Video surveillance system operating modes.
Scientific and technical progress, marked by gigabytes and gigahertz, is marching victoriously across the security systems industry. However, sometimes you hear that not only was the water wetter before, but video surveillance systems were also more convenient. More convenient – everyone certainly admits that there are more megapixels and real-time recording of hundreds of video cameras has only become possible now, but many things were more convenient before. Ancient functional capabilities, known since the times of recording on magnetic tape and surveillance on several CRT monitors, were developed for several decades. Now, video surveillance equipment is often made by completely different people – computer specialists who are often unfamiliar with these old techniques, and therefore there are more and more gigahertz, but it is not always more convenient to work.
In this article, I will try to list the most important algorithmic and organizational techniques that were used 20 years ago and evaluate their usefulness in modern technology.
Recording
Previously, the recording was done on a 3-hour video tape, so, naturally, the question arose: what to do when the tape runs out. Naturally, the answer arose — rewind and write again. And while the tape is rewinding — write to the backup tape recorder. Now you don't need to rewind anything, but the fact of cyclic rewriting (with the deletion of the oldest records) is not implemented everywhere. Many developers sincerely advise to install a larger disk so that it does not overflow. Or regularly clean the disk manually from unnecessary records. They probably do not know that security systems are often left unattended after the ceremonial launch and they work like this for years, right up until the day when the rooster pecks in a certain place, and then the boss comes running and starts frantically looking for the recording from yesterday.
However, it is not that simple. Even in analog systems with a tape recorder, it was understood that it is impossible to mindlessly re-record over important recordings. The simplest solution: if an alarm occurred during the recording of the tape, the re-recording was canceled, and the tape recorder waited until the operator manually reset the alarm. In this way, detectives arriving at the site could see the recording made during the robbery, and not an interesting video about how the detectives arrived. In advanced systems, a separate tape recorder was even installed, which was used to record only in the event of an alarm. In this way, especially important fragments of the recording were saved separately. In modern systems, separate settings are sometimes also found — the amount of disk space for regular and alarm recordings. Just do not forget that the volume of alarm recordings must be sufficient for recording until a specialist arrives who is able to copy all the most interesting things to a removable media. Or automatic re-recording of alarm fragments must be prohibited.
The recording quality in old systems left much to be desired. A VHS tape recorder, especially with a sparse frame rate, is even worse than old silent film tapes. Of course, in the event of an alarm, the quality (at least the frame rate) of alarm video cameras was increased. Nowadays, there is little need to increase the quality, often recording is always in the highest quality, but in many cases it makes sense to broadcast alarm video in high quality for backup recording in a central storage, and not just record locally. At the same time, there are even more problems nowadays: previously, it was impossible to exceed 50 frames per second when recording to a tape recorder. Now, several video streams, simultaneously and sharply increasing their quality (frame rate, compression quality, pixel resolution), can easily overload a file server or network, as a result of which none of these especially important videos will be recorded.
In older systems, the problem of limited recording capabilities for multiple alarm cameras simultaneously was so obvious that several algorithms were developed. For example, each alarm camera is given a minimum guaranteed time of high-quality recording after an alarm. And then, if a large number of cameras remain in the «alarm» state, recording is not done uniformly from all cameras (i.e., all cameras are equally bad), but in turn, allocating higher quality for each camera. Since with higher quality in one channel, the recording quality of the other cameras decreases slightly, thus, there is a chance to improve at least something.
Viewing a recording
There are currently many cutting-edge solutions in this area with user-friendly user interfaces based on computers, touchscreens and flying translucent windows. But we should not forget about old but useful solutions.
The main action when viewing is searching for the next alarm. The senior operator who comes on duty in the morning must view all alarms for the night and decide whether it was shadows running on the ceiling or someone trying to break into the protected facility. Therefore, he quickly watches the recording of each alarm and, if there is nothing interesting, moves on to the next one. It is very inconvenient if, in order to move to the next one, you need to call up several menus and manually select the recording following the one you are viewing. Especially if you need to remember and not forget the name of the one you are currently viewing. The «next alarm» button is the main interface element when viewing a recording.
