Three-dimensional panoramic video fusion monitoring platform
Technical Field
The invention relates to the technical field of video monitoring systems, in particular to a three-dimensional panoramic video fusion monitoring platform.
Background
The monitoring is a physical basis for real-time monitoring of key departments or important places in various industries, and management departments can obtain effective data, image or sound information through the monitoring and memorizing system, and timely monitor and memorize the process of sudden abnormal events so as to provide efficient and timely command and height, police arrangement, case handling and the like. The monitoring system is composed of camera shooting, transmission, control, display, record and registration and the like. The video camera transmits video images to the control host through the coaxial video cable, the control host distributes video signals to each monitor and video equipment, and simultaneously, voice signals needing to be transmitted can be synchronously recorded into the video recorder. Through the control host, an operator can send out an instruction to control the up-down, left-right and left-right actions of the pan-tilt and the operation of focusing and zooming the lens, and the switching between the multi-path camera and the pan-tilt can be realized through the control host.
The prison security system has complex work content and great responsibility, so a monitoring system is required to be arranged, and the monitoring system is mainly used for preventing prisons from escaping, fighting, self-injuring and self-disabling, harassing, rioting, alarming, hurdling prisons, damaging administrative and educational cadres and workers, information leakage, illegal entry of outsiders, holding and carrying out of prohibited articles, processing after the events, storage and filing of historical data and the like.
However, the current security system of prisons generally has the following problems: the existing video monitoring system is imperfect, some regional video monitoring equipment cannot cover the system, and once an emergency occurs, the emergency cannot be verified afterwards. The intrusion alarm system can only carry out overall arming and disarming on the whole site, and layered and regional arming and disarming cannot be realized, and the traditional shot monitoring is difficult to effectively associate two-dimensional monitoring video with the actual three-dimensional environment of a key region due to the lack of the overall scene information of a monitoring region, so that the overall real-time grasping and control on a large scene cannot be realized, and effective means cannot be really provided for a commander to realize macroscopic command monitoring, overall association and comprehensive scheduling.
Disclosure of Invention
The invention aims to provide a three-dimensional panoramic video fusion monitoring platform which has the advantages that panoramic video fusion can be carried out on monitoring videos which are scattered in a monitoring scene and have different field angles and a three-dimensional model of the monitoring scene, space correlation among different video pictures in the scene is formed, integrated monitoring of the three-dimensional panoramic video fusion is realized, and overall three-dimensional monitoring of the whole area in a single picture is realized.
The above object of the present invention is achieved by the following technical solutions:
a three-dimensional panoramic video fusion monitoring platform comprises:
the access part is used for accessing various video resources and data acquired by the camera and integrating various image resources with different sources and different formats;
panoramic three-dimensional fusion monitoring platform: video transmitted by the access part is subjected to three-dimensional panoramic video fusion display, intelligent automatic patrol, three-dimensional panoramic historical video playback, three-dimensional panoramic video associated display, three-dimensional measurement and planning rehearsal, alarm linkage screen hanging display and three-dimensional panoramic video intelligent analysis display and output to display equipment according to requirements;
the three-dimensional panoramic video fusion display part comprises the following modules:
panoramic video concatenation module: matching a plurality of videos provided by the acquisition part by adopting characteristic points, finding out the mapping relation among the images, transforming the plurality of images into the same coordinate system according to the coordinate change relation, and carrying out deformation correction and edge fusion processing to obtain and output a complete panoramic three-dimensional video;
the three-dimensional panorama fusion module: and under the three-dimensional geographic information framework, fusing and displaying a plurality of videos finished by the panoramic video splicing module at corresponding real positions in the three-dimensional model, thereby realizing the overall situation of the three-dimensional overall monitoring.
