Cloud services: from video surveillance to video analytics
Video surveillance as a service (VSaaS, video surveillance as a service) based on cloud infrastructure is one of the most current industry trends. According to IMS Research, the UK, the global VSaaS market amounted to $ 500 million in 2011 and could exceed $ 1 billion in 2014, that is, in 3 years it could double.
At the same time, according to ABI Research, the United States, the market for video analytics tools for business will increase more than 2.5 times from 2011 to 2016, which will reach $ 900 million.
“Video analytics as a service” is located at the intersection of these two markets and emphasizes the focus on video analysis without operator involvement, while the term VSaaS today more often involves the possibility of remote viewing and recording without any analytics.
')
Our article analyzes the state of the market and VSaaS technologies, and also considers the possibility of overcoming the main obstacle to the development of the VSaaS market - limiting the outgoing subscriber channel - with the help of video analytics.
As noted by IMS Research, the growth of the VSaaS market is driven by demand from individual users, small and medium businesses and the state. For certain segments of users, VSaaS is more attractive than classic solutions based on Network Video Recorders (NVR) and Video Management Systems (VMS). Thus, the VSaaS commercialization model assumes that instead of the cost of a hardware and software solution without guarantees of return on investment, the consumer pays for a specific service, for example, video recording, automatic security call, data collection, and analytical reports. VSaaS service is highly scalable in terms of the volume of stored video, the number of observation points and the number of users of the system.
There are about 30 companies in the world that are actively developing the VSaaS direction, and their number is increasing every month. The target markets of these companies are retail chains, builders, small offices and households.
Among the major players, Axis, which developed the Axis AVHS infrastructure for providing VSaaS services through intermediaries, is worth noting. The limitation of Axis AVHS is that the VSaaS service only works with network cameras manufactured by Axis.
In Russia, only a few companies have entered or are planning to enter the VSaaS market. For example, the domestic company iVideon offers a convenient service.
video surveillance via the Internet, Profinegro actively implements business analytics services in retail networks. DSSL Company recently announced
Trassir Cloud service for automatic monitoring of video surveillance systems at the first stage and, apparently, with certain plans for the development of the spectrum of cloud services,
directly associated with video surveillance. Moscow government in
As part of a dangerous city buys video surveillance as a service for
competitive basis.
Thus, the following types of companies appear on the VSaaS market:
The most significant barrier to the expansion of VSaaS services in the world, and especially in Russia, is the insufficient capacity of communication channels outside the local network. Let's compare the data streams generated by the cameras with the average subscriber connection speeds.
On the one hand, even with the use of modern compression algorithms, such as H.264, standard-definition cameras (0.4 megapixels) form data stream from 0.5 to 4 Mbps, and high-definition cameras (13 megapixels) from 1 to 10 Mbit / s under good observation conditions and up to 50 Mbit / s - with bad ones. For systems with a large number of cameras, the costs of transmitting, storing and analyzing data become critical.
On the other hand, the average outgoing channel speed is 4 Mbit / s in the world and 13 Mbit / s in Russia according to NetIndex data for June 2012. When using asymmetric access technologies (for example, LTE, 4G, ADSL or cable modem), the outgoing channel from the subscriber to the VSaaS application is 410 times smaller than the incoming one. Thus, the VSaaS service does not allow for the remote viewing and recording of video from a larger number of cameras, especially high-resolution cameras (more than 1 megapixel).
According to our estimates, in the world the annual growth of video data recorded by cameras is more than 50%, and the increase in channel capacity is only 20%. Thus, the gap between the needs of VSaaS and the capabilities of communication channels is increasing annually by 30% per year.
Video analytics is the only technology that can solve the problem of the subscriber’s outgoing channel, as well as the problem of storing video in the cloud. Despite the advent of cost-effective video storage methods such as LAID (LinearArrayofIdleDisks), storing large
the volume of video in the cloud is the most costly component of the VSaaS service.
Video analytics can be viewed as a specialized encoder that leaves only the data that the user needs in a video. A universal coder, such as H.264, does not “understand” the degree of importance of each element in the image and, therefore, cannot effectively filter redundant data to provide VSaaS services. For example, a standard coder cannot distinguish a small person in the background and numerous droplets of rain in the foreground. If a person and each droplet is encoded with synodic detail, the stream will be substantially redundant for transmission and storage.
As a rule, a universal encoder and video analytics are used together, which allows you to take advantage of each separately.
Obviously, to reduce the load on communication channels, video analytics should work on the subscriber side. In addition, some types of video analytics, such as facial recognition , require uncompressed video analysis to achieve maximum accuracy. For these reasons, cloud infrastructure cannot be cost-effectively used for preprocessing video without any equipment on the subscriber side.
On the other hand, the cloud infrastructure can be effectively used to scale the video surveillance system in the following dimensions:
The higher the accuracy of video analytics, the lower the load on communication channels and cloud storage. Therefore, if the accuracy of the video analytics is known, the service provider or consumer can easily calculate the economic benefit of using video analytics.
Consider an example. For the protection of the perimeter of a large solar station (solar thermal installation) 300 high-resolution cameras (1.2 megapixels) are used. Under normal weather conditions, the total
video stream is 1.8 Gbps. Under unfavorable conditions, that is, when the signal is noisy, for example, at night, the stream almost doubles to 3.5 Gbit / s. The use of conventional motion detector allows you to reduce the amount of video data by an average of 80%, that is, to 0.40.7 Gbit / s.
Unfortunately, the peak load sometimes exceeds this range many times due to the fact that the motion detector reacts to global changes in illumination and weather conditions at once on all cameras.
The use of professional video analytics for detecting people in the solar station has reduced the average load to 0.010.02 Gbit / s, and the peak load to 0.05 Gbit / s. Thus, in comparison with the usual motion detector , video analytics reduced the load on the communication channel and cloud storage by more than 40 times.
Local video storage on the flash card in the camera or on the hard disk in the server gateway (gateway) on the subscriber side allows you to more efficiently use the outgoing channel due to data buffering. When a video analytics event occurs, the camera or local gateway server records video in local memory, which is then transferred to the cloud storage via an available communication channel. For such an event recording mode, the recording speed may be higher than the video data transfer speed limit.
Local storage is also a means of ensuring the fault tolerance of the video surveillance system in case of a temporary disconnection of the communication channel between the camera (local server) and the cloud application.
An example of a more complex way to apply video analytics as part of a VSaaS application is the video data ranking technology developed by Synesis . Here, video transmission via communication channels with limited bandwidth does not occur at the expense of video quality, as in simple DVRs, but due to the priority transmission of the most important information for the user, and the priority of video fragments is automatically determined by video analytics. This approach also improves the efficiency of the use of cloud disk space: the storage time for video data in a local storage or cloud may depend on their importance.
In conclusion, we consider the main differences between the classical video management system (VMS) and the cloud application for the provision of the VSaaS service:
The article assumes that cloud infrastructure is used to provide VSaaS or “Video analytics as a service”. Some categories of users will prefer dedicated servers. The considered findings apply equally to cloud and dedicated server application allocations.
At the same time, according to ABI Research, the United States, the market for video analytics tools for business will increase more than 2.5 times from 2011 to 2016, which will reach $ 900 million.
“Video analytics as a service” is located at the intersection of these two markets and emphasizes the focus on video analysis without operator involvement, while the term VSaaS today more often involves the possibility of remote viewing and recording without any analytics.
')
Our article analyzes the state of the market and VSaaS technologies, and also considers the possibility of overcoming the main obstacle to the development of the VSaaS market - limiting the outgoing subscriber channel - with the help of video analytics.
As noted by IMS Research, the growth of the VSaaS market is driven by demand from individual users, small and medium businesses and the state. For certain segments of users, VSaaS is more attractive than classic solutions based on Network Video Recorders (NVR) and Video Management Systems (VMS). Thus, the VSaaS commercialization model assumes that instead of the cost of a hardware and software solution without guarantees of return on investment, the consumer pays for a specific service, for example, video recording, automatic security call, data collection, and analytical reports. VSaaS service is highly scalable in terms of the volume of stored video, the number of observation points and the number of users of the system.
There are about 30 companies in the world that are actively developing the VSaaS direction, and their number is increasing every month. The target markets of these companies are retail chains, builders, small offices and households.
Among the major players, Axis, which developed the Axis AVHS infrastructure for providing VSaaS services through intermediaries, is worth noting. The limitation of Axis AVHS is that the VSaaS service only works with network cameras manufactured by Axis.
In Russia, only a few companies have entered or are planning to enter the VSaaS market. For example, the domestic company iVideon offers a convenient service.
video surveillance via the Internet, Profinegro actively implements business analytics services in retail networks. DSSL Company recently announced
Trassir Cloud service for automatic monitoring of video surveillance systems at the first stage and, apparently, with certain plans for the development of the spectrum of cloud services,
directly associated with video surveillance. Moscow government in
As part of a dangerous city buys video surveillance as a service for
competitive basis.
Thus, the following types of companies appear on the VSaaS market:
- young companies (startups) focused on specific market segments, for example, iVideon, Profintegro;
- manufacturers of video surveillance systems (VMS), for example, SatelliteInnovation (trademark Mascroscop, DSSL (Trassir);
- camera manufacturers, for example, Axis;
- telecom operators, for example, Megaphone .
The most significant barrier to the expansion of VSaaS services in the world, and especially in Russia, is the insufficient capacity of communication channels outside the local network. Let's compare the data streams generated by the cameras with the average subscriber connection speeds.
On the one hand, even with the use of modern compression algorithms, such as H.264, standard-definition cameras (0.4 megapixels) form data stream from 0.5 to 4 Mbps, and high-definition cameras (13 megapixels) from 1 to 10 Mbit / s under good observation conditions and up to 50 Mbit / s - with bad ones. For systems with a large number of cameras, the costs of transmitting, storing and analyzing data become critical.
On the other hand, the average outgoing channel speed is 4 Mbit / s in the world and 13 Mbit / s in Russia according to NetIndex data for June 2012. When using asymmetric access technologies (for example, LTE, 4G, ADSL or cable modem), the outgoing channel from the subscriber to the VSaaS application is 410 times smaller than the incoming one. Thus, the VSaaS service does not allow for the remote viewing and recording of video from a larger number of cameras, especially high-resolution cameras (more than 1 megapixel).
According to our estimates, in the world the annual growth of video data recorded by cameras is more than 50%, and the increase in channel capacity is only 20%. Thus, the gap between the needs of VSaaS and the capabilities of communication channels is increasing annually by 30% per year.
Video analytics is the only technology that can solve the problem of the subscriber’s outgoing channel, as well as the problem of storing video in the cloud. Despite the advent of cost-effective video storage methods such as LAID (LinearArrayofIdleDisks), storing large
the volume of video in the cloud is the most costly component of the VSaaS service.
Video analytics can be viewed as a specialized encoder that leaves only the data that the user needs in a video. A universal coder, such as H.264, does not “understand” the degree of importance of each element in the image and, therefore, cannot effectively filter redundant data to provide VSaaS services. For example, a standard coder cannot distinguish a small person in the background and numerous droplets of rain in the foreground. If a person and each droplet is encoded with synodic detail, the stream will be substantially redundant for transmission and storage.
As a rule, a universal encoder and video analytics are used together, which allows you to take advantage of each separately.
