CN106657026B - Video flowing Intellisense coprocessing system and its processing method based on GPU - Google Patents
Video flowing Intellisense coprocessing system and its processing method based on GPU Download PDFInfo
- Publication number
- CN106657026B CN106657026B CN201611089038.6A CN201611089038A CN106657026B CN 106657026 B CN106657026 B CN 106657026B CN 201611089038 A CN201611089038 A CN 201611089038A CN 106657026 B CN106657026 B CN 106657026B
- Authority
- CN
- China
- Prior art keywords
- node
- video flowing
- video
- gpu
- cloud server
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L65/00—Network arrangements, protocols or services for supporting real-time applications in data packet communication
- H04L65/60—Network streaming of media packets
- H04L65/75—Media network packet handling
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1097—Protocols in which an application is distributed across nodes in the network for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/12—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
Abstract
A kind of video flowing Intellisense coprocessing system and its processing method based on GPU, is related to technical field of video processing, and the solution is to the technical problems more than prior art difference of overall importance and the occupancy communication resource.The system includes the Cloud Server and multiple video flowing sensing nodes interconnected by wide area network；The Cloud Server is configured with cloud storage unit；The video flowing sensing node includes node camera, node main control module, node CPU module, node communication module；Cloud Server is that each video flowing sensing node distributes a video frame process cycle generator to match with the characteristic of the video flowing sensing node, the processing period sequence operative that each video flowing sensing node is generated according to video frame process cycle generator；Cloud Server carries out depth analysis to the pretreated video information that each video flowing sensing node is sent.System and method provided by the invention, of overall importance good, video analysis effect is comprehensive, and it is few to occupy the communication resource.
Description
Technical field
The present invention relates to video processing techniques, handle more particularly to a kind of video flowing Intellisense collaboration based on GPU
The technology of system and its processing method.
Background technique
With technology of Internet of things, cloud computing, the extensive use of big data analysis technology, the real-time analysis of video flowing becomes to get over
Come more important.Video flowing analysis can be applied to safety monitoring, remote device diagnostics, traffic analysis and intelligent automobile without
People's driving system etc..Currently, the video flowing of high definition mainly uses H.265/H.264 coded format, this coded image is divided
Analysis needs to undergo the processes such as decoding, image preprocessing, image analysis, and mathematical model used by these processes needs very high meter
Resource is calculated to support.Thus, video flowing analysis is carried out at present often uses 2 kinds of modes:
1) video is streamed to local monitoring center, is analyzed by monitoring personnel by eyes；This mode has need to
A large amount of manpower is wanted, and important information is missed due to the fatigue of personnel, further, this analytical model lacks letter of overall importance
Breath, thus people need to improve and abandon this method.
2) video flowing is transferred to the analysis of cloud data and storage center, and large-scale cloud platform is needed to support.This side
Method major defect is: video flowing transmission occupies a large amount of Wide Area Network resource, and cost is excessively high, and scale is difficult to scale up；On a large scale
Cloud platform needs to put into a large amount of human and material resources, consumes a large amount of electric energy, thus is difficult sustainable development.
Summary of the invention
For above-mentioned defect existing in the prior art, technical problem to be solved by the invention is to provide a kind of of overall importance
Good, video analysis effect is comprehensive, and occupy the few video flowing Intellisense coprocessing system based on GPU of the communication resource and
Its processing method.
In order to solve the above-mentioned technical problem, at a kind of video flowing Intellisense collaboration based on GPU provided by the present invention
Reason system, it is characterised in that: including Cloud Server and multiple video flowing sensing nodes；
The Cloud Server is configured with cloud storage unit, and video flowing sensing node, Cloud Server access the same wide area network
Network；
The video flowing sensing node includes node camera, node main control module, node CPU module, node communication mould
Block, node CPU module therein and node main control module interconnect, and the video data output port of node camera is connected to node master
The data collection terminal mouth of module is controlled, node main control module accesses the network where Cloud Server by node communication module, so that
Node main control module can carry out data exchange with Cloud Server.