By adopting the technical scheme, the panoramic video fusion can be carried out on the monitoring videos which are discrete in the monitoring scene and have different field angles and the three-dimensional model of the monitoring scene, the spatial correlation among different video pictures in the scene is formed, the integrated monitoring of the three-dimensional panoramic video fusion is realized, a user can roam in the whole three-dimensional map space, the concerned field details can be captured quickly and accurately, and the use efficiency of the monitoring system can be improved.
The invention is further configured to: the reconstruction of the three-dimensional model comprises the following:
the method comprises the steps of carrying out three-dimensional reconstruction on an outdoor scene by adopting unmanned aerial vehicle three-dimensional scanning, combining GPS position and image video tracking, measuring point positions and attitude angles of a camera, capturing dense point cloud, and then carrying out three-dimensional point cloud reconstruction on a high-performance computing platform.
By adopting the technical scheme, the outdoor scanning reconstruction of the unmanned aerial vehicle is adopted for modeling, and the low-cost and accurate model reconstruction and virtual-real fusion are realized.
The invention is further configured to: the intelligent patrol part is used for presetting patrol paths on the three-dimensional model on the basis of the three-dimensional model, and dynamically forming the optimal patrol paths to carry out depth and breadth patrol by taking all video points as nodes based on a graph theory searching technology, so that automatic video patrol is realized.
By adopting the technical scheme, real-time, quick, non-repetitive and non-missing video traversing cruising and quick cruising combining 'point-line-surface' are realized, and the daily security patrol and control capacity and efficiency are fundamentally improved.
The invention is further configured to: the three-dimensional panoramic historical video playback part is used for reading in a plurality of historical videos of the cameras in storage, visualizing the historical videos into an integral three-dimensional model scene, and uniformly returning all videos to a certain past moment as required to realize forward or reverse playing and searching in a full space-time environment.
By adopting the technical scheme, the source can be traced efficiently in the panoramic scene display mode, the whole playback of the historical events in the whole scene is ensured, the event route is described by the whole picture, the inquiry capability of the historical events and the whole scene reproduction capability of the events are improved, the whole playback of the historical events in the whole scene is ensured by the three-dimensional panoramic historical video playback function, the traditional working mode of depending on shot scattered reverse check is changed, the review efficiency of the historical events can be effectively improved, and the resource investment is reduced.
The invention is further configured to: the panoramic video related display part monitors the three-dimensional panoramic video in a mode of combining the gun type camera and the ball type camera, and the ball machine automatically tracks the target by locking the target in the full-scene video to realize linkage tracking of the target.
By adopting the technical scheme, monitoring personnel do not need to know the position and the observation angle of the dome camera in advance, click the target interest point in the visual field range in the panoramic picture, automatically zoom the dome camera in the panoramic picture according to the position of the target, and adjust the visual angle to monitor details.
The invention is further configured to: the three-dimensional measurement and planning rehearsal part clicks a target or an area to be observed on a three-dimensional scene or a two-dimensional global map, and three-dimensional measurement and planning rehearsal are performed on a three-dimensional model.
By adopting the technical scheme, the user can know the rationality of the deployment scheme and the defense deployment scheme by carrying out three-dimensional measurement and planning rehearsal on the three-dimensional model according to the tracking and monitoring requirements.
The invention is further configured to: the alarm linkage screen-hanging display part organically combines three-dimensional virtual reality display with traditional matrix video, and places cameras such as key entrances and exits on the second screen for display, and the cameras for dynamic alarm also dynamically display on the second screen, and combines global situation with key position screen-dividing details.
By adopting the technical scheme, the alarm linkage screen-hanging display surpasses the limitation of display in a single space of the traditional video monitoring system, and the global unified situation is realized.
The invention is further configured to: the three-dimensional panoramic video intelligent analysis part utilizes computer vision and data analysis technology to carry out cross-camera personnel tracking, tripwire detection, timing forbidden zone detection, people counting and density estimation on a monitored scene.
By adopting the technical scheme, the monitoring effect of the monitoring platform can be further improved, the working intensity of a user is saved, and the working efficiency is improved.
