A kind of panoramic video traffic situation supervisory control system and methodTechnical field
The present invention relates to a kind of traffic situation supervisory control system and method, particularly based on traffic surveillance and control system and the method for panoramic video realization, belong to technical field of intelligent traffic.
Background technology
Constantly perfect along with urban transportation facility, vehicles quantity increases day by day, and traffic pressure sharply increases, and accident rate significantly rises, and intelligent traffic monitoring is imperative, and the construction of intelligent transportation system (ITS) is imperative.Intelligent transportation system (ITS) understands the ruuning situation of road traffic by the telecommunication flow information detected on road, according to the dynamic change of traffic flow, make rapidly traffic guidance to control, alleviate the road degree of crowding, reduce vehicle driving to incur loss through delay, reduce the probability that traffic accident occurs, ensure traffic safety, means of transportation are fully used.
At present, traffic information collection device category is a lot, as ring coil detector, geomagnetism detecting device, electromagnetic detector, microwave detector, supersonic detector, video detection technology etc.But along with the development of video image processing technology, traffic information collection technology based on image shows the unexistent advantage of other collecting devices, to hold an important position in following real-time traffic information collection and treatment technology, video encoder server technology constantly will be developed and be applied more widely.
Based in the magnitude of traffic flow supervisory control system of video detection technology and analytical method, usually setting up enough video cameras along traffic route, by carrying out intellectual analysis to the video of these video cameras, drawing the traffic behavior in this region.
In actual use, there is following shortcoming in above system and method:
1, camera is arranged on road both sides by discrete, and the region that each video pictures covers is limited, causes the magnitude of traffic flow of single-lens analysis to have larger locality, monitoring blind area ubiquity;
Although 2, can effectively make up above-mentioned defect by increasing camera quantity, can bring being multiplied of headend equipment, background analysis memory device, cost sharp rises.
Summary of the invention
In order to overcome the above-mentioned defect under prior art, the object of the present invention is to provide a kind of panoramic video traffic situation supervisory control system and method, this system and method can realize when limited hardware drops into efficiently, the traffic video monitoring of non-blind area.
Technical scheme of the present invention is:
A kind of panoramic video traffic situation method for supervising, implementation step comprises:
A. live video to be monitored is gathered by one or more groups video capture device;
B. gathered live video is spliced into panoramic video;
C. carry out the real-time analysis of flow speeds and vehicle density for described panoramic video, thus obtain traffic state data;
D. to described panoramic video, described traffic state data and comprise scene to be monitored GIS map by information registration and fusion treatment, form GIS live-action map;
E. described GIS live-action map is pushed on monitoring display equipment.
The splicing of described step b preferably includes offline parameter and calculates and real-time splicing two processes.
Described offline parameter computational process mainly calculated for subsequent splices the splicing parameter needing to use, the process calculating splicing parameter can be: carry out characteristic matching to image sequence to be spliced and calculate homography matrix between two between image, the mapping matrix of single image to panoramic video is obtained by LM algorithm optimization, and calculate the mask figure of single image thus, derive this mapping matrix and mask figure as described splicing parameter.
In described real-time splicing process, preferably create the GPU thread that transfer of data, image mapped and color training managing three are parallel, transfer of data thread is used for several monoscopic video (i.e. the aforesaid live video to be monitored often organized video capture device and gather) to upload to GPU video memory and the panoramic video after splicing is downloaded to CPU internal memory from GPU video memory; Image mapped thread is used for pixel and tables look-up mapping, colour correction, Fusion Edges and cutting process; Color training managing thread is used for color correction parameter training and upgrades.
The real-time analysis process of described step c can be: detect moving target and vehicle for the region to be analyzed preset in described panoramic video by background modeling, calculate quantity and/or the density of moving target, when the quantity of moving target and/or density are greater than respective threshold, carry out motion target tracking, calculate wagon flow direction and wagon flow average speed, integrated car stream average speed and vehicle density draw traffic state data and export, and traffic behavior I is defined as flow speedsand vehicle densityfunction:
，
Wherein x, y, z represents the position in region to be analyzed respectively, and α, β are respectively flow speeds factor of influence and vehicle density factor of influence, and Thd1 is that vehicle density is preset.
The detailed process of described steps d can be: first in panoramic video, mark road, set up the corresponding relation between flow analysis region and road section, be plotted in panoramic video according to this corresponding relation by traffic state data with the lines of different colours; Then, adopt interactive video calibration technique to choose two, three-dimensional point pair, calculate the projection matrix M of described panoramic video to described GIS map, realize the registration of described panoramic video and described GIS map, set up the three-dimensional coordinate mapping table of road area image coordinate and corresponding road area in described GIS map in described panoramic video; Magnitude of traffic flow segmentation statistical form is set up afterwards in GIS map, according to three-dimensional coordinate mapping table, the magnitude of traffic flow segmentation statistics in described panoramic video is mapped in described magnitude of traffic flow segmentation statistical form, and draws flow temperature figure according to the statistics in described magnitude of traffic flow segmentation statistical form; Finally respectively described flow temperature figure and described road panoramic video and described GIS map are merged, form GIS temperature figure and GIS realistic picture, wherein said GIS temperature figure comprises described GIS map, described flow temperature figure and described magnitude of traffic flow segmentation statistics, and described GIS realistic picture comprises described GIS map and described road panoramic video.
Preferably, described GIS map shows all the time, described flow temperature figure and described magnitude of traffic flow segmentation statistics are opened as required or close, and described road panoramic video is presented in described GIS map with the form of texture as required, or in described GIS map, play window display.
A kind of panoramic video traffic situation supervisory control system, comprise the video capture device connected successively by information flow direction, GPU concatenation module, traffic flow analysis module and information registration and fusion treatment module, described information registration is connected with the adaptable interface two-way communication of monitoring display equipment with fusion treatment module, described information registration and fusion treatment module are also provided with panoramic video input and GIS map information input terminal, the output of described GPU concatenation module is also connected with described panoramic video input, described video capture device is for gathering live video to be monitored, live video to be monitored is used as the monoscopic video on the basis forming panoramic video, described GPU concatenation module is used for the decoding splicing of described monoscopic video and colour correction process thus forms panoramic video, described traffic flow analysis module is used for moving object detection and traffic behavior assessment in described panoramic video, described information registration and fusion treatment module be used for described panoramic video, described traffic state data and comprise scene to be monitored GIS map by information registration and fusion treatment, form GIS live-action map, described adaptable interface is used for the identification of monitoring display device type and GIS live-action map pushes.
Described panoramic video traffic situation supervisory control system can also be provided with GIS map data memory module, and the output of described GIS map data memory module connects described GIS map information input terminal, and described GIS map data memory module is used for stored GIS map datum.
Described GPU concatenation module is preferably multiple parallel connection, and each described GPU concatenation module correspondence connects a video capture device group, and each described video capture device group comprises one or more described video capture device.
Beneficial effect of the present invention is:
The blank that current panoramic video technology is applied in intelligent traffic monitoring field has been filled up in the traffic situation monitoring that the present invention is based on panoramic video realization, achieve non-blind area traffic monitoring, the Meteorological of hardware device is controlled in limited range, for significantly improving monitoring effect, operating efficiency is monitored in raising, effective guarantee traffic safety provides efficient solution simultaneously.
Being formed in panoramic video process, adopt offline parameter to calculate and have the real-time splicing of three parallel GPU threads, ensure that the real-time of panorama Graphics Processing well, is the guarantee that provides the foundation based on the traffic situation monitoring of panoramic video.
The present invention adopts specific state machine when traffic flow analysis, makes the result of traffic state analysis more effective, credible.