Obviously, to reduce the load on communication channels, video analytics should work on the subscriber side. In addition, some types of video analytics, such as facial recognition , require uncompressed video analysis to achieve maximum accuracy. For these reasons, cloud infrastructure cannot be cost-effectively used for preprocessing video without any equipment on the subscriber side.
On the other hand, the cloud infrastructure can be effectively used to scale the video surveillance system in the following dimensions:
- storage video and metadata video analytics;
- connection of new objects of observation (for example, outlets);
- implementation of new metadata analysis and archive search functions;
- service a large number of users.
The higher the accuracy of video analytics, the lower the load on communication channels and cloud storage. Therefore, if the accuracy of the video analytics is known, the service provider or consumer can easily calculate the economic benefit of using video analytics.
Consider an example. For the protection of the perimeter of a large solar station (solar thermal installation) 300 high-resolution cameras (1.2 megapixels) are used. Under normal weather conditions, the total
video stream is 1.8 Gbps. Under unfavorable conditions, that is, when the signal is noisy, for example, at night, the stream almost doubles to 3.5 Gbit / s. The use of conventional motion detector allows you to reduce the amount of video data by an average of 80%, that is, to 0.40.7 Gbit / s.
Unfortunately, the peak load sometimes exceeds this range many times due to the fact that the motion detector reacts to global changes in illumination and weather conditions at once on all cameras.
The use of professional video analytics for detecting people in the solar station has reduced the average load to 0.010.02 Gbit / s, and the peak load to 0.05 Gbit / s. Thus, in comparison with the usual motion detector , video analytics reduced the load on the communication channel and cloud storage by more than 40 times.
Local video storage on the flash card in the camera or on the hard disk in the server gateway (gateway) on the subscriber side allows you to more efficiently use the outgoing channel due to data buffering. When a video analytics event occurs, the camera or local gateway server records video in local memory, which is then transferred to the cloud storage via an available communication channel. For such an event recording mode, the recording speed may be higher than the video data transfer speed limit.
Local storage is also a means of ensuring the fault tolerance of the video surveillance system in case of a temporary disconnection of the communication channel between the camera (local server) and the cloud application.
An example of a more complex way to apply video analytics as part of a VSaaS application is the video data ranking technology developed by Synesis . Here, video transmission via communication channels with limited bandwidth does not occur at the expense of video quality, as in simple DVRs, but due to the priority transmission of the most important information for the user, and the priority of video fragments is automatically determined by video analytics. This approach also improves the efficiency of the use of cloud disk space: the storage time for video data in a local storage or cloud may depend on their importance.
In conclusion, we consider the main differences between the classical video management system (VMS) and the cloud application for the provision of the VSaaS service:
- VSaaS involves the use of more complex technologies to control the communication channel between the object of observation and the cloud, as well as between the cloud and the subscriber. It is necessary to take into account restrictions on the speed of transmission, as well as the presence of complex topologies with address translation (NAT) and firewalls.
- VMS assumes the presence of permanent operators in front of monitors of security television, while VSaaS subscribers most often connect to the system through the browser only after receiving an alarm message or for analyzing the video archive (reports).
- As a consequence of points 1 and 2, video analytics play a more important role in VSaaS than VMS. Video analytics is used to control data transmission, generating alarm messages and reports.
- The VSaaS user interface should be implemented on the basis of a browser or mobile phone, while VMS often uses installable applications.
- Local storage support is more important in VSaaS applications than VMS.
- VSaaS should serve significantly more users and cameras than VMS.
- As a rule, VMS applications are configured by professional system integrators, and VSaaS applications are end users, which place higher demands on the simplicity of the VSaaS interface.
- VSaaS service operators and / or their subscribers are more sensitive to the cost of the final equipment (cameras, local video servers) than VMS users.
The article assumes that cloud infrastructure is used to provide VSaaS or “Video analytics as a service”. Some categories of users will prefer dedicated servers. The considered findings apply equally to cloud and dedicated server application allocations.
Source: https://habr.com/ru/post/147117/
All Articles