En Ru Es
Send request and get quotations for your project.
Send request and get quotations for your project.

Artyom Razumkov, CEO of Macroscop: Suspect tracking as a new word in the world of video archiving

Artem Razumkov, the CEO of Macroscop, professional IP camera software manufacturer, talks to us in this interview about a new approach to Suspect inter-camera tracking and a new function for this technology – the indexing of video based on special features.

The suspect inter-camera tracking: the essence of technology

Artem Razumkov: For many years our company has been developing technology which allows one to index videos based on the special features of subjects. 

Using this technology, you can search video archives based on different criteria. However, Macroscop conducted research in this area and found out that these functions are used rather rarely and much less frequently than we would like.  Currently, they are not very popular. Based on feedback from our customers, we can say that they have little need to search a video on a daily basis for, say, a person dressed in a red jacket and blue jeans heading in a certain direction. And for this reason, our customers don’t rush out to find out how the function works and how to prepare a query for a particular video searching task. 

The next challenge we tried to address was how to use indexing technologies based on special features in order to achieve certain goals which our users would more often like to achieve. In our research, we discovered that our users don’t just need to pick out a certain person on single camera and to trace his route, but to trace his route as he passes from one camera to another, and so on. Often our customers need to know where a subject appears in a video, where it goes after that, and where it exits. In other words, it’s necessary to trace the whole path from the beginning to the very end and to be able to replay the whole chain of events associated with a subject in a video and to put together the route of their movement.

The less routine, the faster results

Users who work with video surveillance systems and use video achieves most often need to use the video indexing function. Traditionally (without using our indexing with special features), the process is as follows: An operator looks at an image on one camera, searches for a suspected person, browses the video archive, and checks to see whether or not this person appears on any other camera. Normally this process takes a lot of time. 

And that is how the process appears when the new video indexing technology is employed. It is very simple: the camera captures the appearance in a video of a specific person, and the operator is able, by one click of a button, to select automatically all the appearances of this person from the cameras’ viewpoints.  It is then possible to track all of this subject’s movements easily:  where they came from, and when, where, and how they moved. Within a particular surveillance system, the indexing technology is able to locate people with similar features who were within the cameras’ viewpoints during or within a specific period of time. Usually, the system finds a lot of different people with similar features and presents them in the form of snapshots. The operator only needs to click on those images and subjects which need to be tracked. Within a minute or so, it’s possible to track all the movements of a particular person, and even to create a step-by-step video of the movements.

In my opinion, this feature can make a great contribution to the process of working with video archives.

It really works well because the approach is completely different

The idea of such an analysis is not new. A similar approach has been used in the marketplace for many years. However, our technology gives us an opportunity to trace in literally three clicks the course of events in a video. This is the first point. 

Secondly, many companies are involved in inter-camera tracking, but they are attempting to make it work in an automatic mode. Their approach is to track objects based not on specific features but based on the proximity of cameras to each other and the fact that the cameras’ viewpoints intersect. Only if cameras are near to one another is it possible to track a subject’s movements. And usually, this approach does not work very effectively. When the system operates in the automatic mode, it "confuses," or misreads objects. 

Our approach is different. The essence of our indexing method rests on the fact that the most difficult task, which is to identify a certain subject, is not carried out automatically, which often results in misinterpretations, but is delegated to a human being who makes this decision. And as I pointed out, the idea is not at all new, but in our case, it really works well because the approach is completely different. A representative of one of the leading companies in the video surveillance market has reviewed our system and said the following: “This is the best inter-camera tracking I've ever seen."

The variety of filters available

Razumkov: The list is rather long. The system allows you to apply multiple filters, such as the size of an object, its proportions, whether it is in a horizontal or vertical position, its direction of movement, the particular location of a subject, and much more. From this standpoint, I can confidently say that our product is rather unique. At the same time, we understand that it is practically impossible to find an absolutely unique product within a modern marketplace. I truly believe that.

Versatility of practical application

Razumkov: Also, there is one more important feature of our technology. It is not geared toward a specific vertical market. A lot of video analysis functions are designed to solve very limited tasks. Any new request will require a complex reconfiguration of the system, installation of additional equipment, etc.

For example, let’s take a look at a very simple task: people counting analytics. Unlike a lot of others, this function works quite reliably for almost all developers on the market. The fine quality of this function is achieved because the equipment is configured for this particular pre-set task.

However, our approach is different, because we never know in advance what a user is going to want from the system. That is why the system’s scope of application is very wide.

In fact, it is like working with a big data.  At first, we collect a large volume of data, then we analyze it and reveal certain patterns within it, and then give the users the opportunity to find answers to their questions. Thus, the same modules can be used for a variety of tasks, which significantly expands the scope of their application. It all depends on the tasks. Quite often it leads to a situation where a certain module was initially designed to work for security needs. However, it ends up being used for other tasks as well, which have nothing to do with security. And we have to be ready for that as well.

This interview was first published in Ru-bezh magazine

Download free Demo version now to check out how it works! 

We use cookies on this site to ensure the best service possible. Read more