VSaaS 2025: Future Video Surveillance Technologies


    Over the past 15 years, video surveillance has changed a lot. We switched from analog cameras to digital ones, the resolution of the matrixes has increased significantly, the autotune of contrast and illumination has become better. Cameras are now configured in the cloud - it is easy to store video and launch video analytics modules in the same place.

    The global market for VSaaS is growing at a rate of 22% annually and will reach $ 6 billion by 2022. Network bandwidth and traffic costs are still a barrier, but in the coming years, almost all systems will switch to cloud solutions. There are several reasons for the dominance of clouds, but the main thing is that VSaaS is much more than just video surveillance.

    Many in the industry are wondering what global changes are waiting for us in the next 5-7 years. Most likely, the systems we are accustomed to now will look different. Relying on some recent trends, we can now visualize a picture of VSaaS solutions for the near future.

    Video analytics and new algorithms

    The video surveillance system created by the Chinese company SenseTime

    The inaccuracy of the algorithms, the large hardware requirements, the high cost and the complexity of end-use have hampered the development of analytical modules for many years. Learning neural networks inhibited the quality of the video itself. It got to the point that some companies hired actors to shoot training videos with the formulation of illegal actions.

    New methods of deep learning allowed us to analyze video data faster and cheaper than ever before. Video analytics has become available in home cameras and for business. The Ivideon service, for example, offers a face recognition system for businesses at a price of 1,700 rubles per camera.


    Face recognition is just the beginning. We are already seeing the first results of combining weak (as yet) artificial intelligence with video surveillance systems. In the project IC Realtime Ella AI is used for video stream analysis and instant search. Ella can recognize hundreds of thousands of queries in natural language, allowing users to search for frames with specified objects: animals, people in clothes of a certain color, or even individual brands of cars.


    In the article "The eye in the sky»Researchers describe a system in which video data from a quadcopter is analyzed by a neural network trained using deep learning to evaluate people's postures and compare them with postures that researchers have identified as“ violent ”. The project included five poses: strangulation, punching and kicking, shooting and stabbing. Scientists hope that the "eye in the sky" will be used to detect crimes in public places and at major events.

    Although deep learning methods have helped define a person in a video stream, it is still difficult to make a step towards “perfect” algorithms. Therefore, systems that detect criminals are so imperfect. The car is difficult to understand who is in front of her, a criminal or a benevolent person who just decided to hug a friend. At the moment, the system works with an accuracy of 94% in the definition of "violent" poses, but the more people appear in the frame, the lower this indicator - the accuracy decreased to 79% while analyzing the actions of 10 people.

    The pace of technical progress leaves no room for doubt - by 2025, accuracy will be close to 100% on any number of people in a crowd. Opportunities to recognize illegal actions will reduce the number of specialized personnel (police, security guards), thereby significantly leveling the human factor in relation to the control of complex social situations.


    Algorithms also help to improve the work of even outdated hardware. The team of scientists presented the “computational periscopy” algorithm, which allows any IP camera to literally “look around the corner”. The algorithm is based on the analysis of object shadows, dropped on any surface.

    To obtain an image of an object located around the corner, the algorithm requires an image of the surface on which the shadow of the object falls. At the moment, to “calculate” the image “from the corner”, you need only a regular computer, which will do all the necessary calculations in no more than 48 seconds. Higher-powered computers will do this even faster.

    Access for various devices

    Swiveling speed camera Nobelic 4225Z-ASD with 25x zoom. One of the most “powerful” cameras in the Ivideon

    VSaaS store allows access to camera images and information notifications from computers and mobile devices, but most importantly, it combines different types of equipment. You can connect IP cameras, old analog cameras, thermal cameras, various sensors to the cloud. At the same time, the cameras are stationary, on board ships, in trucks, even on drones.

    While many manufacturers are trying to make their cameras “smarter”, VSaaS solutions allow you to “upgrade” any existing cameras that require only an Internet connection. That is why we effectively help modernize any object for observation. A business owner does not spend a lot of money building a system from scratch. Old analog cameras work in conjunction with the most advanced equipment to date - 4K cameras.

    Camera manufacturers have only recently started selling 4K devices. Not least because it’s difficult to store videos of this resolution. Fortunately, in the cloud service the limit is set to the archive depth in time, and not in terabytes.

    With the advent of 4K video surveillance, you can save on the equipment itself - a single high-definition camera can replace several ordinary cameras. UHD also offers more opportunities for zoom, face recognition and license plate technologies. 4K quality and other video processing formats such as H.265 are likely to be the next step in the industry.

    Above, we mentioned an algorithm that helps the camera literally see what happens around the corner. Scientists from Scotland solved this problem at the level of "iron". The system is a set of two devices - a "photon gun", which scientists fired at the floor and wall, located on the opposite side of the corner, and a special light-sensitive matrix based on avalanche photodiodes that recognize even single photons and convert light into an electrical signal due to the photoelectric effect.

    The photons from the gun beam, reflecting from the surface of the wall and floor, collide and are reflected from the surface of all the objects that are behind the wall. Some of them fall into the detector, reflecting again from the wall, which allows, based on the time of the beam, to determine the position, shape and appearance of what is hiding around the corner.

    This decision brings us one more step closer to a future in which cameras will become truly all-seeing.


    One of the control points of Santa Anita Park , USA.

    Adding more cameras increases the amount of video data collected, but reduces useful information. In other words, the vast majority of recorded video is never viewed. This is not due to the lack of interesting content - rather, because of the gradual decrease in the concentration of attention of a person.

    In the future, people will no longer constantly look at the monitor. Cloud systems will begin to fully control themselves. People will only receive information that really matters to them. In terms of service, a system with 10 thousand cameras will cease to differ from the system of 10 distributed cameras.

    Already, cloud-based video surveillance services allow you to build distributed systems with any number of connected objects and equipment, with all broadcasts from remote sites represented in a single personal account. The user can distribute access rights to hundreds of cameras at the same time through the cloud, change settings, access the archive and connect video analytics modules.

    Autonomous robot with a Knightscope video surveillance camera capable of recognizing license plates and identifying smartphones by MAC and IP addresses

    Perhaps the only point that is difficult to optimize is the installation of systems, but there is still room for progress. Thanks to advances in wireless data transmission, cameras today do not need physical network connections. However, the cameras still need power.

    Existing solutions with autonomous power sources (for example, solar panels) for real wireless systems are not widely used, and they still have their limitations - cost, size and, ultimately, dependence on the Sun itself.

    Today, power technology is being created via Wi-Fi and other wireless solutions - in the future, fully wireless cameras will appear. And let's not forget about robots - commands of autonomous devices can act as a whole. These groups with multiple cameras can be used to observe and collect information, as well as in hazardous work, where human presence is undesirable.

    Robots can be drones, autonomous ground vehicles, or even humanoid devices going through the crowd. If you have robots that interact with each other, you can set any task for them: monitoring, monitoring security, reporting on the situation.

    The VSaaS video surveillance market continues to show strong growth worldwide. Cloud-based surveillance technology has proven to be extremely important, especially for monitoring security and reducing business costs. Therefore, everything is invested in the "clouds" - from small businesses to the government - to give impetus to the development of modern video surveillance devices. Attracting investment to critical infrastructure laid the foundation for a new world in which some may see “threats of total control” according to the “Chinese” model, and others - unique opportunities for the implementation of complex, exciting projects.

    Also popular now: