Method for verifying and improving offline streaming media advertisement putting effect

Abstract

The invention provides a method for verifying and improving an offline streaming media advertisement putting effect. The method comprises a multimedia advertisement board, an offline data processing unit, an image acquisition unit and a cloud processing unit, the multimedia advertisement board is used for playing a plurality of advertisements stored locally in turn, each advertisement is provided with an independent digital label, and the digital labels are associated with the offline data processing unit to determine the currently played advertisement content; the image acquisition unit is a camera shooting module, the camera shooting module directly faces an area influenced by advertisement putting, the camera shooting module is provided with a specific visual field breadth so as to determine an image acquisition range, and acquired image data is sent to the offline data processing unit for primary processing; the offline data processing unit is in data intercommunication with the cloud processing unit and is used for data analysis; the method has the advantages that the offline advertisement putting effect is fed back in time, the privacy security of offline users is protected, and the marketing efficiency of manufacturers is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of streaming media advertising computers, and in particular to a method for verifying and improving the effect of offline streaming media advertising delivery. Background Art

[0002] Streaming media advertising is divided into two types: online and offline. It is very easy to get feedback from online advertising push, and feedback can be achieved through real interactive operation data such as playback volume, completion volume, and click volume. It is an important means of new media development.

[0003] Offline streaming media ads are generally designed based on large billboards and are usually installed in areas with high traffic, such as subways, buses, supermarkets, etc. Some merchants of mass goods usually place a large number of video ads in these areas to expand the promotion of new products or brands.

[0004] However, although the content of video ads is shot and completed by a professional team, its effect is difficult to predict. Generally, feedback is obtained through market research and online surveys by professional third-party companies. The entire feedback chain is too long, and even the feedback chain is longer if it is evaluated by the conversion rate of the product sold. The demand for advertising effects for fast-moving consumer goods or new products is more urgent, and this feedback is often not matched.

[0005] For example, in the invention patent application with application number CN202110548351.6 and invention name: Information delivery effect attribution method and device, a delivery effect attribution method based on a comparative experiment is provided. However, this method is based on background data analysis and has a lag. It can only be used to analyze the reasons for the delivery effect, but cannot obtain feedback in real time and correct the delivery method.

[0006] For example, the invention patent with application number CN202110826377.2 and invention name: Offline smart screen advertising anti-fraud system and alarm based on machine learning provides a method for preventing advertising fraud based on data labels, but this technical solution can only provide feedback based on the advertising data in the background of the machine, and cannot obtain feedback on the real offline effects.

[0007] Based on the above analysis, there is currently a lack of direct and effective technical means for providing feedback on the effects of offline advertising, which is a technical problem that technical personnel in this field urgently need to solve. Summary of the invention

[0008] The purpose of the present invention is to address the deficiencies of the prior art and thereby provide a method for verifying and improving the effect of offline streaming media advertising, which is used to provide timely feedback on the effect of offline advertising, protect the privacy and security of offline users, and improve the marketing efficiency of manufacturers.

[0009] In order to achieve the above-mentioned purpose, the technical solution adopted by the present invention is: a method for verifying and improving the effect of offline streaming media advertising, including a multimedia advertising board, an offline data processing unit, an image acquisition unit and a cloud processing unit; The multimedia advertising board is used to rotate a number of locally stored advertisements, wherein each advertisement is provided with an independent digital tag, and the digital tag is associated with an offline data processing unit to determine the content of the currently played advertisement; The image acquisition unit is a camera module, and the camera module is facing the area affected by the advertisement placement. The camera module is set with a specific field of view width to determine the range of image acquisition, and sends the acquired image data to the offline data processing unit for preliminary processing; The offline data processing unit communicates data with the cloud processing unit for data analysis; Perform the following steps: Step 1) The offline data processing unit collects the digital tags of the current advertisement and establishes an independent data subset. At the same time, the image acquisition unit sends the image or video data collected during the current advertisement period to the offline data processing unit and stores it in the data subset; Step 2) the offline data processing unit processes the image or video data stored in the data subset; Step 2.1) Using gesture recognition technology, obtain the flow data of people in the current data subset; Step 2.2) using face recognition technology, obtaining the number of faces in the current data subset facing the image acquisition unit; Step 2.3) based on the human flow data analyzed in step 2.1), filter out the human flow that has exceeded the first set time threshold and has not left the current field of view; Step 2.4) based on the face quantity data analyzed in step 2.2), screen out the number of faces whose face data can still be collected after exceeding the second set time threshold; Step 3) The offline data processing unit aggregates the data obtained from step 2.1) to step 2.4) and stores them in the current data subset, deletes the image or video data, binds the digital label of the current advertisement, and uploads it to the cloud processing unit; Step 4) The cloud processing unit evaluates the delivery effect based on the feedback data; Step 4.1) Compare the traffic data collected in step 2.1) and step 2.3), calculate the traffic base of the advertising area, the duration of traffic passing through the advertising area, and the ratio of the two, and compare them with the set standard evaluation data to determine the cost-effectiveness of advertising in the current area; Step 4.2) compares the number of faces collected in step 2.2) and step 2.4), calculates the number of people who watched the ad, the number of people who watched the ad for more than the set time, and the ratio of the two, and compares them with the set standard evaluation data to determine the effect of the current advertising.

[0010] Based on the above, in step 1), the data subset is stored in the cache data; In step 2.1), when reading the image or video data in the data subset, the offline data processing unit first performs mosaic processing on the valid frames of the image data or video data in the entire image domain, and the degree of mosaic processing is: based on the current image resolution, the degree to which the human posture can be distinguished is determined, and then the human posture recognition is performed; In step 2.2), when reading the image or video data in the data subset, the offline data processing unit first performs mosaic processing on the valid frames of the image data or video data in the entire image domain. The degree of mosaic processing is based on the degree to which the face can be distinguished according to the current image resolution, and then the face recognition is performed.

[0011] Based on the above, the camera module is a wide-format camera.

[0012] Based on the above, in step 2.4), when judging whether the collected face data exceeds the time threshold, the pixel coordinate accumulation method is used for statistics, including the following steps: Step 2.4.1) Arrange the acquired images by time, extract valid frames from the acquired video and arrange them by time; Step 2.4.1) Identify the face in the current image or valid frame and mark the pixel range where the face is located in red, record the coordinates of all pixel ranges marked in red, and mark each subsequent image or valid frame in red and record the coordinates; Step 2.4.2) Calculate the number of coordinates of the red-marked area in all images or valid frames. If the number exceeds the set threshold, it is considered that the time threshold has been exceeded.

[0013] Based on the above, in step 2.4.2), the pixels in the red-marked area contain multiple pixels. When the pixel overlap of the red-marked areas in two different images or valid frames exceeds 40%, they are considered to be the same red-marked pixel range. If the number of times the same red-marked pixel range appears in different images or valid frames exceeds the set threshold, it is considered to have exceeded the time threshold.

[0014] The present invention has substantial characteristics and progress compared with the prior art. Specifically, the present invention has the following advantages: 1. Through this self-feedback technology, the advertising effect of offline streaming media advertising, that is, the impact on offline people, can be digitized in real time, thereby achieving real-time and effective verification. After obtaining these data, merchants can adjust advertising strategies in a timely manner, or even change advertising content, to improve the effect of offline advertising.

[0015] 2. Based on the statistics of on-site traffic and the length of time people pass through the delivery area, we can determine the characteristics of the current environment, especially the relationship between traffic and residence time, and then determine the expected cost-effectiveness of advertising in the current environment, and then determine the delivery strategy.

[0016] 3. Based on the face recognition statistics on the scene, it is possible to determine how many people in the current flow of people are facing the direction of the advertisement. At the same time, the time the faces stay in the current direction is also counted, so that the most intuitive advertising appeal, the advertisement entering the crowd's field of vision and other effective data can be obtained, and then the attractiveness of the current advertising content to the offline customer group can be determined. On the one hand, it can be used as a basis for judging the influence of the advertisement itself, and on the other hand, it can be used as a basis for judging the effectiveness of the advertising area.

[0017] 4. In order to protect the personal privacy of offline customer groups and prevent personal image data from being stolen, after the image or video data is collected, it will be stored in the cache data for analysis, and before the analysis, it will be pre-mosaic processed to a degree that can distinguish human posture or can be used for face recognition, thereby avoiding the possibility of human image data being stolen. Since the data is in the cache, it will be deleted immediately after the recognition is completed, avoiding the long-term retention of data and protecting the privacy of the customer group.

[0018] 5. When judging the stay time of a face, in order to solve the problem of excessive amount of computing data, a fuzzy processing method is used to solve this problem. Specifically, when identifying a key frame or image, the face area is marked in red and converted into a coordinate data set, and then accumulated frame by frame or all images. If the accumulated number of data in the same coordinate data set exceeds the set threshold, it is considered that the face stays for more than the preset time. In order to facilitate statistics, fuzzy processing is further performed during accumulation. If the number of overlapping coordinates between data sets exceeds 40%, it can be regarded as the same red-marked pixel range. This can greatly reduce the difficulty of calculation and the amount of data, and reduce the pressure on offline data processing units. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of a method for verifying and improving the effect of offline streaming media advertising in the present invention.

[0020] Figure 2 This is a statistical effect diagram of the human flow after mosaic processing in the present invention.

[0021] Figure 3 This is a statistical effect diagram of the mosaic processing of the human face in the present invention. DETAILED DESCRIPTION

[0022] The technical solution of the present invention is further described in detail below through specific implementation methods.

[0023] like Figure 1-Figure 3 As shown, a method for verifying and improving the effect of offline streaming media advertising includes a multimedia advertising board, an offline data processing unit, an image acquisition unit and a cloud processing unit.

[0024] The multimedia advertising board is used to rotate a number of locally stored advertisements, wherein each advertisement is provided with an independent digital tag, and the digital tag is associated with an offline data processing unit to determine the currently played advertisement content, in order to identify the played advertisement content so as to bind the statistical data of the current advertisement and send it to the background for analysis.

[0025] The image acquisition unit is a camera module. In this embodiment, a wide-format camera is used to ensure that the statistical range width meets the requirements as much as possible. The camera module is facing the area affected by the advertising. The camera module is set with a specific field of view width to determine the range of image acquisition and send the acquired image data to the offline data processing unit for preliminary processing.

[0026] Specifically, the width of the wide-format camera itself can be greater than or equal to the width of the captured field of view, that is, within the width of the wide-format camera, draw the area to be counted, and only perform statistical analysis on the flow of people passing through the area to ensure a unified benchmark for analysis and statistics.

[0027] The offline data processing unit communicates data with the cloud processing unit for data analysis, and places the previous image processing process offline. On the one hand, it reduces the pressure of cloud data processing, and on the other hand, it avoids uploading image data to protect the privacy of offline people.

[0028] Perform the following steps: Step 1) The offline data processing unit collects the digital label of the current advertisement and establishes an independent data subset. At the same time, the image acquisition unit sends the image or video data collected during the current advertising period to the offline data processing unit and stores it in the data subset. The data subset is stored in the cache data, which can be deleted after use to avoid the original image or video data being stored in the hard disk or other effective storage area and being restored to valid data, thereby effectively protecting the data.

[0029] Step 2) The offline data processing unit processes the image or video data stored in the data subset.

[0030] Step 2.1) Use posture recognition technology to obtain the pedestrian flow data in the current data subset; specifically, when reading the image or video data in the data subset, the offline data processing unit first performs mosaic processing on the valid frames of the image data or video data in the entire image domain, and the degree of mosaic processing is: based on the current image resolution, the degree to which the human posture can be distinguished is taken as the standard, and then the human posture is recognized.

[0031] In the mosaic image, the human body is still roughly a block of pixels with a head, body and legs, and is still typical. Therefore, the mosaic image does not affect the success rate of human body recognition. Moreover, since the factors are simplified and the number of pixels is greatly reduced, the computing pressure of the processor can be reduced to a certain extent, thereby improving recognition efficiency. Of course, its most critical function is to protect the safety of crowd images.

[0032] Step 2.2) Use face recognition technology to obtain the number of faces in the current data subset that are facing the image acquisition unit. Specifically, when the offline data processing unit reads the image or video data in the data subset, it first performs mosaic processing on the valid frames of the image data or video data in the entire image domain. The degree of mosaic processing is based on the degree to which faces can be distinguished according to the current image resolution, and then face recognition is performed.

[0033] Similar to step 2.1), in the mosaicked image, the characteristic contours of the human face are still typical, and can roughly include reference objects such as eyeballs, nose, ears, face shape, hair and mouth, and their positional relationships with each other. In addition, due to the reduction in the number of pixels, the computing pressure is reduced and the efficiency is improved.

[0034] Step 2.3) Based on the human flow data analyzed in step 2.1), the human flow that has exceeded the first set time threshold and has not left the current field of view width is screened out.