The invention is further configured to: the cross-camera personnel tracking function adopts a method based on sensor identification and visual fusion, and the cross-camera tracking of the sportsman is realized by synthesizing the face comparison alarm notifications.
By adopting the technical scheme, the user can pay attention to the global motion track of a certain person or object, continuous tracking is carried out in the three-dimensional panoramic video, and monitoring is facilitated.
The invention is further configured to: the tripwire detection function adopts a Gaussian model for background modeling, foreground target extraction is carried out by using background difference, all points on a virtual line are scanned, and when a detection target passes through the tripwire, if the detection target is a bidirectional tripwire, the system directly alarms; if the tripwire is a one-way tripwire, the target movement direction is determined by using the Euler distance of the target color histogram, and if the target movement direction is consistent with the crossing forbidding direction, the system alarms and is used for setting a forbidden zone for alarming for a sensitive area or a position.
The invention is further configured to: the timing forbidden zone detection function sets a forbidden zone for a certain zone in the three-dimensional panoramic space according to the needs of a user, and within a specified time range, if a moving object enters the forbidden zone, the system gives an alarm, and the cruise camera is adjusted back to the alarm zone for alarm confirmation and processing.
By adopting the technical scheme, the alarm device can be used for setting a forbidden zone in a fixed time range in a certain fixed area for alarming, does not need to be monitored all the time by personnel, reduces the working strength and improves the monitoring accuracy.
The invention is further configured to: the people counting and density estimation function performs people counting based on machine learning, extracts input image foreground to obtain different features, including: edge features, foreground features, gray level co-occurrence matrices, HOG features, and the like; carrying out perspective transformation on the obtained features, and increasing the weight of the image pixel features with long distance; and performing regression in different regression modes according to different estimated densities to obtain statistics of the number of people in the monitored scene.
By adopting the technical scheme, the method can be used for counting the number of people at key intersections or areas in a certain place, estimating the density of the intersections and giving an alarm for abnormal density change.
The invention is further configured to: the camera and the sensor in the three-dimensional model are displayed by adopting the POI labels, so that the condition that a plurality of cameras related to the physical position can be observed at the same time under one visual angle is ensured.
In conclusion, the beneficial technical effects of the invention are as follows:
1. the method comprises the steps of performing panoramic video fusion on discrete traditional monitoring videos with different field angles and a three-dimensional model of a monitoring scene to form spatial correlation among different video pictures in the scene, realizing an integrated monitoring system for three-dimensional panoramic video fusion, and supporting functions of global three-dimensional monitoring of an integral area in a single picture, monitoring detail display in sub-pictures, user-definable video intelligent automatic patrol, intelligent analysis and the like;
2. the video real-time, rapid, non-repetitive and non-missing video traversal cruise and the rapid cruise combining the point, line and plane are realized, and the daily security patrol control capability and efficiency are fundamentally improved;
3. the monitoring personnel do not need to know the position and the observation angle of the dome camera in advance, click the target interest point in the visual field range in the panoramic picture, the dome camera in the panoramic picture can automatically zoom according to the position of the target, and the visual angle is adjusted to monitor the details.
Drawings
FIG. 1 is a schematic diagram of the architecture of the present invention.
Fig. 2 is a schematic diagram of a panoramic three-dimensional fusion monitoring platform architecture of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
Referring to fig. 1, the three-dimensional panoramic video fusion monitoring platform disclosed by the invention comprises a system platform architecture and a system hardware architecture. The system platform architecture includes the following five modules: the system comprises a data layer, an access layer, a data management scheduling layer, a data processing layer and an application layer.
Firstly, a data layer: the system is mainly responsible for collecting and accessing basic data of the whole platform, and the accessible data types are rich and diverse, including video data (panoramic video, network gunshot, network dome camera and other platform resources) and environment data (three-dimensional models and GIS data), and cover structured and unstructured data types.
