What types of detection are useful in video surveillance. Mechanisms and functions


    A notification signal when motion is detected (detection) in the field of view of the camera is a basic function, without which it is impossible to imagine a modern video surveillance system. However, even this simple and user-friendly function has many nuances that affect the cost and quality of the entire system.

    Today we will introduce you to the various types of detection used in CCTV cameras, talk about their advantages and disadvantages, compare in detail the devices integrated with the cloud, and with analytics “on board”.

    We offer several solutions that are based on algorithms for determining the object and motion in the frame:

    From this list, only Ivideon Counter 3D is presented as a separate device that allows you to count store visitors and create analytical reports: by traffic (weekly and total) and hourly load, as well as note a large flow of visitors on the timeline under the video.

    Motion Detector- This is part of the video analytics system available for all cameras of the Ivideon service. It determines the movement in the selected area, and if the user did not specify it, then the detector works throughout the frame zone. The detector is based on an algorithm that creates a background image and compares subsequent frames with the background. Thus, it is possible to avoid false positives with a small change in the frame. For example, the detector will not work if a bird flies outside the window. But it will work if a motionless sitting person begins to move his hand. In addition to comparing frames with the background, additional motion detection algorithms are used to increase the quality of work.

    Queue DetectorIs a video analytics module that is used to detect a queue based on counting goals in a frame. The detector was developed using machine learning, that is, its accuracy increased as it was used.

    The detector is useful for owners of medium-sized businesses or network retailers, where there is a high risk of queuing. The analytical module generates reports, groups the queues according to the zones of occurrence and shows segments of video from cameras on which the moment of queue formation is visible. Reports are available for viewing in your account, and can be exported in .csv format.

    Face recognition- video analytics technology, which is most often needed by large companies in the banking and insurance sectors, where there are high requirements for security and observance of official secrets. We wrote more about it here (link to an article about emotions).

    All these technologies appeared after a long evolution, which began with analog detectors.

    Analog External Motion Detector

    A passive infrared sensor with a pyroelectric detects infrared radiation levels.

    Earlier on the market for video surveillance systems, analog cameras with a standard resolution dominated, as a rule, 576 p. To activate recording when motion was detected, separate external sensors were used.

    Heat-sensitive infrared sensors were installed in them, which reacted to the appearance of a heat source (person) in the control zone and determined the fact of its movement. The signal from the sensor activated the recording of the image from the camera to the cassette of the VCR or the alarm in the security system.
    Standard IR motion sensor

    This technology is used today due to its low cost, high sensitivity and the ability to work with cheap analog cameras, but all the advantages are offset by serious problems using an analog detector.

    Firstly, its sensor covers with its “watchful eye” the entire protected area. You have no way to close a certain sector from the detector, in which, for example, there may be movement of warm air masses. In this case, you will encounter periodic false positives, and the protection of the object will receive a headache.

    Of course, you can try to minimize false alarms by adjusting the sensitivity of the detector and changing the viewing angle, but this is not always the best solution.

    Secondly, various atmospheric phenomena can lead to false alarms of the sensor: snow, rain or dust. It is also quite sensitive to artificial interference.

    Determine the activity in the frame

    Multiplexer for working with analog cameras ( c )

    With the development of digital video processing technologies, it became possible to integrate activity detectors into multiplexers, although the cameras themselves were still analog at that stage.

    The task of the multiplexer is to reduce the video stream from several cameras into a single video sequence for demonstration on security monitors and recording on a Time Lapse video recorder.

    That period can be considered a turning point in the development of detection technology. And there is no typo: we are talking about recognizing activity in specially marked areas of the frame - this operation was performed by the multiplexer.

    It could be programmed to determine activity in a video sequence coming from one or several cameras. For this purpose, the frame was divided into zones (as a rule, 256 zones of 16x16 squares), and in the process of setting up the system, a specialist could choose which zones to analyze activity in and in which not.

    Unlike the analog motion sensor, an alarm signal was generated when there was activity only in the selected areas, and not over the entire frame area. This approach provided greater flexibility and ease of control, minimizing false alarms of the system.

    Another advantage of this technology can be considered the ability to activate recording with a standard frame rate in the presence of activity on the camera. In the multiplexer, it was possible to configure the recording of the video stream from a specific camera in full-frame mode, so that later it would receive normal dynamic video without jerking and frame skipping, with a smooth transfer of the motion effect. Video from other cameras continued to be recorded at a lower frame rate.

    The technology has significant advantages over analogue motion detection. First of all, fewer false positives due to the flexible configuration of control zones. Such a system was practically not affected by precipitation.

    Progress does not stand still, and standard Full HD analog cameras have been replaced by digital Full HD cameras, and analog multiplexers and Time Lapse video recorders have been replaced by digital recorders, as well as local and cloud video servers. All this led to the fact that users of digital video surveillance systems received additional features, which we will tell you about below.