Further, the video flowing sensing node is provided with node cabinet, node main control module, node CPU module, section
Point communication module is installed in node cabinet, equipped with the alignment lens direction for control node camera on node cabinet
Nodal head, the control port of nodal head are connected to the control signal output mouth of node main control module, the node camera
It is mounted on nodal head, equipped with the node heat dissipation ventilation dress for the electric component heat dissipation in node cabinet on node cabinet
It sets, radiator fan, air filter module and Temperature Humidity Sensor, and temperature and humidity is provided in the node radiation ventilator
The inductive signal output port of sensor is connected to the data collection terminal mouth of node main control module.
The processing method of video flowing Intellisense coprocessing system provided by the present invention based on GPU, feature exist
In:
A node serial number id is set for each video flowing sensing node, and sets one for each video flowing sensing node
It include for identifying the work shape of all parts in video flowing sensing node in node state vector SV, node state vector SV
Multiple state elements of state；
After the starting of video flowing sensing node, node main control module sends a packet to Cloud Server by node communication module
Registration information packet containing this node registration information, and wait the feedback information packet of Cloud Server, the note of video flowing sensing node
Volume information includes the node serial number id, node network address ip, node state vector SV of video flowing sensing node；
GPU algorithms library GS is pre-deposited in cloud storage unit；
After Cloud Server receives the registration information packet that video flowing sensing node is sent, first detect whether cloud storage unit stores
There is the registration information of the video flowing sensing node；
It is first the view if Cloud Server if being stored with the registration information of the video flowing sensing node in cloud storage unit
Frequency stream sensing node distributes an a video frame process cycle generator CG and GPU algorithm subset GPUS, then is the video flowing
Sensing node constructs a feedback information packet and is handed down to the video flowing sensing node；
If the not stored registration information for having the video flowing sensing node in cloud storage unit, first from reception if Cloud Server
To registration information packet in obtain the registration information of the video flowing sensing node, and acquired registration information is stored in cloud storage
Unit, and an a video frame process cycle generator CG and GPU algorithm subset GPUS is distributed for the video flowing sensing node,
Then a feedback information packet is constructed again and is handed down to the video flowing sensing node；
Cloud Server includes the node of video flowing sensing node in the feedback information packet for the building of video flowing sensing node
The video frame process cycle generator that number id, node network address ip and Cloud Server are distributed by video flowing sensing node
The GPU algorithm subset GPUS that the process cycle T and Cloud Server of CG is distributed by video flowing sensing node；
Video frame process cycle generator CG is the period sequencer for generating processing period sequence, and
In video frame process cycle generator CG processing period sequence generated, the duration of each processing period is equal, video
The process cycle T of frame process cycle generator CG refers to the duration of single processing period；
GPU algorithm subset GPUS is the subset of GPU algorithms library GS, is extracted from cloud storage unit by Cloud Server
The set of GPU algorithm；
In video flowing sensing node, node main control module obtains the feedback letter from Cloud Server by node communication module
Breath packet, and according to process cycle T, the GPU algorithm subset GPUS for the video frame process cycle generator CG for distributing to this node come
The workflow for constructing node CPU module, and is sent to node CPU module, so that node CPU module is according to distributing to this node
Video frame process cycle generator CG process cycle T, GPU algorithm subset GPUS work；
In video flowing sensing node, node main control module is according to the video frame process cycle generator CG for distributing to this node
Process cycle T from node camera obtain current video frame frameImg, and from distribute to this node video frame handle week
Phase generator CG obtains current processing time section t, constructs video according to current processing time section t and current video frame frameImg
Frame GPU process object VFPO is sent to node CPU module；
Video frame GPU deal with objects VFPO by current processing time section t, video flowing VST, current video frame frameImg,
GPUAS is constituted, and current processing time section t therein is generated by the video frame process cycle generator CG for distributing to this node,
Video flowing VST is obtained from node camera, and current video frame frameImg is to be dealt with current in current processing time section t
Frame, GPUAS are according to distributing to set by the GPU algorithm subset GPUS of this node for handling the set of algorithms of video flowing VST
It closes；
In video flowing sensing node, node CPU module deals with objects the GPUAS in VFPO to current according to video frame GPU
Video frame frameImg is handled, and constructing one after being disposed includes current processing time section t, video flowing VST
And the video frame processing result object VFPRO of processing result Rels, then video frame processing result object VFPRO is sent to again
Node main control module, it includes video frame processing result object VFPRO and the section of this node that node main control module constructs one again
The processing result message package of point number id, node network address ip, and Cloud Server is sent to by node communication module；
After Cloud Server receives the processing result message package that video flowing sensing node is sent, according in processing result message package
Node serial number id, node network address ip find out the video frame process cycle generator for distributing to the video flowing sensing node
CG, and a new current processing time section tt is obtained from video frame process cycle generator CG, then from processing result message
Video flowing VST and processing result Rels is extracted in video frame processing result object VFPRO in packet, according still further to new current place
Time period t t is managed, depth analysis is carried out to extracted video flowing VST, processing result Rels using GPU algorithms library GS, and will be deep
Degree analysis result is stored in cloud storage unit.