Second, access layer: the method comprises video data access and multi-element data access, realizes formatting, normalization and standardization processing of data accessed to a video in terms of data format and control signaling, and formats, normalization and standardization processing of accessed multi-source data in terms of data format, data encapsulation and transmission protocol, so that the inside of a system processes various accessed data according to a uniform format, the complexity of data format conversion inside the system is reduced, and the data processing efficiency inside the system is improved.
Thirdly, a data management scheduling layer: the data management and scheduling in the system are mainly completed, and the forwarding, distribution, storage, association and linkage relation configuration among the data and the like of various data are realized.
And fourthly, a data processing layer: the system is in charge of carrying out depth processing on data distributed from a lower layer, and finishing optimization processing on a video through a computer graphic image processing algorithm embedded in the system, wherein the optimization processing comprises panoramic three-dimensional processing and video optimization, and the panoramic three-dimensional processing comprises video splicing processing, model data optimization and virtual-real fusion processing; the video optimization comprises the optimization processing of chromatic aberration correction, video enhancement, electronic anti-shake and electronic fog penetration, and provides more fine data service for upper-layer application.
Fifthly, application layer: the system is a unique interface for interaction between a terminal user and a site panoramic monitoring system, provides a basic environment for service operation, basic applications related to videos (real-time on-demand and video playback), related applications based on panoramic videos, a series of applications based on video images and the like (three-dimensional panoramic video fusion display, intelligent automatic patrol, three-dimensional panoramic historical video playback, three-dimensional panoramic video association display, three-dimensional measurement and planning preview, alarm linkage screen hanging display and three-dimensional panoramic video intelligent analysis) for the user on the basis of a database, a map engine and a three-dimensional engine, and provides a secondary development interface to meet the requirement of other system data sharing of the user.
The system hardware architecture includes the following three systems: the system comprises an access part, a basic service platform and a panoramic three-dimensional fusion monitoring platform.
Firstly, an access part: the video management and control system is used for seamless access of domestic and foreign mainstream video monitoring equipment, video resources and data of various sources and formats acquired by front-end cameras distributed in monitoring places are accessed to a panoramic video monitoring management system to build a video management and control system with unified encoding and decoding standards, unified networking protocols, unified control protocols, unified numbering rules, unified image labels and unified position marks, and image resources of various sources and different formats are integrated in the system.
II, a basic service platform: the method provides basic service for video monitoring, and performs fusion processing, graphic processing and other operations on the incoming multi-channel video of the access part through the video analysis server through the access management server.
Thirdly, a panoramic three-dimensional fusion monitoring platform: and processing and displaying the video processed by the basic service platform according to the actual requirement of the user and outputting the processed video to the display equipment.
Referring to fig. 2, the panoramic three-dimensional fusion monitoring platform comprises the following seven functions:
1. the three-dimensional panoramic video fusion display part comprises a panoramic video splicing module and a three-dimensional panoramic fusion module. The panoramic video splicing module carries out image preprocessing on different shot videos, a characteristic point matching strategy is adopted to find out the mapping relation between two images, then the two images are transformed into the same coordinate system according to the coordinate change relation, and the complete large-range panoramic video is obtained through deformation correction and edge fusion processing. In order to meet the real-time performance and the accuracy of monitoring, the panoramic video monitoring is divided into an initialization part and a normal monitoring part. An initialization part, which determines the spatial position relationship of the camera through image matching and introduces an optimization idea to determine the splicing gap position of two images; and the normal monitoring part carries out a rapid fusion scheme aiming at each frame of video information according to the parameters determined in the initialization stage, thereby realizing real-time and accurate panoramic video monitoring.