    Camera activity detector

    ( c )

    The advent of powerful DSPs (digital signal processors) for video processing has enabled developers to integrate motion detectors (and not only them) directly into the camera itself. We will call such cameras "smart", or "cameras with analytics."

    You can select the desired zone in the frame, the activity in which you want to track, and at the same time, as a rule, you are not limited to squares of a certain size. Smart cameras allow you to set zones of any size and shape and activate recording, for example, on an installed flash drive. And of course, they can transmit an alarm.

    Smart cameras are not limited to determining the presence of movement in the frame. More advanced models can send notifications about various events at the facility, for example, by e-mail.

    Some cameras have the ability to set the time interval for motion tracking. And of course, most modern cameras with analytics also highlight moving objects in the frame that the operator needs to pay attention to.

    Note that now technologies that were previously inaccessible to most surveillance systems are gaining popularity:

    • face recognition;
    • recognition of license plates and other objects;
    • recognition of emotions.

    If your surveillance system needs to automatically identify or recognize people and objects, then you need to pay great attention not only to choosing a camera. A very important role is also played by its location and lighting at the facility. For example, the presence of backlight can significantly reduce recognition accuracy, especially if your camera does not have a backlight compensator or is not active.

    It is also necessary to take into account the size that the recognized object occupies in the frame. According to the requirements of the European standard EN 50 132-7, for the detection, recognition and identification of people and objects, the concept of "pixel density", or the number of image pixels per 1 m of horizontal distance to the observed target, is introduced.

    These requirements are shown in the table.
    Type of activityChallenges and OpportunitiesOld parameter (% of height)Alternative parameter mm / pixThe number of pixels per 1 m horizontally (for reference)
    MonitoringCrowd monitoring and control5% of frame height8012
    DetectionGuaranteed detection of people in the frame10 % от высоты кадра4025
    НаблюдениеОпределение характеристик и особенностей человека, например одежды25 % от высоты кадра1662
    РаспознаваниеРаспознавание известных оператору людей50 % от высоты кадра8125
    ИдентификацияКачество достаточное для идентификации человека100 % от высоты экрана4250
    ИнспектированиеВозможность 100% идентификации, исключающей сомнения400 % от высоты экрана11000

    As you can see, for reliable detection of people in the frame, a density of at least 40 mm / pixel is required. It is clear that recognizing people is a rather difficult task, therefore, the requirements for image quality are very high.

    And if we take, for example, the problem of recognizing car numbers, then for this the height of license plates and letters should be approximately 15 pixels, which is equivalent to 200 mm / pixel.

    In addition to high-quality images, the most important role in any detection is played by the algorithms used.
    External and built-in

    detection. Detection programs can be integrated into the camera or located on a remote video or cloud server.

    Consider the advantages and disadvantages of these approaches.

    Advantages of built-in detection:
    • the user immediately receives a ready-made video surveillance system with detection. He does not need to think about how it works and which detection algorithm to choose, the camera manufacturer did everything for him;
    • so the second, albeit controversial, advantage of this approach is the ease of installation and configuration of equipment. Why this plus is controversial, we will describe in the chapter on cloud technology for motion detection.

    This is where the benefits end.

    Let's talk about the disadvantages of built-in detection.

    Any function built into the camera increases its cost. After all, the developer needs to create the appropriate software and debug and test it. And the buyer must pay for all this.

    As a rule, most of the detection algorithms built into the camera are quite simplified. After all, it’s easier for the manufacturer to write the software once and sell it for as long as possible.

    Fullhan 8520 processor in a Dahua camera ( c )

    High-quality detection algorithms require powerful DSPs to operate, and the more powerful the processor, the more expensive the camera will be. It turns out that the buyer is actually “tied” to a strictly defined detection algorithm wired into the camera, which may not be very effective, and the manufacturer is in no hurry to release an update. Or he can’t always do this in the case of the old hardware. It is easier for the manufacturer to release a new, more expensive camera than to spend time and money on finalizing the old one.

    When planning to buy a camera, especially from a manufacturer from a low price segment, think about whether you need "features" in the form of built-in detection. And now let's explain why.

    Advantages and disadvantages of external detection


    The algorithms will analyze the video stream sent to the cloud or local server and give the result. This option is implemented in Ivideon.

    Unlike the motion detection algorithm wired into the camera, work on improving the technologies used in our cloud does not stop. In particular, we recently updated the settings of the motion detector in the personal account of the service.

    The user can make all the settings in the office remotely from a PC or any gadget. Now we have added the "Detection" section, in which you can adjust the zone and sensitivity of the motion detector. There are similar settings for the sound path.

    The motion detector setup is simple and intuitive: the user does not need to think about how effective the detection algorithms are and how to update them. Our programmers deal with these issues.

    Using a cloud-based video surveillance service makes it easy to organize any detection, regardless of the type and price of cameras installed at your facility. You can set up motion detection or activate face recognition. You do not need to purchase new cameras, just select the desired option in your personal account service.

    Also popular now: