Advertisement playing system and method based on image identification

[0024] The advertising delivery system based on image recognition of the present invention, such as figure 1 Shown, including:

[0025] The advertisement display platform 100 is used for the display of advertisements; among them, the advertisement display platform 100 includes: TV sets, advertising LED/LCD display screens, personal computers, advertising posters, etc.;

[0026] The image collection system 101 is used to collect images of people (people with different degrees of attention to advertisements), and can transmit the image data obtained after the collected images are analyzed and processed to the network; among them, the image collection system 101 includes: Generally visible light camera, infrared camera, thermal imaging camera, etc.; the image acquisition system 101 may be one or more; in addition, the image acquisition system 101 may also include a camera capable of detecting depth information, where the depth information is Refers to the distance between the imaging object and the camera. This is mainly relative to the normal camera. Usually the camera can only capture flat information. For example, a group of people captured by a universal camera can only see these people from a flat angle. Can analyze the front of the camera by other non-intuitive methods, and the one standing behind (judged by the size of the image of the person or by the way of occlusion), and the camera with depth detection can directly indicate the distance of the object from the camera through the brightness For example, the farther away from the camera, the darker the image, otherwise, the brighter the image; and for the network, wired or wireless methods can be used;

[0027] The data analysis and processing module 102 is used to analyze and process the collected images, and to summarize and overall analyze the data of the image collection system 101 in the network to obtain the index data required by the user. Among them, the content of data aggregation and overall analysis includes: the number of groups interested in advertising products, the number of potential customers, the types of people interested in products, the possible sales volume of regional product promotion, etc.

[0028] In addition, the advertisement delivery system of the present invention may further include: an audience 103 who pays attention to the advertisement and an audience 104 passing by.

[0029] In the advertising delivery system based on image recognition of the present invention, the analysis and processing of images can be processed in the image acquisition system 101 or the data analysis and processing module 102. Among them, the image processing content includes: facial expression recognition, character gender analysis and recognition, character approximate age analysis and recognition, character standing and walking gesture recognition, multiple character target recognition, etc.

[0030] Therefore, compared with the traditional advertising delivery system, the present invention mainly has an additional image collection and analysis mechanism, which can automatically analyze the effect of advertising delivery and information collection and analysis, that is: viewers watch the advertisement in front of the advertising system After the advertisements are displayed on the display platform, the audience’s images are collected by the data acquisition system and undergo preliminary image processing (such as image filtering, compression, etc.). The image information is summarized to the data analysis and processing module through the network, and the image information is collected by the data acquisition system. Analysis and processing of facial expressions, gender, age, and advertising audience statistics.

[0031] Therefore, for the advertisement delivery system of the present invention, its implementation method includes the steps:

[0032] (1) By adding the image acquisition system 101 to the advertisement display platform 100, the degree of attention of the advertisement audience to the advertisement is collected in real time;

[0033] (2) Combined with facial expression recognition, the image acquisition system 101 or the data analysis processing module 102 automatically analyzes the advertising audience's feedback on the advertisement;

[0034] (3) Combined with the analysis and recognition of human body characteristics, the data analysis and processing module 102 automatically analyzes and estimates the age structure and gender of the customer group to provide a reference for subsequent advertising.

[0035] Among them, the specific implementation steps of the above method, such as figure 2 Shown, including:

[0036] (1) The image acquisition system 101 performs image acquisition, and the image acquisition system 101 around the advertising display platform 100 captures an image of a certain area facing the advertising display platform 100;

[0037] (2) Analyze and identify whether there is a new audience (person) entering the image acquisition area or an audience leaving from the image acquisition area, if not, analyze whether there is an audience in the image area, if not, go back to step (1);

[0038] If an audience enters the image collection area, execute step (3), if an audience leaves the area, execute step (5);

[0039] (3) Analyze the characteristics of the new audience and establish an ID number for the audience to provide reference for the subsequent analysis;

[0040] (4) According to the ID established above, track the specific actions of the viewer within the image collection range, analyze the age and gender of the viewer, identify the facial expression of the viewer, and count the time that the viewer watches the advertisement, etc.;

[0041] (5) If it is detected that the audience is leaving from the image collection area, the system records the ID number of the leaving person for subsequent analysis;

[0042] (6) Whether the audience with the corresponding ID number has paid attention to the advertisement information during the process of searching, if not, just pass through the image collection area in front of the advertising system and count the ID into the potential advertising audience; if you have followed the advertisement, get the corresponding ID audience’s facial expression information, analyze whether they are interested in the advertisement;

[0043] (7) Accumulate the number of people interested in the advertisement, and analyze the age structure and gender of the people interested in the advertisement through physical characteristics.

[0044] The specific steps of the above method are described in detail as follows:

[0045] 1) First, the recognition and analysis of the overall posture of the human body

[0046] When a person enters the image collection area, the system will establish an ID number for the person based on the general characteristics of the human body. This general feature includes the height of the human body, fat and thin, style and color of clothing, length of hair, color and hairstyle, skin color, etc. (the face setting will not be involved here, because a considerable number of people may not see Advertising, the system cannot get the positive face information). There are many detection methods for height, fat and thinness, and hair length. Here are two examples: a. Through a camera with depth detection function or multi-camera collaborative work (you can test the specific distance of the person) to obtain the distance information, combined with the human body The height, short, fat, and thin size of the planar imaging is calculated to calculate the height information of the actual person; b. By setting a certain reference icon in the image acquisition area (for example, draw a zebra-like stripe on the ground, according to the audience standing on which stripe Determine the distance from the advertising screen), analyze the distance between the audience and the advertising screen, and then calculate the height of the actual characters. The colors of clothes, hair, and skin are directly analyzed based on the color of the camera's imaging. Considering that this information is only used to distinguish people in the image collection area, just look at the color contrast of different people, and the system environment light source is not too demanding. harsh. Certain samples will be collected before the audience ID number is established for the styles of clothes and hair and hairstyles. When creating the ID number, the samples are matched with the ID number, and follow-up tracking of the person is provided as a reference (in order to improve the accuracy of person tracking, during the tracking process, When the character's orientation changes, update the sample library corresponding to the ID number).

[0047] 2) Secondly, the analysis of the human body's expression

[0048] After obtaining the ID number of each person in the image capture area, when someone starts to look at the advertisement screen, the image capture system 101 can capture the face picture of the person. At this time, in order to determine the user's interest in the advertisement, it is necessary to The facial expressions of the characters are recognized.

[0049] The principle of facial expression recognition is relatively simple. It mainly analyzes the changes in the emotions of the characters according to the changes in the characteristics of the mouth, eyes, eyebrows and cheeks. At present, many institutions have developed face-related expression recognition methods. The common method is to compare and analyze facial expression classes to obtain the weight of which type of expression the facial expression belongs to. In this system, secondary statistics will be performed based on these weights and the classification of expressions in the emotion sample library, and the following three results are summarized: interesting attention, no special expressions, boredom and disgust. For example, when the weights of expressions such as happy smile, surprise, and curiosity account for a large proportion (more than other expressions), the statistics are that the audience is paying attention in an interesting way.

[0050] 3) Finally, it is the age judgment and gender judgment of the characters.

[0051] The basis of this judgment is similar to face recognition, which is mainly analyzed through the comprehensive feature library of the human body. For example, gender recognition requires comprehensive judgments based on the collection of facial features, the length and hairstyle of hair, the style and color of clothes, and shoes to obtain corresponding results. Judgment of age is more difficult, mainly through comprehensive judgments such as facial features, hair styles, clothes, styles and colors of shoes, etc. In order to improve the accuracy of judgment, it is also necessary to combine the actions of characters to assist judgment.

[0052] Through the above judgments, statistics and classification, the audience information of advertisements can be quickly transmitted to advertisers and advertised product suppliers, and the effect of advertising and the value of advertising investment can be analyzed more accurately, so that relevant personnel can timely Adjusting the advertising strategy has a greater role in promoting product promotion.