Further, after the starting of video flowing sensing node, node main control module detects this section by node state vector SV
The working condition of each component in point, node main control module and node CPU module, node communication module communication for coordination, mutually report work
Make state.
Further, after video flowing sensing node VSN starting, node main control module is detected by node state vector SV
There is when trouble unit in this node, building one includes that the warning message packet of the node state vector SV of this node is concurrent
Give Cloud Server.
Video flowing Intellisense coprocessing system and its processing method provided by the invention based on GPU, using GPU mould
Block constructs video flowing sensing node, and video flowing sensing node carries out Preprocessing to video using the powerful processing capacity of GPU,
Then pretreated information is transferred to Cloud Server, global video information analysis is then carried out by Cloud Server；The present invention
It has the advantages that
1) utilize GPU(graphics processor) building video flowing sensing node have powerful processing capacity, can be to figure
As executing Gaussian transformation, optical flow analysis, Kalman's operation scheduling algorithm to image preprocessing, it might even be possible to execute the group of these algorithms
It closes, therefore the information content of pretreated result greatly reduces；
2) burden of communication system is greatly reduced, communication system is often based upon wireless network (such as 3G), due to pretreatment
Information content afterwards greatly reduces, therefore the communication occupancy volume of 3G network reduces, therefore saves communication cost；And due to communication
The reduction of amount considerably increases the number of the video flowing access of cloud processing center, improves the treatment scale of system, thus expands
It is of overall importance, make video analysis effect more comprehensively；
3) each video flowing sensing node has oneself independent video frame process cycle generator, and Cloud Server can be with
Support the Cooperative Analysis of the video flowing sensing node of thousands of different process cycles, thus have greater flexibility and
Scalability；
4) GPU Processing Algorithm can be formulated according to the GPU processing capacity of each video flowing sensing node, there is isomery meter
The good support of calculation ability；
5) GPU processing unit has the characteristic of low-cost and high-performance, can promote the widely available of video flowing sensing node,
And then promote the development in an all-round way based on video Internet of Things industry；A such as video influenza with 2000 stream processing units
Know that the cost of node does not exceed 4000 yuans, as the development of GPU technology is universal, this cost will be will be greatly reduced.
Detailed description of the invention
Fig. 1 is the structural schematic diagram of the video flowing Intellisense coprocessing system based on GPU of the embodiment of the present invention；
Fig. 2 is the video flowing perception in the video flowing Intellisense coprocessing system based on GPU of the embodiment of the present invention
The structural schematic diagram of node.
Specific embodiment
The embodiment of the present invention is described in further detail below in conjunction with Detailed description of the invention, but the present embodiment is not used to limit
The system present invention, all that protection scope of the present invention should all be included in using similar structure and its similar variation of the invention, present invention
In pause mark indicate the relationship of sum, the English alphabet in the present invention is case sensitive.
As Figure 1-Figure 2, a kind of video flowing Intellisense collaboration processing based on GPU provided by the embodiment of the present invention
System, it is characterised in that: including Cloud Server CLOUD and multiple video flowing sensing node VSN；
The Cloud Server CLOUD is configured with cloud storage unit CDB, video flowing sensing node VSN, Cloud Server CLOUD
Access the same Wide Area Network；Cloud Server CLOUD is cloud access center, and Cloud Server CLOUD is responsible for saving with video flowing perception
Point VSN interaction, and each video flowing sensing node VSN is managed, each video flowing sensing node VSN is sent pretreated
Video information carry out depth analysis, cloud storage unit CDB is for storing video data information and each video flowing sensing node
The management information of VSN；
The video flowing sensing node VSN include node camera ca, node main control module mc, node CPU module gpum,
Node communication module cm, node CPU module gpum therein and node main control module mc interconnection, the video counts of node camera ca
The data collection terminal mouth of node main control module mc is connected to according to output port, node main control module mc is connect by node communication module cm
Enter the network where Cloud Server CLOUD, node main control module mc and Cloud Server CLOUD is enabled to carry out data exchange；Section
Point main control module mc, node CPU module gpum apply same model (such as NVIDIA CUDA) cooperated computing, complete to view
The analysis (such as Gaussian transformation, optical flow analysis, Kalman's operation etc.) of frequency image；
The video flowing sensing node VSN is provided with node cabinet, node main control module mc, node CPU module gpum, section
Point communication module cm is installed in node cabinet, equipped with the alignment lens side for control node camera ca on node cabinet
To nodal head pla, the control port of nodal head pla is connected to the control signal output mouth of node main control module mc, institute
It states node camera ca to be mounted on nodal head pla, equipped with node erecting by overhang sm and for node cabinet on node cabinet
The node radiation ventilator VS of interior electric component heat dissipation, node cabinet are mounted on landing object by node erecting by overhang sm
On, the bottom of node cabinet offers air inlet, and the bottom air inlet of node cabinet is provided with air filter module, institute
It states and is provided with radiator fan, air filter module and Temperature Humidity Sensor in node radiation ventilator VS, and temperature and humidity passes
The inductive signal output port of sensor is connected to the data collection terminal mouth of node main control module mc.
The processing method of video flowing Intellisense coprocessing system based on GPU provided by the embodiment of the present invention,
It is characterized in that:
A node serial number id is set for each video flowing sensing node VSN, and is set for each video flowing sensing node VSN
Include and all parts one in video flowing sensing node VSN in a fixed node state vector SV, node state vector SV
One corresponding multiple state elements, and there are two types of state values for each state elements in node state vector SV, it is therein
The working condition that a kind of state value is used to identify component corresponding to the state elements is normal operating conditions, another state value
Working condition for identifying component corresponding to the state elements is malfunction；
After video flowing sensing node VSN starting, node main control module mc passes through node communication module cm to Cloud Server
CLOUD sends one and includes the registration information packet of this node registration information, and waits the feedback information of Cloud Server CLOUD
Packet, the registration information of video flowing sensing node VSN includes node serial number id, the node network address of video flowing sensing node VSN
Ip, node state vector SV；
It is the prior art that GPU algorithms library GS(GPU algorithms library GS is pre-deposited in cloud storage unit CDB)；
After Cloud Server CLOUD receives the registration information packet that video flowing sensing node VSN is sent, cloud storage unit is first detected
Whether CDB is stored with the registration information of video flowing sensing node VSN；
If being stored with the registration information of video flowing sensing node VSN, Cloud Server in cloud storage unit CDB
CLOUD then first distributes a video frame process cycle generator CG and a GPU algorithm subset for the video flowing sensing node VSN
GPUS, then construct a feedback information packet for the video flowing sensing node VSN and be handed down to video flowing sensing node VSN；
If the not stored registration information for having video flowing sensing node VSN, Cloud Server in cloud storage unit CDB
CLOUD then first obtains the registration information of video flowing sensing node VSN from the registration information packet received, and will be acquired
Registration information is stored in cloud storage unit CDB, and distributes a video frame process cycle generator for the video flowing sensing node VSN
Then CG and GPU algorithm subset GPUS constructs a feedback information packet again and is handed down to video flowing sensing node VSN；
Cloud Server CLOUD includes video flowing perception section in the feedback information packet for video flowing sensing node VSN building
The view that node serial number id, node network address ip and the Cloud Server CLOUD of point VSN is distributed by video flowing sensing node VSN
The GPU that the process cycle T and Cloud Server CLOUD of frequency frame process cycle generator CG is distributed by video flowing sensing node VSN
Algorithm subset GPUS；
Video frame process cycle generator CG is the period sequencer for generating processing period sequence, and
In video frame process cycle generator CG processing period sequence generated, the duration of each processing period is equal, video
The process cycle T of frame process cycle generator CG refers to the duration of single processing period；
Cloud Server CLOUD is by each video flowing sensing node VSN video frame process cycle generator CG's distributed
Process cycle T and GPU algorithm subset GPUS is to be formulated according to the video processing requirement of each video flowing sensing node VSN, respectively
The video processing requirement of a video flowing sensing node VSN can be set manually, can also according to the rule settings customized in advance,
Process cycle T and GPU the algorithm subset for the video frame process cycle generator CG that different video flowing sensing node VSN is distributed
GPUS can be it is identical, be also possible to it is different, such as distribute to first video flowing sensing node VSN video frame processing
The value of the process cycle T of Cycle generator CG is T1, distributes to the video frame process cycle of second video flowing sensing node VSN
The value of the process cycle T of generator CG is T2, and the video frame process cycle for distributing to m-th of video flowing sensing node VSN occurs
The value of the process cycle T of device CG is Tm, then the value of T1, T2, Tm can be identical, be also possible to different；
GPU algorithm subset GPUS is the subset of GPU algorithms library GS, is by Cloud Server CLOUD from cloud storage unit CDB institute
The set of the GPU algorithm of extraction；
In video flowing sensing node VSN, node main control module mc is obtained by node communication module cm and is come from Cloud Server
The feedback information packet of CLOUD, and calculated according to process cycle T, GPU for the video frame process cycle generator CG for distributing to this node
Method collection GPUS come construct node CPU module gpum workflow (building node CPU module gpum workflow method be
The prior art), and it is sent to node CPU module gpum, so that node CPU module gpum is according to the video frame for distributing to this node
Process cycle T, the GPU algorithm subset GPUS of process cycle generator CG works；
Cloud Server CLOUD issues holder regulating command, video flowing perception section to video flowing sensing node VSN by network
In point VSN, node main control module mc driving node camera ca work, and adjusted according to the holder that Cloud Server CLOUD is issued
Instruction carrys out control node holder pla operation, so that the holder regulating command that node camera ca is issued according to Cloud Server CLOUD
It locks video and captures window, and is periodical according to the process cycle T for the video frame process cycle generator CG for distributing to this node
Generation handle period sequence so that node camera ca according to it is generated processing period sequence periodicity capture number
According to frame, start video acquisition procedure；
In video flowing sensing node VSN, node main control module mc is sent out according to the video frame process cycle for distributing to this node
The process cycle T of raw device CG obtains current video frame frameImg from node camera ca, and from the video for distributing to this node
Frame process cycle generator CG obtains current processing time section t, according to current processing time section t and current video frame frameImg
Building video frame GPU process object VFPO is sent to node CPU module gpum；
Video frame GPU deal with objects VFPO by current processing time section t, video flowing VST, current video frame frameImg,
GPUAS is constituted, and current processing time section t therein is generated by the video frame process cycle generator CG for distributing to this node,
Video flowing VST is obtained from node camera ca, and current video frame frameImg is to be dealt in current processing time section t works as
Previous frame, GPUAS are according to distributing to set by the GPU algorithm subset GPUS of this node for handling the algorithm of video flowing VST
Set；
In video flowing sensing node VSN, node CPU module gpum deals with objects the GPUAS in VFPO according to video frame GPU
Current video frame frameImg is handled, and constructing one after being disposed includes current processing time section t, video
The video frame processing result object VFPRO for flowing VST and processing result Rels, then again by video frame processing result object VFPRO
It is sent to node main control module mc, it includes video frame processing result object VFPRO that node main control module mc constructs one again, and
The processing result message package of the node serial number id of this node, node network address ip, and be sent to by node communication module cm
Cloud Server CLOUD；
After Cloud Server CLOUD receives the processing result message package that video flowing sensing node VSN is sent, according to processing result
Node serial number id, node network address ip in message package find out the video frame processing for distributing to video flowing sensing node VSN
Cycle generator CG, and a new current processing time section tt is obtained from video frame process cycle generator CG, then from
Video flowing VST and processing result Rels is extracted in video frame processing result object VFPRO in reason results message, according still further to
New current processing time section tt carries out depth to extracted video flowing VST, processing result Rels using GPU algorithms library GS
Analysis, and depth analysis result is stored in cloud storage unit CDB.
In the embodiment of the present invention, after video flowing sensing node VSN starting, node main control module mc passes through node state vector
SV detects the working condition of each component in this node, node main control module mc and node CPU module gpum, node communication module
Working condition, and the temperature and humidity sensing in node main control module mc and node radiation ventilator VS are mutually reported in cm communication for coordination
Device communication for coordination, mutually reports working condition.
In the embodiment of the present invention, after video flowing sensing node VSN starting, node main control module mc passes through node state vector
SV detects that building one includes the alarm signal of the node state vector SV of this node there is when trouble unit in this node
Breath wraps and is sent to Cloud Server CLOUD.
In the embodiment of the present invention, node cabinet is equipped with each cooling vent in front and back, and the cooling vent respectively fills
There is a node radiation ventilator VS；
When video flowing sensing node VSN is run, node main control module mc passes through in two node radiation ventilator VS
The humidity of two cooling vents of Temperature Humidity Sensor detection node cabinet is previously provided with one in node main control module mc
Humidity threshold；
If the humidity of two cooling vents is both greater than pre-set humidity threshold, node main control module mc is controlled
Radiator fan in front nodal point radiation ventilator VS is dried forward, and controls the radiation air in posterior nodal point radiation ventilator VS
Fan is dried backward；
If the humidity of two cooling vents is both less than equal to pre-set humidity threshold, and preceding cooling vent
Humidity be greater than the humidity of rear cooling vent, then control the radiator fan in two node radiation ventilator VS to fore blow
Wind；
If the humidity of two cooling vents is both less than equal to pre-set humidity threshold, and preceding cooling vent
Humidity be less than the humidity of rear cooling vent, then control the radiator fan in two node radiation ventilator VS to after-blow
Wind；
By the control to the radiator fan in two node radiation ventilator VS, realize to warm and humid in node cabinet
The adjusting of degree.
The embodiment of the present invention has carried out case verification, and experimental situation is as follows:
Video flowing sensing node VSN, node main control module mc use Intel (Intel) Intel Core i3, node CPU module
Gpum uses 300W PC power source using inscription speed GTX670M 2G D5 independent display card, power supply；
Only consider that video flowing sensing node VSN completes the execution of one group of GPU algorithm in some process cycle T, utilizes road
Needed for the video analysis of the crowded monitoring scene in road, Kalman Algorithm, light are completed in a process cycle T to a video frame
Flow algorithm, Gauss algorithm, the time of three kinds of algorithm process 1080P, the mono- frame image are respectively as follows: Kalman Algorithm and have used 0.07 second,
Optical flow algorithm has been used 0.09 second, and Gauss algorithm has used 0.05；
It can be seen that when a length of 0.07+0.09+0.05=0.21 second of process cycle T, it can be seen that the embodiment of the present invention can
To complete the pretreatment of the Kalman Algorithm, optical flow algorithm, Gauss algorithm of image in 1 second, treated, and image is artwork master
Picture is stored by compression, greatly reduces traffic.
Claims (4)
1. a kind of processing method of the video flowing Intellisense coprocessing system based on GPU, it is characterised in that: the system includes
Cloud Server and multiple video flowing sensing nodes；
The Cloud Server is configured with cloud storage unit, and video flowing sensing node, Cloud Server access the same Wide Area Network；
The video flowing sensing node includes node camera, node main control module, node CPU module, node communication module,
In node CPU module and node main control module interconnect, the video data output port of node camera is connected to node master control mould
The data collection terminal mouth of block, node main control module accesses the network where Cloud Server by node communication module, so that node
Main control module can carry out data exchange with Cloud Server,
The processing method of the system are as follows:
A node serial number id is set for each video flowing sensing node, and sets a node for each video flowing sensing node
It include for identifying all parts working condition in video flowing sensing node in state vector SV, node state vector SV
Multiple state elements；
After the starting of video flowing sensing node, node main control module sends one to Cloud Server by node communication module and includes
The registration information packet of this node registration information, and the feedback information packet of Cloud Server is waited, the registration letter of video flowing sensing node
Breath includes node serial number id, node network address ip, the node state vector SV of video flowing sensing node；
GPU algorithms library GS is pre-deposited in cloud storage unit；
After Cloud Server receives the registration information packet that video flowing sensing node is sent, first detect whether cloud storage unit is stored with this
The registration information of video flowing sensing node；
It is first the video flowing if Cloud Server if being stored with the registration information of the video flowing sensing node in cloud storage unit
Sensing node distributes an a video frame process cycle generator CG and GPU algorithm subset GPUS, then is video flowing perception
Node constructs a feedback information packet and is handed down to the video flowing sensing node；
If the not stored registration information for having the video flowing sensing node in cloud storage unit, first from receiving if Cloud Server