Of course, there should be the ability to view in fast and slow motion and view a still frame. It was not easy to implement a high-quality still frame on magnetic tape. Do you think it is easier to do now? In most modern super-duper-sophisticated recorders, you cannot stop recording anywhere — only on one of the reference frames, which can go even less often than once a second. Theoretically, it is possible to stop on an intermediate frame, although it will have a slightly reduced quality. But in practice, not everyone can do this. Or they simply do not consider it important. So you will jump every second during frame-by-frame viewing. In one frame, the suitcase is there, and in the next, it is not. And who took it — go figure. Another old feature that some computer system developers are now so proud of is searching for changes or movement when viewing a recording. Of course, on analog magnetic tape, the capabilities were limited. But fast forwarding up to a sharp change in illumination in the frame was possible in many systems. And some even allowed using a full-featured motion detector during fast forwarding, with settings for what, where, and where should move in order to stop fast forwarding. Alas, gigahertz and megapixels have not yet made this option available on every single device.
Often, when viewing, you want to watch several channels (several video cameras) at the same time. This is, of course, possible. The only problem is that synchronous viewing in fast or slow motion of several video channels is not so easy to implement now, with the widespread introduction of MPEG4, H.264 and similar very good algorithms. Some do it. But earlier, this was automatically possible on any most primitive multiplexer.
Observation
There are many useful functions when monitoring in real time (and when viewing a recording). For example, if you are monitoring several video cameras on one or several adjacent monitors, you need to equalize their brightness and contrast. A typical situation: several outdoor cameras at night, several indoor cameras in brightly lit rooms. If you do not take measures, the eye simply will not perceive details in one of the channels. Usually the eye will adjust to a bright screen, and a dark one will seem black. Yes, of course, all current systems allow you to adjust the brightness and contrast. But they usually suggest doing this manually after any switch in the set of viewed video channels.
Modern TVs have a function of automatic brightness correction when the lighting in the room changes. Some video monitors in the past also had such a function. Some had the ability to save several sets and switch between sets, for example, «day-night» with one button on all monitors at once. But how often have you seen such a function in computer monitors used today for viewing and surveillance? And what should the operator do if the lighting suddenly switches to emergency mode (practically goes out) and all the screens are glaringly bright? Or, conversely, in the middle of a night shift in the pleasant semi-darkness, there is suddenly an alarm, full light, everyone is on guard, and the screens seem like black windows with unclear spots on them.
Number of channels/windows/screens
As electronics became cheaper, systems with 20-30 cameras began to be considered small, and systems with 200-400 cameras have long ceased to surprise anyone. However, I am surprised that many still suggest installing 1 screen for 100 cameras and watching them all in turn, or, on the contrary, recommend installing a wall of 16 monitors and displaying 16 windows on each — admire how beautiful 256 micro-windows look at the same time.
Yes, of course, a video surveillance system is not a TV, there is no operator and director who will carefully display all the most interesting moments of the match on a single screen. But a person, in principle, is not able to glance at more than 3-4, well, maximum 10 windows. In the best case, he will look at them one by one. 30 years ago, in large systems, it was not economic reasons at all that prevented the installation of hundreds of monitors. The experience of organizing the operator service led to the following optimal ratio: 1-2 large monitors (spot monitors in English terminology) for examining details, conveniently located right in front of the operator. In particular, it is on one of them that the image will be when the operator controls the rotating camera. In addition, several (up to 10) small monitors (alarm monitors), which in a normal situation are completely black and do not attract attention, and in the event of an alarm they show a queue of alarm cameras that need to be assessed by the operator. With the spread of multiplexers capable of providing multi-window display, central monitors have sometimes been used to temporarily display many windows (essentially icons) at once, making it easier for the operator to select the desired camera.