And the three-dimensional panoramic fusion module fuses and displays a plurality of videos finished by the panoramic video splicing module at corresponding real positions in the three-dimensional model under a three-dimensional geographic information frame, so that the overall situation of the three-dimensional overall monitoring is realized. The three-dimensional model carries out three-dimensional reconstruction on an outdoor scene through three-dimensional scanning of the unmanned aerial vehicle, combines the GPS position with image video tracking, measures the point location and the attitude angle of the camera, captures dense point cloud, and then carries out three-dimensional point cloud reconstruction on a high-performance computing platform. Under the condition of a large scene, unmanned aerial vehicles can be used for formation to carry out scanning integration. And the method combining manual modeling and photo modeling is adopted indoors. Therefore, the problem that the traditional shot lacks the whole scene information of a monitoring area and is difficult to effectively associate a two-dimensional monitoring video with the actual three-dimensional environment of a key area is solved, a user can roam in the whole three-dimensional map space, the concerned site details can be quickly and accurately captured, the large scene can be globally grasped and controlled in real time, and effective means can be really provided for a commander to realize macroscopic command monitoring, integral association and comprehensive scheduling.
In addition, the cameras and the sensors in the monitoring scene are displayed by adopting POI labels, so that the condition that a plurality of cameras related to the physical position can be observed at the same time under one visual angle is ensured. The method comprises the steps of displaying various types of sensors in a labeling mode, wherein the numerical values of the sensors can be displayed on the labels, and the labels can be informed of the numerical values of the sensors through appearance and flashing when an alarm occurs.
2. The intelligent automatic patrol part presets a patrol path on the three-dimensional model on the basis of the three-dimensional model to realize automatic video patrol, and organically combines a monitoring picture with the three-dimensional model, so that a user can roam and patrol the panoramic video in a more intuitive mode, thus being beneficial to patrol personnel to acquire and understand information more thoroughly and making judgment quickly and accurately. In addition, based on a graph theory search technology, the optimal patrol path is dynamically formed by taking all the video points as nodes to carry out depth and breadth patrol, and the daily security patrol capacity and efficiency are fundamentally improved. According to the actual situation of the monitoring scene and the user demand, automatic roaming cruise, full-coverage cruise, multi-azimuth cruise on point, line and plane and the like can be further realized, the visual fatigue of the patrol security and control personnel is greatly reduced, and the monitoring efficiency and the camera utilization rate are improved.
3. The three-dimensional panoramic historical video playback part reads in a plurality of historical videos of cameras in storage, visualizes the historical videos into an integral three-dimensional model scene, uniformly recalls all videos to a certain past moment according to user needs, realizes forward or reverse playing and searching in a full space-time environment, changes a single-path storage mode of traditional videos, avoids limitations of historical video retrieval and playback, and reduces a large amount of manpower and material resources required for searching a certain path of historical video in a large amount of stored videos. Efficient tracing is carried out in a panoramic scene display mode, the whole playback of the historical events in the whole scene is ensured, the event context is described by a whole picture, and the inquiry capability of the historical events and the whole scene reproduction capability of the events are improved. The three-dimensional panoramic historical video playback function ensures that the historical events are played back integrally in the whole scene, the traditional working mode of depending on shot scattering and reverse checking is changed, the back checking efficiency of the historical events can be effectively improved, and the resource investment is reduced.
4. The panoramic video related display part monitors the three-dimensional panoramic video in a mode of combining the gun type camera and the ball type camera, and the ball machine automatically tracks the target by locking the target in the full-scene video to realize linkage tracking of the target. Therefore, monitoring personnel do not need to know the position and the observation angle of the dome camera in advance, target interest points in the visual field range are clicked in the panoramic picture, the dome camera in the panoramic picture can be automatically zoomed according to the position of the target, and the visual angle is adjusted to monitor details.
5. The three-dimensional measurement and planning rehearsal part clicks a target or an area to be observed on a three-dimensional scene or a two-dimensional global map, and three-dimensional measurement and planning rehearsal are performed on a three-dimensional model, so that a user can know the rationality of a deployment scheme and a defense deployment scheme according to tracking and monitoring requirements.
6. The alarm linkage screen-hanging display part organically combines three-dimensional virtual reality display with traditional matrix video, realizes global unified situation, resides position cameras such as key entrances and exits in the second screen for display, dynamically displays the position cameras on the second screen, combines the global situation with the key position split screen details, and keeps the advantages of the global situation and the key position split screen details.