The registration information of the video flowing sensing node is obtained in registration information packet, and acquired registration information is stored in cloud storage list
Member, and an a video frame process cycle generator CG and GPU algorithm subset GPUS is distributed for the video flowing sensing node, so
It constructs a feedback information packet again afterwards and is handed down to the video flowing sensing node；
Cloud Server includes the node serial number of video flowing sensing node in the feedback information packet for the building of video flowing sensing node
The video frame process cycle generator CG's that id, node network address ip and Cloud Server are distributed by video flowing sensing node
The GPU algorithm subset GPUS that process cycle T and Cloud Server are distributed by video flowing sensing node；
Video frame process cycle generator CG is for generating the period sequencer for handling period sequence, and video
In frame process cycle generator CG processing period sequence generated, the duration of each processing period is equal, at video frame
The process cycle T of reason Cycle generator CG refers to the duration of single processing period；
GPU algorithm subset GPUS is the subset of GPU algorithms library GS, is to be calculated by Cloud Server from the extracted GPU of cloud storage unit
The set of method；
In video flowing sensing node, node main control module obtains the feedback information from Cloud Server by node communication module
Packet, and according to process cycle T, the GPU algorithm subset GPUS for the video frame process cycle generator CG for distributing to this node come structure
The workflow of node CPU module is built, and is sent to node CPU module, so that node CPU module is according to distributing to this node
Process cycle T, the GPU algorithm subset GPUS of video frame process cycle generator CG works；
In video flowing sensing node, node main control module according to the video frame process cycle generator CG for distributing to this node place
It manages cycle T and obtains current video frame frameImg from node camera, and sent out from the video frame process cycle for distributing to this node
Raw device CG obtains current processing time section t, constructs video frame according to current processing time section t and current video frame frameImg
GPU process object VFPO is sent to node CPU module；
Video frame GPU deals with objects VFPO by current processing time section t, video flowing VST, current video frame frameImg, GPUAS
It constitutes, current processing time section t therein is generated by the video frame process cycle generator CG for distributing to this node, video flowing
VST is obtained from node camera, and current video frame frameImg is present frame to be dealt in current processing time section t,
GPUAS is according to distributing to set by the GPU algorithm subset GPUS of this node for handling the algorithm set of video flowing VST；
In video flowing sensing node, node CPU module deals with objects the GPUAS in VFPO to current video according to video frame GPU
Frame frameImg is handled, and constructing one after being disposed includes current processing time section t, video flowing VST and place
The video frame processing result object VFPRO for managing result Rels, is then sent to node for video frame processing result object VFPRO again
Main control module, it includes video frame processing result object VFPRO and the node volume of this node that node main control module constructs one again
The processing result message package of number id, node network address ip, and Cloud Server is sent to by node communication module；
After Cloud Server receives the processing result message package that video flowing sensing node is sent, according to the section in processing result message package
Point number id, node network address ip find out the video frame process cycle generator CG for distributing to the video flowing sensing node, and
A new current processing time section tt is obtained from video frame process cycle generator CG, then from processing result message package
Video flowing VST and processing result Rels is extracted in video frame processing result object VFPRO, according still further to new current processing time
Section tt carries out depth analysis to extracted video flowing VST, processing result Rels using GPU algorithms library GS, and by depth analysis
As a result it is stored in cloud storage unit.
2. the processing method of the video flowing Intellisense coprocessing system according to claim 1 based on GPU, feature
Be: the video flowing sensing node is provided with node cabinet, and node main control module, node CPU module, node communication module are equal
It is mounted in node cabinet, the nodal head equipped with the alignment lens direction for control node camera on node cabinet, section
The control port of point holder is connected to the control signal output mouth of node main control module, and the node camera is mounted on node cloud
On platform, equipped with the node radiation ventilator for the electric component heat dissipation in node cabinet on node cabinet, the node dissipates
Radiator fan, air filter module and Temperature Humidity Sensor, and the induction of Temperature Humidity Sensor are provided in hot ventilation device
Signal output port is connected to the data collection terminal mouth of node main control module.
