Video analytics and machine vision – from theory to practice

0
226

Video surveillance and video analytics are used today not only for security purposes. Increasingly, they are positioned as modern and effective tools for business, while along with traditional IP cameras, machine vision systems are used to solve a certain range of tasks.Technologies, which until recently seemed expensive and complex, are gradually becoming more widespread and accessible. Vision cameras and video analytics are already widely integrated into a variety of applications, and their applications extend beyond niche, specialized segments. We have already written about what machine vision cameras are and how they differ from CCTV cameras.

MACHINE VISION TO THE MASS

Inexpensive machine vision cameras can significantly improve video surveillance in transport, trade and in a number of other areas. According to Lars Brinkmann, Basler’s Regional Sales Manager, the new CMOS technology allows for better image quality and increased system throughput. CCD technology is gradually leaving the market: 74% of Basler products use CMOS sensors from Sony (Ace cameras) and ON Semiconductor (Python) with resolutions up to 5.3 megapixels @ 60 fps with USB 3.0.

The company offers its customers platforms with embedded software for the development of various vertical solutions – in particular, for the needs of medicine, biology, retail, for use in ATMs and in transport. Oftentimes, such applications require the identification of the smallest details, and machine vision cameras come to the rescue. Moreover, the cost of such a solution, including a printed circuit board, a camera and machine vision lens for it, sometimes does not exceed 200 euros.

INTELLIGENT VIDEO SURVEILLANCE: DESIGN PRINCIPLES

The new generation of IP cameras with HD and higher resolution, high frame rates and high-speed interfaces places serious demands on data transmission and processing systems. And to solve computer vision problems, such as computer graphics (in the film industry), video surveillance, medical research or surgery, and others, it is necessary to work with uncompressed data received directly from an machine vision lens. As a rule, processing such a volume of information requires significant computing resources, but, as noted by Mikhail Lepeshkin, director of advanced developments at Qoollo, other approaches are also possible.

In some cases, the problems of transmitting and receiving large amounts of data are solved by localizing video processing and load balancing, and storage of large amounts of heterogeneous data is carried out using specialized hybrid storages that support working with images and video, structured and metadata.

When designing security systems, one has to take into account an impressive list of threats and applicable regulatory documents, says Aleksey Sofiyskiy, head of the engineering department at Autonomous Systems. Video monitoring assumes both work with the archive and operational video surveillance.

When designing such systems, it is necessary to agree with all users of the facility a plan for placing cameras and viewing areas, draw up a work schedule, provide for maintenance of systems, analyze the conditions of their operation over a long period of operation, ensure tracking of the timing of the implementation of engineering infrastructure, and also allocate a place for installation active equipment. Often, structures that obstruct the view are subsequently mounted at large transport facilities, and the likelihood of such an event should be taken into account when designing any video system.

Usually, with the help of the video cognition system, it is possible to quickly recognize crowds of people, as well as to identify and record the facts of their inadequate movement (fights, falls, chaotic movement, sudden acceleration, crossing the forbidden zone), to find left or disappeared objects in the controlled area, to reveal smoke or open the fire.

At the same time, it is important to understand what tasks the customer needs to solve in order to help him determine realistic goals, since the capabilities of video detectors are usually overestimated. It is often helpful to conduct a feasibility study for the implementation of each subsystem. An indispensable condition is pilot testing of the system in real-world conditions.

When choosing a video analytics system, one should be guided by the terms of its use, the type of calculator used and the type of video analytics itself. The functions of the computer, the “brain” of the system, can be transferred to the camera, to industrial units or servers. Each of the options has its own advantages and disadvantages. Cheaper solutions are usually functionally weaker.

In real conditions, video analytics often provide completely different results than the manufacturer promises. For its effective use, the shooting angle is very important, as well as the stability of the background and lighting – the appearance of shadows and flares. Depending on these conditions, different detectors and algorithms behave differently.

Video analytics is usually based on background subtraction, motion detection and subsequent classification of the object (in this case, the correct use of a scale grid is important) or on machine vision algorithms using trained classifiers (formation of classifying features). Both options have their drawbacks. The first one is more common – such a solution is cheaper. Machine vision algorithms are more accurate, but they require more computing resources.

An interesting perspective is opening up when using machine vision cameras to classify and count visitors in retail. According to Markus Müller, video expert in retail at Basler, they can be used to optimize sales based on the statistics they collect – visitor recognition and counting, classification and tracking, and heat mapping. Another area of ​​application of machine vision cameras is face recognition.

VIDEO SURVEILLANCE AND CLOUDS

Cloud services can also be used to solve video analytics problems. For example, Smarttec  Security offers the uCountit service  on the Amazon retail platform. It consolidates metadata collected from customer-installed Smartec Neyro IP cameras and meters and then processes them using VCA Technology video analytics. With its help, you can generate a variety of reports (statistics and graphs of visits for a specified period), build heat maps and even view “live video” transmitted from a point of sale.

Video surveillance services based on the service model are excellent for small and medium-sized businesses, says Maxim Garusev, director of the marketing department at MGTS. According to MGTS, up to 15% of small and medium-sized enterprises are ready to use cloud video surveillance services, although 85% of such clients have already installed certain solutions. And the number of potential customers from the SMB segment is growing. There are several reasons: frequent moves from office to office due to changes in rental rates, unwillingness to bear capital costs for their own video surveillance, lack of trained specialists for the installation  and operation of their video surveillance system.

Recently, more and more companies are paying attention not only to the security function of video surveillance – they are beginning to perceive video surveillance as an additional tool for carrying out their activities and monitoring the state of the business. Needs gradually began to change: from an autonomous software and hardware complex, video surveillance is turning  into a cloud service that provides access to viewing images from video cameras from anywhere in the world, providing opportunities for role-based access to such viewing and archive recordings, and in addition, guaranteeing the safety archive records.

The main consumers of the video surveillance service from MGTS are small businesses from the service sector: law firms, beauty salons, construction organizations, cafes and restaurants, shops. In this case, video surveillance does not so much perform a security function as it becomes an effective tool for business. Cameras, of course, help to strengthen security measures, but at the same time they provide surveillance of the trading floor and the checkout area, the room for receiving  visitors, the area for unloading goods, the work of staff – that is, they allow you to monitor compliance with labor discipline and the level of service provided.

Installation of cameras above cash registers and in retail premises allows detecting emergency situations and carrying out their analysis. In the warehouse and in production halls, you can monitor the movement of goods and the employment of workers. With the help of cameras, it is convenient to observe the entire territory of the enterprise, even if there are several objects located in different places.

In this way, video surveillance service providers are able to offer small businesses attractive solutions to the problems they face most frequently today. However, obstacles remain: the high cost of storing video and the significant load on data transmission channels in the case of streaming HD video. It is possible to somewhat reduce the cost of storage with the help of special configuration disk arrays, optimized for video archives.

The operator’s service involves not only maintenance, but also the installation of equipment. And the presence of a project library, the use of standard elements for fast project design, and the proven practice of network surveys simplify system deployment.

Over time, users have a need for additional services that require sophisticated video analytics, for example, for the organization of passive protection of the territory (virtual perimeter, penetration into a controlled area, message about movement in the frame on a schedule) or when using video analytics as a marketing tool that allows manage sales (counter of visitors, heat and kinetic maps, control of empty counters, opening and closing of points of sale).

As a result of the development of such  services, video surveillance and video analytics are becoming modern  and effective tools  for business. The installation of cameras helps to increase the level of security  in retail and other premises,  and also allows you to strengthen control  over the work of employees and improve service.