The multi-source information such as road traffic panoramic video, traffic flow data, magnitude of traffic flow temperature figure and GIS map converges by the present invention, registration and fusion, display, provide the monitored results data with very abundant information amount, a greater variety of magnitude of traffic flow monitoring demand can be met.
Accompanying drawing explanation
Fig. 1 is that GPU walks abreast splicing flow process;
Fig. 2 is panoramic video magnitude of traffic flow detection procedure;
Fig. 3 is panoramic video traffic flow analysis state transition diagram;
Fig. 4 is converging information handling process;
Fig. 5 is panoramic video traffic situation supervisory control system theory of constitution block diagram.
Embodiment
See Fig. 1-4, the invention provides a kind of panoramic video traffic situation method for supervising, utilize one or more groups video capture device (as video camera) the collection site video being arranged on higher position, panorama traffic video is formed by efficient post-processing technology, and on this panoramic video, carry out magnitude of traffic flow detection and body posture potential analysis, form road network real-time traffic situation map, finally panoramic video, traffic behavior and GIS map are carried out Multi-source Information Fusion, by being pushed on long-range PC or mobile device after the conversion of equipment adaptable interface.
The performing step of the method comprises:
A. live video to be monitored is gathered by one or more groups video capture device;
B. gathered live video is spliced into panoramic video;
C. carry out the real-time analysis of flow speeds and vehicle density for described panoramic video, thus obtain traffic state data;
D. to described panoramic video, described traffic state data and comprise scene to be monitored GIS map by information registration and fusion treatment, form GIS live-action map;
E. described GIS live-action map is pushed on monitoring display equipment.
The splicing of described step b comprises offline parameter and calculates and real-time splicing two processes, and flow process is see Fig. 1.
Described offline parameter computational process mainly calculated for subsequent splices the splicing parameter needing to use, the process calculating splicing parameter is: carry out characteristic matching to image sequence to be spliced and calculate homography matrix between two between image, the mapping matrix of single image to panoramic video is obtained by LM algorithm optimization, and calculate the mask figure of single image thus, derive this mapping matrix and mask figure as described splicing parameter.Because video camera attitude substantially remains unchanged after Installation and Debugging complete, therefore video pre-filtering adopts the mode of calculated off-line, result of calculation is saved as subsequent treatment parameter, realizes once calculating, repeatedly uses, improve treatment effeciency.
In described real-time splicing process, create the GPU thread that transfer of data, image mapped and color training managing three are parallel, each monoscopic video in scene to be monitored that transfer of data thread is used for described video capture device to gather uploads to GPU video memory and the panoramic video after splicing is downloaded to CPU internal memory from GPU video memory; Image mapped thread is used for pixel and tables look-up mapping, colour correction, Fusion Edges and cutting process, realizes the normalization of video color; Color training managing thread is used for color correction parameter training and upgrades.Map and the video fusion stage at video, the strategy that processing procedure adopts GPU whole process to accelerate, greatly improve the frame processing time.
Adopt offline parameter to calculate and the omnidistance real-time splicing mode accelerated of GPU, ensure that the real-time of the panorama Graphics Processing process of step b well.
See Fig. 2, the real-time analysis process of described step c is: detect moving target and vehicle for the region to be analyzed preset in described panoramic video by background modeling, calculate quantity and/or the density of moving target, when the quantity of moving target and/or density are greater than respective threshold, feed-forward technology is adopted to carry out motion target tracking, complete moving object classification, and calculate wagon flow direction and wagon flow average speed, when flow speeds is greater than respective threshold, integrated car stream average speed and vehicle density draw traffic state data (the increase and decrease state of such as wagon flow average speed) and export.Traffic behavior I is defined as flow speedsand vehicle densityfunction:
，
Wherein x, y and z represent the position in region to be analyzed respectively, one group of xyz represents a region, section to be analyzed, α and β is respectively flow speeds factor of influence and vehicle density factor of influence, and preferred default value is 0.5, Thd1 is vehicle density preset (value), and preferred default value is 0.4.
Wherein, vehicle density refers to the vehicle number distributed in the unit length in an a certain instantaneous interior track, and it represents the intensity of vehicle distribution, formula is: K=N/L(pcu/km), wherein K is vehicle density, and N is the vehicle number (pcu) in bicycle road segment segment, and L is road section length.
Step c mainly weighs the density case of wagon flow by the quantity analyzing vehicle in the average speed of vehicle operating in region to be analyzed and region, only have and meet vehicle fleet size and/or density when being greater than respective threshold, the result that flow speeds is analyzed is just effective, again by the state of flow speeds, calculate vehicle density, the state machine (as shown in Figure 3) that final formation one is annular.
As shown in Figure 4, the detailed process of described steps d can be: first in panoramic video, mark road, set up the corresponding relation between flow analysis region and road section, according to this corresponding relation, traffic state data is plotted in panoramic video with the lines of different colours; Then, first adopt video calibration technique to choose two, three-dimensional point pair, calculate the projection matrix M of described panoramic video to described GIS map, realize the registration of described panoramic video and described GIS map, set up the three-dimensional coordinate mapping table of road area image coordinate and corresponding road area in described GIS map in described panoramic video; Magnitude of traffic flow segmentation statistical form is set up afterwards in GIS map, according to three-dimensional coordinate mapping table, the magnitude of traffic flow segmentation statistics in described panoramic video is mapped in described magnitude of traffic flow segmentation statistical form, and draws flow temperature figure (namely creating flow temperature figure) according to the statistics in described magnitude of traffic flow segmentation statistical form; Finally described flow temperature figure and described road panoramic video and described GIS map are merged respectively, form GIS temperature figure and GIS realistic picture, all information is all browsed by the mode of three-dimensional range, can overlook panorama, can watch details again.
Wherein said GIS temperature figure comprises described GIS map, described flow temperature figure and described magnitude of traffic flow segmentation statistics, and described GIS realistic picture comprises described GIS map and described road panoramic video.
Described GIS map shows all the time, and described flow temperature figure and described magnitude of traffic flow segmentation statistics are opened as required or close; Described road panoramic video is presented in described GIS map with the form of texture as required, or in described GIS map, play window display.Multiple display modes is the facility that user provides more choices and monitors.
As shown in Figure 5, present invention also offers a kind of panoramic video traffic situation supervisory control system, comprise video capture device, GPU concatenation module, traffic flow analysis module and the information registration and fusion treatment module that connect successively by information flow direction, described information registration is connected with the adaptable interface two-way communication of monitoring display equipment with fusion treatment module, described information registration and fusion treatment module are also provided with panoramic video input and GIS map information input terminal, and the output of described GPU concatenation module is also connected with described panoramic video input.
Described video capture device is for gathering live video to be monitored, and live video to be monitored is used as the monoscopic video on the basis forming panoramic video; Described GPU concatenation module is used for the decoding splicing of described monoscopic video and colour correction process thus forms panoramic video; Described traffic flow analysis module is used for moving object detection and traffic behavior assessment in described panoramic video; Described information registration and fusion treatment module be used for described panoramic video, described traffic state data and comprise scene to be monitored GIS map by information registration and fusion treatment, form GIS live-action map; Described adaptable interface is used for identification and the propelling movement of GIS live-action map of monitoring display device type, the type by device gateway identification subscriber equipment specifically, receive the request of data of subscriber equipment, and information registration and fusion treatment module are submitted in this request, simultaneously by the result data retransmission of information registration and fusion treatment module to subscriber equipment.
Described panoramic video traffic situation supervisory control system is also provided with GIS map data memory module, and the output of described GIS map data memory module connects described GIS map information input terminal, and described GIS map data memory module is used for stored GIS map datum.
Described GPU concatenation module is multiple parallel connection, and each described GPU concatenation module correspondence connects a video capture device group, and each described video capture device group comprises one or more described video capture device.