camera
The pile No. information analysis of position goes out the geographical coordinate where camera, and according to the geographical coordinate of meteorological observation station location, calculating is taken the photograph
As the space length of head and weather station, there is visibility observation to each camera shooting association one according to apart from nearest principle
Weather station;5, start big data analysis module, according to the related information of camera and weather station, high speed is monitored
Picture is handled with visibility value, generates the sample database for being used for machine learning;6, start visibility recognition training module, to root
Model is trained according to the sample file that previous step generates, saves training parameter file;7, visibility is instructed on test set
Practice result to test;8 are directed to new monitoring picture data, and machine learning model is called to carry out visibility estimation;9, starting connects
Mouth service routine estimates message according to visibility, obtains the image data in message, visibility appraising model is called to provide and can see
The estimation of angle value, and return to visibility value.
The present invention is using highway camera shooting data as object, by machine learning and artificial intelligence technology, through user's interaction
Module, freeway surveillance and control image data acquiring module, highway camera location information extraction module, camera and gas
As station observation data association module, visibility observation data acquisition module, freeway surveillance and control image and visibility big data point
Analyse module, the visibility recognition training module based on depth convolutional Neural grid, real time job scheduler module, visibility estimation inspection
Module, interface service module collective effect are tested, road visibility information is automatically extracted from high speed monitoring image, is able to achieve pair
Sudden mist of office realizes effectively monitoring, provides monitoring information for road foggy weather passage, formulates correlation for relevant department and answer
Valid data support is provided to measure.The following whole nation highway planning will realize system-wide and the whole network monitoring camera-shooting, according to this
Inventive technique, a camera are equivalent to a visibility visualizer, are equivalent to and realize that the whole nation is high on the interval of 2-10km
Fast highway visibility real-time monitoring, it is possible to reduce the huge cost that visibility meter is laid, while increasing the utilization of monitoring camera
Value more can provide advantageous Information Assurance to reduce the traffic weather disaster that dense fog induces.Based on above-mentioned, so of the invention
With good application prospect.
Basic principles and main features and advantages of the present invention of the invention have been shown and described above, for this field skill
For art personnel, it is clear that invention is not limited to the details of the above exemplary embodiments, and without departing substantially from spirit of the invention or
In the case where essential characteristic, the present invention can be realized in other specific forms.Therefore, in all respects, should all incite somebody to action
Embodiment regards exemplary as, and is non-limiting, the scope of the present invention by appended claims rather than on state
Bright restriction, it is intended that including all changes that fall within the meaning and scope of the equivalent elements of the claims in the present invention
It is interior.Any reference signs in the claims should not be construed as limiting the involved claims.
In addition, it should be understood that although this specification is described in terms of embodiments, but not each embodiment is only wrapped
Containing an independent technical solution, this description of the specification is merely for the sake of clarity, and those skilled in the art should
It considers the specification as a whole, the technical solutions in the various embodiments may also be suitably combined, forms those skilled in the art
The other embodiments being understood that.
Claims (9)
1. the system based on meteorological observation, road camera shooting visibility identification, it is characterised in that public including user interactive module, high speed
Number is observed in road monitoring image data acquisition module, the location information extraction module of highway camera, camera and weather station
According to relating module, visibility observation data acquisition module, freeway surveillance and control image and visibility big data analysis module, it is based on
Visibility recognition training module, real time job scheduler module, the visibility of depth convolutional Neural grid estimate inspection module, interface
Service module.
2. the system according to claim 1 based on meteorological observation, road camera shooting visibility identification, it is characterised in that high speed
Highway monitoring image data acquiring module allows to import the video file of different video compressed encoding and different-format storage, with
The shooting interval of setting extracts image data, identifies the shooting time of image;Allow carry out different images format between
Conversion, supports the picture formats such as JPG, BMP, PNG, TIFF of mainstream;It supports to carry out the size of image cutting and resolution ratio tune
It is whole;It supports to carry out color stretching and setting contrast to image.
3. the system according to claim 1 based on meteorological observation, road camera shooting visibility identification, it is characterised in that high speed
The location information extraction module of highway camera, the camera site of highway camera is represented with road pile No., according to height
The pile No. information of fast highway, extracts the latitude and longitude information of each camera position, with highway number, pile No., shooting day
Phase and time establish image Naming conventions, carry out Uniform Name to all images by specification, all images are input to picture number
According in library, in addition to image data, there are also highway number, camera latitude and longitude information, date information include year, month, day,
When, point.
4. the system according to claim 1 based on meteorological observation, road camera shooting visibility identification, it is characterised in that camera shooting
Head observes data association module with weather station, and function is as follows: A, with the variable distance of setting, according to highway pile No. coordinate
File generated buffer area;B automatically extracts the weather station being located in buffer area according to the geographical coordinate of weather station;
C calculates space length of each camera apart from weather station, takes distance according to the coordinate of the position of camera and weather station
Space correlation visibility observation station of the smallest weather station as camera.
5. the system according to claim 1 based on meteorological observation, road camera shooting visibility identification, it is characterised in that can see
Degree observes data acquisition module, what expressway weather observation station associated by acquisition camera or periphery meteorological station were observed
Visibility, temperature, relative temperature, wind speed, wind direction information;Extract the latitude and longitude information of each meteorological station, with site number,
Date and time is observed to be named specification to the meteorological data file of observation, the meteorology in all buffer areas is seen by specification
Measured data file carries out Uniform Name, is input to all in database, in addition to meteorological element data, there are also weather station warps
Latitude information, date information included year, month, day, hour, min, Road Weather calamity source assessment models, to periphery weather station 30 years
Reorganization history meteorological measuring is handled, the frequency that statistics frost sleet, low visibility, strong wind and precipitation occur respectively
It is secondary, calamity source grade classification is carried out according to related grade scale.
6. the system according to claim 1 based on meteorological observation, road camera shooting visibility identification, it is characterised in that high speed
Highway monitoring image and visibility big data analysis module, mainly by distributed memory system, distributed resource scheduling frame, big
Data Computational frame, scientific research sample generate, five part of sample preprocessing forms.
7. the system according to claim 1 based on meteorological observation, road camera shooting visibility identification, it is characterised in that be based on
The visibility recognition training module of depth convolutional Neural grid is realized under the support in intelligent algorithm library based on depth convolution mind
Monitoring photo sky identification function through grid;The building of visibility identification model, parameter adjustment and training are provided;Training parameter
The update of file.
8. the system according to claim 1 based on meteorological observation, road camera shooting visibility identification, it is characterised in that in real time
Job scheduling module provides historical data processing function, realizes the processing and storage of photo and weather observation data in the past;It provides
Visibility model training function;Real time job Parameter File is set, realizes that backstage automatically processes real time data, reads monitoring in real time
Picture simultaneously provides visibility estimated value.
9. the system according to claim 1 based on meteorological observation, road camera shooting visibility identification, it is characterised in that application
Method the following steps: 1, pass through the location information extraction module of highway camera, freeway surveillance and control image data
Acquisition module is acquired processing to freeway surveillance and control image and related data;2, believed by the position of highway camera
Breath extraction module, visibility observation data acquisition module are acquired expressway weather observation station data;3, by that can see
Degree observation data acquisition module handles meteorological station visibility data;4, data are observed by camera and weather station and are closed
Gang mould block carries out camera and weather station spatial match;5, pass through freeway surveillance and control image and visibility big data analysis
Module realizes the sky detection for being based on depth convolutional neural networks (DCN);6, pass through can see based on depth convolutional Neural grid
Recognition training module is spent, realizes that the visibility based on depth convolutional neural networks (DCN) identifies.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810329654.7A CN110188586A (en) | 2018-04-13 | 2018-04-13 | System and application method based on meteorological observation, road camera shooting visibility identification |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810329654.7A CN110188586A (en) | 2018-04-13 | 2018-04-13 | System and application method based on meteorological observation, road camera shooting visibility identification |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110188586A true CN110188586A (en) | 2019-08-30 |
Family
ID=67713912
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810329654.7A Withdrawn CN110188586A (en) | 2018-04-13 | 2018-04-13 | System and application method based on meteorological observation, road camera shooting visibility identification |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110188586A (en) |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110910378A (en) * | 2019-11-29 | 2020-03-24 | 南通大学 | Bimodal image visibility detection method based on depth fusion network |
CN111341118A (en) * | 2020-02-28 | 2020-06-26 | 长安大学 | System and method for early warning of mist on grand bridge |
CN111526193A (en) * | 2020-04-17 | 2020-08-11 | 中国科学院上海微系统与信息技术研究所 | Distributed weather station system based on fog calculation and block chain platform and working method |
CN112258853A (en) * | 2020-10-19 | 2021-01-22 | 洛阳云感科技有限公司 | Chain visibility monitoring and early warning system in highway fog zone becomes more meticulous |
CN112649900A (en) * | 2020-11-27 | 2021-04-13 | 上海眼控科技股份有限公司 | Visibility monitoring method, device, equipment, system and medium |
CN113034902A (en) * | 2021-03-05 | 2021-06-25 | 阿里云计算有限公司 | Traffic control method and device and electronic equipment |
CN113487873A (en) * | 2021-05-20 | 2021-10-08 | 刘铁拴 | Intelligent detection system for road traffic safety |
CN113609406A (en) * | 2021-07-27 | 2021-11-05 | 杭州鸿泉物联网技术股份有限公司 | High-altitude operation trend information sharing method, system and equipment based on geocoding |
CN113611123A (en) * | 2021-10-09 | 2021-11-05 | 智广海联(天津)大数据技术有限公司 | Auxiliary study and judgment treatment method and system for muck truck |
CN113704207A (en) * | 2021-08-10 | 2021-11-26 | 南京信息工程大学 | High-altitude meteorological data compression and decoding method based on video compression technology |
CN113870560A (en) * | 2021-09-27 | 2021-12-31 | 公安部道路交通安全研究中心 | Highway portrayal evaluation and safety risk research and judgment system based on multi-source data |
CN113901092A (en) * | 2021-12-07 | 2022-01-07 | 中国气象局气象探测中心 | File management method and system for racing field operation monitoring |
CN114360270A (en) * | 2021-07-26 | 2022-04-15 | 云南省交通规划设计研究院有限公司 | Method and system for studying and judging maximum allowable speed of highway under adverse weather influence |
WO2022176643A1 (en) * | 2021-02-16 | 2022-08-25 | 株式会社フジタ | Information provision system |
CN115035723A (en) * | 2022-07-11 | 2022-09-09 | 长安大学 | Highway visibility monitoring and early warning system based on deep learning |
US20220380990A1 (en) * | 2021-05-26 | 2022-12-01 | Here Global B.V. | Apparatus and method for controlling a functional road object |
CN116486347A (en) * | 2023-02-21 | 2023-07-25 | 河南省气象探测数据中心(河南省气象档案馆) | Method and device for monitoring and capturing fog based on scale-invariant feature transformation image recognition |
CN117269011A (en) * | 2023-11-13 | 2023-12-22 | 珠海光焱科技有限公司 | Meteorological visibility judging method based on laser light path measuring system |
-
2018
- 2018-04-13 CN CN201810329654.7A patent/CN110188586A/en not_active Withdrawn
Cited By (26)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110910378B (en) * | 2019-11-29 | 2021-11-02 | 南通大学 | Bimodal image visibility detection method based on depth fusion network |
CN110910378A (en) * | 2019-11-29 | 2020-03-24 | 南通大学 | Bimodal image visibility detection method based on depth fusion network |
CN111341118A (en) * | 2020-02-28 | 2020-06-26 | 长安大学 | System and method for early warning of mist on grand bridge |
CN111526193A (en) * | 2020-04-17 | 2020-08-11 | 中国科学院上海微系统与信息技术研究所 | Distributed weather station system based on fog calculation and block chain platform and working method |
CN112258853A (en) * | 2020-10-19 | 2021-01-22 | 洛阳云感科技有限公司 | Chain visibility monitoring and early warning system in highway fog zone becomes more meticulous |
CN112649900A (en) * | 2020-11-27 | 2021-04-13 | 上海眼控科技股份有限公司 | Visibility monitoring method, device, equipment, system and medium |
WO2022176643A1 (en) * | 2021-02-16 | 2022-08-25 | 株式会社フジタ | Information provision system |
CN113034902A (en) * | 2021-03-05 | 2021-06-25 | 阿里云计算有限公司 | Traffic control method and device and electronic equipment |
CN113487873A (en) * | 2021-05-20 | 2021-10-08 | 刘铁拴 | Intelligent detection system for road traffic safety |
US20220380990A1 (en) * | 2021-05-26 | 2022-12-01 | Here Global B.V. | Apparatus and method for controlling a functional road object |
CN114360270B (en) * | 2021-07-26 | 2023-02-14 | 云南省交通规划设计研究院有限公司 | Method and system for studying and judging maximum allowable speed of highway under adverse weather influence |
CN114360270A (en) * | 2021-07-26 | 2022-04-15 | 云南省交通规划设计研究院有限公司 | Method and system for studying and judging maximum allowable speed of highway under adverse weather influence |
CN113609406B (en) * | 2021-07-27 | 2024-02-20 | 杭州鸿泉物联网技术股份有限公司 | Overhead operation wind situation information sharing method, system and equipment based on geocoding |
CN113609406A (en) * | 2021-07-27 | 2021-11-05 | 杭州鸿泉物联网技术股份有限公司 | High-altitude operation trend information sharing method, system and equipment based on geocoding |
CN113704207A (en) * | 2021-08-10 | 2021-11-26 | 南京信息工程大学 | High-altitude meteorological data compression and decoding method based on video compression technology |
CN113704207B (en) * | 2021-08-10 | 2023-05-09 | 南京信息工程大学 | High-altitude meteorological data compression and decoding method based on video compression technology |
CN113870560A (en) * | 2021-09-27 | 2021-12-31 | 公安部道路交通安全研究中心 | Highway portrayal evaluation and safety risk research and judgment system based on multi-source data |
CN113611123B (en) * | 2021-10-09 | 2021-12-17 | 智广海联(天津)大数据技术有限公司 | Auxiliary study and judgment treatment method and system for muck truck |
CN113611123A (en) * | 2021-10-09 | 2021-11-05 | 智广海联(天津)大数据技术有限公司 | Auxiliary study and judgment treatment method and system for muck truck |
CN113901092B (en) * | 2021-12-07 | 2022-03-18 | 中国气象局气象探测中心 | File management method and system for racing field operation monitoring |
CN113901092A (en) * | 2021-12-07 | 2022-01-07 | 中国气象局气象探测中心 | File management method and system for racing field operation monitoring |
CN115035723A (en) * | 2022-07-11 | 2022-09-09 | 长安大学 | Highway visibility monitoring and early warning system based on deep learning |
CN116486347A (en) * | 2023-02-21 | 2023-07-25 | 河南省气象探测数据中心(河南省气象档案馆) | Method and device for monitoring and capturing fog based on scale-invariant feature transformation image recognition |
CN116486347B (en) * | 2023-02-21 | 2023-10-10 | 河南省气象探测数据中心(河南省气象档案馆) | Method and device for monitoring and capturing fog based on scale-invariant feature transformation image recognition |
CN117269011A (en) * | 2023-11-13 | 2023-12-22 | 珠海光焱科技有限公司 | Meteorological visibility judging method based on laser light path measuring system |
CN117269011B (en) * | 2023-11-13 | 2024-01-30 | 珠海光焱科技有限公司 | Meteorological visibility judging method based on laser light path measuring system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110188586A (en) | System and application method based on meteorological observation, road camera shooting visibility identification | |
CN109165582B (en) | Urban street garbage detection and cleanliness assessment method | |
CN107742125B (en) | Deep machine learning to predict and prevent adverse conditions at structural assets | |
CN106056695B (en) | A kind of intelligent mobile inspection terminal based on Internet of Things | |
CN110443969B (en) | Fire detection method and device, electronic equipment and storage medium | |
CN109285376A (en) | A kind of bus passenger flow statistical analysis system based on deep learning | |
CN109255288A (en) | A kind of road surface breakage detection method, device and terminal device | |
CN109344753A (en) | A kind of tiny fitting recognition methods of Aerial Images transmission line of electricity based on deep learning | |
CN111033318A (en) | Local weather prediction | |
CN105181898A (en) | Atmospheric pollution monitoring and management method as well as system based on high-density deployment of sensors | |
CN109375290B (en) | Cross-sea bridge fog monitoring system based on machine learning and application method thereof | |
CN110414359A (en) | The analysis of long distance pipeline unmanned plane inspection data and management method and system | |
KR101183105B1 (en) | Method of establishing information of cloud data and establishing system of information of cloud data | |
CN112668375B (en) | Tourist distribution analysis system and method in scenic spot | |
CN110400462B (en) | Track traffic passenger flow monitoring and early warning method and system based on fuzzy theory | |
CN112710798B (en) | Water body blue algae identification system and method | |
CN114240868A (en) | Unmanned aerial vehicle-based inspection analysis system and method | |
CN111339826A (en) | Landslide unmanned aerial vehicle linear sensor network frame detection system | |
CN116205541B (en) | Method and device for evaluating influence of local pollution source on environmental air quality | |
CN117351521B (en) | Digital twinning-based power transmission line bird detection method, system, medium and equipment | |
CN115585731A (en) | Space-air-ground integrated hydropower station space state intelligent monitoring management system and method thereof | |
CN115063442A (en) | Power transmission line hidden danger target tracking method, equipment and medium | |
Won et al. | Particulate matter estimation from public weather data and closed-circuit television images | |
CN117037088A (en) | Positioning method and system for thermal power plant coal ash transport vehicle based on edge calculation | |
CN117197374A (en) | Digital twinning-based intelligent community system |
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 | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20190830 |
|
WW01 | Invention patent application withdrawn after publication |