7. The three-dimensional panoramic video intelligent analysis part comprises an intelligent analysis and detection function, and cross-camera personnel tracking, tripwire detection, timing forbidden zone detection, people counting and density estimation are carried out on a monitored scene by utilizing a computer vision and data analysis technology so as to meet the specific requirements of operation and maintenance and security of a certain place.
The cross-camera personnel tracking function adopts a method based on sensor identification and visual fusion, and the alarm notifications are synthesized through face ratio and the like, so that cross-camera tracking of the personnel is realized, a user can pay attention to the global motion track of a certain person or object, continuous tracking is carried out in a three-dimensional panoramic video, and continuous observation of the personnel motion tracks crossing different cameras is realized.
And the tripwire detection function manually draws a virtual line on the video image as a tripwire according to the self requirement of a user, and when the system detects that a moving target crosses the tripwire, namely the intersection is formed between the pixel point set of the image foreground target and the pixel point set of the tripwire on the image, the system sends out an alarm. In the tripwire detection, a Gaussian model is used for background modeling, foreground target extraction is carried out by using background difference, all points on a virtual line are scanned, and when a detection target passes through the tripwire, if the detection target is a bidirectional tripwire, the system directly alarms; if the tripwire is a one-way tripwire, the target movement direction is determined by using the Euler distance of the target color histogram, and if the target movement direction is consistent with the crossing forbidding direction, the system alarms and is used for setting a forbidden zone for alarming for a sensitive area or a position.
The timing forbidden zone detection function sets a forbidden zone for a certain zone in the three-dimensional panoramic space according to the needs of a user, and within a specified time range, if a moving object enters the forbidden zone, the system gives an alarm, and the cruise camera is adjusted back to the alarm zone for alarm confirmation and processing. The alarm device is used for setting forbidden zones in a certain fixed area within a fixed time range for alarming.
The people counting and density estimation function performs people counting based on machine learning, extracts input image foreground to obtain different features, including: edge features, foreground features, gray level co-occurrence matrices, HOG features, and the like; carrying out perspective transformation on the obtained features, and increasing the weight of the image pixel features with long distance; and performing regression in different regression modes according to different estimated densities to obtain statistics of the number of people in the monitored scene. The method for estimating the density of the people stream adopts a method combining various motion characteristics and corrects the characteristics through a perspective normalization algorithm. The system is used for counting the number of people at key intersections or areas in a certain place, estimating the density of the intersections and alarming for abnormal density change.
The three-dimensional panoramic video fusion monitoring platform is used as a core system of system integration, the linkage function of a monitoring system, an access control system, a fire fighting system and an environment monitoring system is realized, when any one system generates an abnormal alarm, the three-dimensional panoramic video monitoring platform can receive a node switching value signal sent by the system and display the node switching value signal, the alarm can be displayed in three colors, and the three colors respectively represent three different abnormal types. And then displaying the video of the abnormal alarm area in the monitoring platform, wherein the monitoring platform system can also provide alarm audio prompt, spherical camera steering, central control room flashing alarm lamp real-sense prompt and the like.
According to the invention, through building and optimizing video resources of key areas in a certain place, the discrete traditional monitoring videos with different field angles and the three-dimensional model of the monitoring scene are subjected to panoramic video fusion to form spatial correlation among different video pictures in the scene, so that an integrated monitoring system for three-dimensional panoramic video fusion is realized, and the functions of global three-dimensional monitoring of the whole area in a single picture, monitoring detail display in sub-pictures, user-defined video intelligent automatic patrol, intelligent analysis and the like are supported.
The embodiments of the present invention are preferred embodiments of the present invention, and the scope of the present invention is not limited by these embodiments, so: all equivalent changes made according to the structure, shape and principle of the invention are covered by the protection scope of the invention.