3. the processing method of the video flowing Intellisense coprocessing system according to claim 1 based on GPU, feature
Be: after the starting of video flowing sensing node, node main control module detects each component in this node by node state vector SV
Working condition, node main control module and node CPU module, node communication module communication for coordination mutually report working condition.
4. the processing method of the video flowing Intellisense coprocessing system according to claim 3 based on GPU, feature
Be: after video flowing sensing node VSN starting, node main control module detects in this node exist by node state vector SV
Trouble unit when, building one include this node node state vector SV warning message packet and be sent to cloud service
Device.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611089038.6A CN106657026B (en) | 2016-12-01 | 2016-12-01 | Video flowing Intellisense coprocessing system and its processing method based on GPU |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611089038.6A CN106657026B (en) | 2016-12-01 | 2016-12-01 | Video flowing Intellisense coprocessing system and its processing method based on GPU |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106657026A CN106657026A (en) | 2017-05-10 |
| CN106657026B true CN106657026B (en) | 2019-08-30 |
Family
ID=58813902
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611089038.6A Active CN106657026B (en) | 2016-12-01 | 2016-12-01 | Video flowing Intellisense coprocessing system and its processing method based on GPU |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106657026B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108922172B (en) * | 2018-06-19 | 2021-03-05 | 上海理工大学 | Road congestion monitoring system based on vehicle characteristic matrix sequence change analysis |
| CN111045606B (en) * | 2019-12-13 | 2021-08-27 | 西安奥卡云数据科技有限公司 | Extensible cloud scale IOT storage method and device and server |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9191625B2 (en) * | 2008-09-26 | 2015-11-17 | Janos Redei | System and methods for transmitting and distributing media content |
| CN103905833B (en) * | 2013-07-12 | 2017-06-27 | 上海易钧信息技术有限公司 | Based on cloud computing distributed network video data digging acquisition system |
| CN106066825B (en) * | 2016-07-01 | 2018-10-12 | 上海理工大学 | GPU cluster power consumption monitoring system and its monitoring method |
-
2016
- 2016-12-01 CN CN201611089038.6A patent/CN106657026B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN106657026A (en) | 2017-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107067365A (en) | The embedded real-time video stream processing system of distribution and method based on deep learning | |
| CN104167818B (en) | Intelligent remote inspection system in linkage with GIS substation integrated automation system and method | |
| CN109068145A (en) | Distribution type intelligent video analysis system, method, apparatus, equipment and storage medium | |
| CN106657026B (en) | Video flowing Intellisense coprocessing system and its processing method based on GPU | |
| CN102638456A (en) | Intelligent analysis method for mass real-time video code stream based on cloud computation and system thereof | |
| Wang et al. | A scalable distributed architecture for intelligent vision system | |
| CN106296788B (en) | Across the computer room Cluster Rendering of one kind disposes realization system | |
| CN111159093B (en) | Heterogeneous intelligent computing system | |
| Sun et al. | mVideo: Edge computing based mobile video processing systems | |
| CN202488597U (en) | Autonomous cooperation video monitor node apparatus for intelligent object identification and track | |
| Mora et al. | Distributed architectures for intensive urban computing: a case study on smart lighting for sustainable cities | |
| CN111652058B (en) | Computer face recognition device | |
| Acharya et al. | Workload-aware task placement in edge-assisted human re-identification | |
| CN108288025A (en) | A kind of car video monitoring method, device and equipment | |
| CN106155813A (en) | Managing computing devices in a computing system | |
| CN103106638A (en) | Graphic processing device for real-time image processing | |
| DE102022114517A1 (en) | HYBRID COOLING SYSTEMS FOR DATA CENTERS | |
| Ang et al. | Embedded intelligence: Platform technologies, device analytics, and smart city applications | |
| Imran et al. | Implementation of wireless vision sensor node for characterization of particles in fluids | |
| CN207624035U (en) | A kind of construction site supervisory systems based on recognition of face | |
| CN117009997A (en) | Informationized processing device based on internet of things | |
| CN112528265A (en) | Identity recognition method, device, equipment and medium based on online conference | |
| CN206364590U (en) | A kind of electric power remote monitoring and control system | |
| CN104994343A (en) | Intelligent video analysis system | |
| CN111487912B (en) | Central intelligent control system in venue |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |