Application of pattern recognition theory in the advertising environment
Hi, Habr. He wrote a pseudo-scientific article, then took part in the university conference and published it on one site. The article was written 9 months ago and only now decided to publish it on Habré. Before reading, let me remind you once again: a pseudoscientific article.
The fate of the modern world is such that it cannot exist without advertising. Wherever you look: on the Internet, at public transport stops and on the transport itself, in magazines and newspapers and even in the cinema - there is advertising everywhere. At first glance it may seem that only advertisers benefit from advertising, but advertising also has a positive effect on consumers. People don’t have to search for what they need for hours; all information gets into their own eyes. The brain of a modern person automatically selects the necessary, and all the surplus is filtered by the internal firewall.
The volumes of information are growing, the rhythm of life is increasing, and the question arises: “How to make people see only the right advertisement, and in general, information that is interesting to them?”. The correct solution of this issue favorably affects both the consumer and the seller. For the consumer, the flow of unnecessary information is reduced, freeing up such a precious time nowadays. And the manufacturer receives "one hundred percent return".
A similar idea underlies search results on the Internet, as well as in contextual advertising. On this principle, Google (in Russia, Yandex is engaged in a similar activity) earns millions of dollars daily. It is hard to imagine that for some 10 years a small company has the opportunity to compete with such high-tech business titans as Apple and Microsoft.
')
All this suggests that people actively use the system, both for advertising and for the purchase of goods and services. The system itself processes the stream of information entered by it, and on the basis of these preferences, the person receives what he may be interested in. But there is one “but”: such a model is efficient today only in terms of feedback (the user must report his preferences).
The author proposes to use image recognition for more flexible advertising in everyday life, not related to the Internet. For example, LED street monitors will display brand advertising based on the condition: is this brand currently the most popular among passersby. Terms of advertising can be changed, but the principle will be one:
The system performs dynamic analysis of each frame in real time, identifying labels on the image. To do this, the system constantly refers to the database of labels. It is a database in which samples of labels are stored, with which real objects are compared in a video file frame. The key point in identifying an object is the so-called “functional vectors” - a set of supporting properties of each object: the speed of movement on the video file, color attributes, and shape.
To save resources, the system can periodically sample the frame of a video sequence. To do this, the system will calculate how much time the average person needs to go from the way to the disappearance of the camera from the "field of view". The time will depend on the method and location of the camera.
It should be noted that the installation of equipment is a difficult task [5]. The fact is that the quality of object recognition directly depends on the installation location of the camera and the lighting. It is advisable that the camera be installed in such a way that the person’s chest / back is directed directly at the camera lens.
Based on the above conditions, it would be preferable to use not a video camera, but a camera. The use of the camera will make it possible to identify the labels of passers-by walking a considerable distance from the camera. Since image resolution will be very high, to optimize processing by the system it is necessary to divide them into parts. Each part of the image will be considered a separate task. The simultaneous execution of several tasks significantly increase the speed of the system.
Today, the situation is such that the distribution of this technology is difficult:
Flexible management can be facilitated by a “single database”, which will contain a huge amount of advertising of various brands, on the filling of which will be a large number of advertising agencies. Achieving this condition will eliminate the following consequences:
To effectively use this technology, you must provide advertisers with a graphical interface. With it, they will be able to specify the parameters that will determine the display of their advertising on the screen:
The need for the use of research by commercial organizations is necessary for science. The strongest impetus to the development of science from technology companies was given by government procurement. This is followed by the interest of commercial organizations and government subsidies. Grants do not happen much, especially in such a technical area as pattern recognition.
Taking into account this fact, it should be noted that the bodies of the Ministry of Internal Affairs use foreign pattern recognition systems [5]. Those. even for state structures in our country there is no development of pattern recognition technology that can help in catching criminals. Therefore, this technology should be developed for global commercial use, so that science in our country can be developed, not only by our country, but also with the participation of foreign capital.
In our high-tech century, a breakthrough can be accomplished only by works that are based on the intersection of humanitarian and technical disciplines.
Link to article
PS By posting a link to an article in one of the publics, I read a comment that it would be more appropriate to embed RFID tags in things. How do you like this idea? It seems to me that the tags are bulky, and it is possible to identify the object at close distances.
The fate of the modern world is such that it cannot exist without advertising. Wherever you look: on the Internet, at public transport stops and on the transport itself, in magazines and newspapers and even in the cinema - there is advertising everywhere. At first glance it may seem that only advertisers benefit from advertising, but advertising also has a positive effect on consumers. People don’t have to search for what they need for hours; all information gets into their own eyes. The brain of a modern person automatically selects the necessary, and all the surplus is filtered by the internal firewall.
The volumes of information are growing, the rhythm of life is increasing, and the question arises: “How to make people see only the right advertisement, and in general, information that is interesting to them?”. The correct solution of this issue favorably affects both the consumer and the seller. For the consumer, the flow of unnecessary information is reduced, freeing up such a precious time nowadays. And the manufacturer receives "one hundred percent return".
A similar idea underlies search results on the Internet, as well as in contextual advertising. On this principle, Google (in Russia, Yandex is engaged in a similar activity) earns millions of dollars daily. It is hard to imagine that for some 10 years a small company has the opportunity to compete with such high-tech business titans as Apple and Microsoft.
')
All this suggests that people actively use the system, both for advertising and for the purchase of goods and services. The system itself processes the stream of information entered by it, and on the basis of these preferences, the person receives what he may be interested in. But there is one “but”: such a model is efficient today only in terms of feedback (the user must report his preferences).
Principle of operation
The author proposes to use image recognition for more flexible advertising in everyday life, not related to the Internet. For example, LED street monitors will display brand advertising based on the condition: is this brand currently the most popular among passersby. Terms of advertising can be changed, but the principle will be one:
- the camera scans passers-by, identifying brand labels;
- the system calculates the number of brands according to different characteristics: the number of a certain brand among a passer-by, the total number of identical brands among passers-by, etc .;
- ranks data in a certain way;
- gives the most suitable advertising.
System operation
The system performs dynamic analysis of each frame in real time, identifying labels on the image. To do this, the system constantly refers to the database of labels. It is a database in which samples of labels are stored, with which real objects are compared in a video file frame. The key point in identifying an object is the so-called “functional vectors” - a set of supporting properties of each object: the speed of movement on the video file, color attributes, and shape.
To save resources, the system can periodically sample the frame of a video sequence. To do this, the system will calculate how much time the average person needs to go from the way to the disappearance of the camera from the "field of view". The time will depend on the method and location of the camera.
It should be noted that the installation of equipment is a difficult task [5]. The fact is that the quality of object recognition directly depends on the installation location of the camera and the lighting. It is advisable that the camera be installed in such a way that the person’s chest / back is directed directly at the camera lens.
Based on the above conditions, it would be preferable to use not a video camera, but a camera. The use of the camera will make it possible to identify the labels of passers-by walking a considerable distance from the camera. Since image resolution will be very high, to optimize processing by the system it is necessary to divide them into parts. Each part of the image will be considered a separate task. The simultaneous execution of several tasks significantly increase the speed of the system.
Implementation difficulties
Today, the situation is such that the distribution of this technology is difficult:
- the camera must have a high-resolution matrix (real-time scanning) [6];
- Recognition methods must be of very high quality (they must be able to “see” an object at an angle, seeing only a part of it).
Flexible management can be facilitated by a “single database”, which will contain a huge amount of advertising of various brands, on the filling of which will be a large number of advertising agencies. Achieving this condition will eliminate the following consequences:
- without a “unified database,” the system will not justify the means;
- advertising agencies will be difficult to unite among themselves;
- a monopolist may appear in whose hands advertising companies of a significant number of brands are concentrated.
To effectively use this technology, you must provide advertisers with a graphical interface. With it, they will be able to specify the parameters that will determine the display of their advertising on the screen:
- in which areas of the city can their ads appear on the screen;
- how many people are needed to display ads;
- whether advertising can be displayed in the usual format, without using the technology of “pattern recognition” or in cases when the “number of brand” is zero.
Financing
The need for the use of research by commercial organizations is necessary for science. The strongest impetus to the development of science from technology companies was given by government procurement. This is followed by the interest of commercial organizations and government subsidies. Grants do not happen much, especially in such a technical area as pattern recognition.
Taking into account this fact, it should be noted that the bodies of the Ministry of Internal Affairs use foreign pattern recognition systems [5]. Those. even for state structures in our country there is no development of pattern recognition technology that can help in catching criminals. Therefore, this technology should be developed for global commercial use, so that science in our country can be developed, not only by our country, but also with the participation of foreign capital.
In our high-tech century, a breakthrough can be accomplished only by works that are based on the intersection of humanitarian and technical disciplines.
Link to article
PS By posting a link to an article in one of the publics, I read a comment that it would be more appropriate to embed RFID tags in things. How do you like this idea? It seems to me that the tags are bulky, and it is possible to identify the object at close distances.
Source: https://habr.com/ru/post/203020/
All Articles