By the way, about PTZ cameras. On primitive matrix switches it was easy to implement a variant when, when an alarm camera (fixed) was displayed on one spot monitor, a suitable PTZ camera was simultaneously displayed on the second one, aimed at a pre-programmed position, as similar as possible to the picture from the fixed camera. If you want to examine something in more detail, grab a joystick and control the camera on the adjacent monitor.
However, one thing was difficult to implement at that time. If one PTZ camera monitors several sectors in which an alarm occurred, then it would have to be queued several times in the alarm queue, and in the small pictures (alarm preview) it would be worth showing a freeze frame of this angle, and not the current signal from this camera, which has already turned to another alarm.
Modern graphical interfaces very successfully use multi-window display technology to select by direct indication (on a touchscreen monitor — by actually directly poking the screen with your finger) the camera that you want to examine more closely. The disadvantage of most modern systems is that surprisingly few of them are capable of working in a coordinated manner with several displays on one workstation. Most systems, even network and multi-operator ones, have grown out of classic computer operating systems, which assume that one operator has one, maximum two displays.
Let me remind you that the concept of an alarm message queue (alarm cameras in terms of video surveillance) has long been established. If the system claims to be integrated, the alarm queue is the main operator interface. The operator must explicitly manually reset each alarm (confirm the perception of this alarm). In specialized video surveillance equipment, such as multi-channel video recorders, this is also easily implemented. But in the most modern distributed systems with web cameras and a web interface, everything is not so rosy. In most cases, the operator interface is located on one computer and can hardly simultaneously display thumbnail images from several web cameras for preview. Web cameras cannot send a very small stream for preview, and it is impossible to receive a high-quality image from many cameras at once and then locally reduce the image — even gigabit Ethernet will hardly pass 50 channels of D1 video format. Many systems do not mention what will happen if 20-30 video cameras are in the alarm queue at the same time. Or they don’t even provide more than 3–4 icons for this in the operator interface.
But many modern systems, unlike the old ones, allow you to easily change the sequence of viewing alarms. The operator can select the next camera to view (a graphical interface with a mouse or touchscreen helps). Previously, it was inevitable to go through all the alarms in a row one by one.
But the biggest problem with viewing alarm cameras from the queue in the old days was related to the delay. By the time the operator reached the next camera, everything was already calm there. You need to be able to see what was in this camera at the moment of the alarm, and not right now. This is what the first digital cameras were used for — freeze frame recorders. Modern systems can easily rewind the picture on any channel to the alarm, show both the freeze frame and the video, return to the current moment (live video) and again easily jump back in time. More precisely, in principle, they could do it. Moreover, they simply have to do it. With one button, as the main action (just like when viewing the recording archive, the main action is to go to the next alarm). At the beginning of the alarm — at the current moment. At the alarm — at now. Alas, the magnificent, diverse, undoubtedly useful and powerful capabilities of modern systems often clutter the operator interface so much that, in principle, all this can be done, but the number of operator actions necessary to achieve the goal is unacceptably large. Duty operators tend to use one or two buttons. The ones they use every day. Especially in an emergency.
As a joke: the operator of a fire alarm and extinguishing system has two buttons — «reset» (he often presses it when there is a false «fire» signal from dusty sensors) and «cancel powder launch». Both are large, beautiful and have large inscriptions. When someone accidentally turned on the extinguishing start, during the start delay the operator pressed the usual «reset» button 20 times, but never thought to press the «cancel start» button, which he had never needed before. The result — two days of work as a vacuum cleaner.
By the way, when you are watching a live picture in real time and want to go back and take a closer look at what just happened, which button do you press? Pause or rewind? It is obvious to you. In all modern systems this would be very easy to implement. Now remember how many menus you need to enter for this, go to the recording playback mode, select the time (current minus 10 seconds), etc.