[0035] This shows the specific characteristics of the current environment. When the flow of people is large but the people stay for a short time, it means that people are less concerned about the advertisements placed in the environment, such as in the aisles of stations, subways, and overpasses with large flow of people. Due to the tight schedule, it is difficult for people to pay attention to video ads. When the flow of people is large and people stay for a long time, it means that the crowd is in a relaxed state, and their attention to the surrounding environment will increase, and their attention to the video content will also increase. This means that the advertising here is cost-effective and easy to be seen, which can be used as a judgment principle for selecting key areas for advertising.

[0036] If the traffic volume is small, the proportion of placement will be reduced.

[0037] Step 2.4) Based on the face quantity data analyzed in step 2.2), the number of faces whose face data can still be collected after exceeding the second set time threshold is screened out.

[0038] Specifically speaking, the scenarios can be roughly divided into two categories. One category is that the advertisements are placed on the side of the direction of crowd movement, that is, when the crowd is walking, their faces are not directly facing the advertising screen, or there is an angle. At this time, the facial data obtained will be more effective and can fully illustrate the attractiveness of the advertising content. The data on the dwell time can further determine whether the quality of the advertising content is high enough, which can be used as a principle for evaluating the quality of advertising content.

[0039] The other type is that the advertisements are placed in the positive direction of the flow of people. People are walking in this direction, so most of the faces are within the recognizable range. In this case, the advertisement content will inevitably enter the person's field of vision. Even if the person's eyes are not focused on the advertisement content at the moment, the advertisement content can still effectively enter the field of vision and form an advertisement impression. By analyzing the facial data and dwell time at this time, an effect similar to that of human flow can be achieved, which can be used to judge the cost-effectiveness of the current advertisement position.

[0040] Since there is a certain amount of computational pressure in determining the dwell time of a face image, in order to reduce the computational pressure on the processor, it is necessary to optimize the statistics of the face dwell time.

[0041] In step 2.4), when judging whether the collected face data exceeds the time threshold, the pixel coordinate accumulation method is used for statistics, including the following steps: Step 2.4.1) Arrange the acquired images by time, extract valid frames from the acquired video and arrange them by time; Step 2.4.1) Identify the face in the current image or valid frame and mark the pixel range where the face is located in red, record the coordinates of all pixel ranges marked in red, and mark each subsequent image or valid frame in red and record the coordinates; Step 2.4.2) Calculate the number of coordinates of the red-marked area in all images or valid frames. If the number exceeds the set threshold, it is considered that the time threshold has been exceeded.

[0042] That is to say, for each frame or each image, the number of red-marked areas is counted once, and the way of counting the number is to collect the coordinate points where they are located to form the number of data about the coordinate points; Then the number of red-marked areas in all images or valid frames is cumulatively calculated, that is, the total number of red-marked areas that appear (with the face facing the advertising screen) during the entire advertising time period. If the total data exceeds the threshold, it is considered to have exceeded the set time threshold.

[0043] The background can be divided into different stay time periods according to different orders of magnitude as a reference for judging the stay time.

[0044] In step 2.4.2), the pixels in the red-marked area contain multiple pixels. When the pixel overlap of the red-marked areas in two different images or valid frames exceeds 40%, they are considered to be the same red-marked pixel range. If the number of times the same red-marked pixel range appears in different images or valid frames exceeds the set threshold, it is considered to have exceeded the time threshold.

[0045] This situation actually solves a special situation, namely the number of people who stop to watch the advertisement. This situation is the best data to reflect the quality of the advertisement, so it needs to be counted independently.

[0046] The principle is that when people stop to watch, the pixel coordinates of the red area will no longer change. When the cumulative number exceeds the set value, it means that people stop to watch the video for a long time, which in turn reflects that the advertising content is more attractive. The duration can be defined as 5-10s.

[0047] Step 3) The offline data processing unit aggregates the data obtained from step 2.1) to step 2.4) and stores them in the current data subset, deletes the image or video data, binds the digital label of the current advertisement, and uploads it to the cloud processing unit.

[0048] Step 4) The cloud processing unit evaluates the delivery effect based on the feedback data.

[0049] Specifically, step 4.1) compares the traffic data collected in step 2.1) and step 2.3), calculates the traffic base of the advertising area, the duration of traffic passing through the advertising area, and the ratio of the two, and compares them with the set standard evaluation data to determine the cost-effectiveness of advertising in the current area.

[0050] Step 4.2) compares the number of faces collected in step 2.2) and step 2.4), calculates the number of people who watched the ad, the number of people who watched the ad for more than the set time, and the ratio of the two, and compares them with the set standard evaluation data to determine the effect of the current advertising.

[0051] By using this solution to manage offline advertising, we can obtain timely feedback on advertising. On the one hand, we can screen out environments with better advertising, and on the other hand, we can judge the quality of advertising through the feedback results, and then make a series of advertising strategies such as modifying the duration, content, and location of advertising.

[0052] Finally, it should be noted that: The above describes in detail the preferred implementation of this patent, but this patent is not limited to the above implementation. Various changes can be made within the knowledge scope of ordinary technicians in this field without departing from the purpose of this patent.

Claims

1. A method for verifying and improving the effect of offline streaming media advertising, characterized by: It includes multimedia advertising board, offline data processing unit, image acquisition unit and cloud processing unit; The multimedia advertising board is used to rotate a number of locally stored advertisements, wherein each advertisement is provided with an independent digital tag, and the digital tag is associated with an offline data processing unit to determine the content of the currently played advertisement; The image acquisition unit is a camera module, and the camera module is facing the area affected by the advertisement placement. The camera module is set with a specific field of view width to determine the range of image acquisition, and sends the acquired image data to the offline data processing unit for preliminary processing; The offline data processing unit communicates data with the cloud processing unit for data analysis; Perform the following steps: Step 1) The offline data processing unit collects the digital tags of the current advertisement and establishes an independent data subset. At the same time, the image acquisition unit sends the image or video data collected during the current advertisement period to the offline data processing unit and stores it in the data subset; Step 2) the offline data processing unit processes the image or video data stored in the data subset; Step 2.1) Using gesture recognition technology, obtain the flow data of people in the current data subset; Step 2.2) using face recognition technology, obtaining the number of faces in the current data subset facing the image acquisition unit; Step 2.3) based on the human flow data analyzed in step 2.1), filter out the human flow that has exceeded the first set time threshold and has not left the current field of view; Step 2.4) based on the face quantity data analyzed in step 2.2), screen out the number of faces whose face data can still be collected after exceeding the second set time threshold; Step 3) The offline data processing unit aggregates the data obtained from step 2.1) to step 2.4) and stores them in the current data subset, deletes the image or video data, binds the digital label of the current advertisement, and uploads it to the cloud processing unit; Step 4) The cloud processing unit evaluates the delivery effect based on the feedback data; Step 4.1) Compare the traffic data collected in step 2.1) and step 2.3), calculate the traffic base of the advertising area, the duration of traffic passing through the advertising area, and the ratio of the two, and compare them with the set standard evaluation data to determine the cost-effectiveness of advertising in the current area; Step 4.2) compares the number of faces collected in step 2.2) and step 2.4), calculates the number of people who watched the ad, the number of people who watched the ad for more than the set time, and the ratio of the two, and compares them with the set standard evaluation data to determine the effect of the current advertising.

2. The method for verifying and improving the effect of offline streaming media advertising according to claim 1, characterized in that: In step 1), the data subset is stored in cache data; In step 2.1), when reading the image or video data in the data subset, the offline data processing unit first performs mosaic processing on the valid frames of the image data or video data in the entire image domain, and the degree of mosaic processing is: based on the current image resolution, the degree to which the human posture can be distinguished is determined, and then the human posture recognition is performed; In step 2.2), when reading the image or video data in the data subset, the offline data processing unit first performs mosaic processing on the valid frames of the image data or video data in the entire image domain. The degree of mosaic processing is based on the degree to which the face can be distinguished according to the current image resolution, and then the face recognition is performed.

3. The method for verifying and improving the effect of offline streaming media advertising according to claim 1 or 2, characterized in that: The camera module is a wide-format camera.

4. The method for verifying and improving the effect of offline streaming media advertising according to claim 3, characterized in that: In step 2.4), when judging whether the collected face data exceeds the time threshold, the pixel coordinate accumulation method is used for statistics, including the following steps: Step 2.4.1) Arrange the acquired images by time, extract valid frames from the acquired video and arrange them by time; Step 2.4.1) Identify the face in the current image or valid frame and mark the pixel range where the face is located in red, record the coordinates of all pixel ranges marked in red, and mark each subsequent image or valid frame in red and record the coordinates; Step 2.4.2) Calculate the number of coordinates of the red-marked area in all images or valid frames. If the number exceeds the set threshold, it is considered that the time threshold has been exceeded.

5. The method for verifying and improving the effect of offline streaming media advertising according to claim 4, characterized in that: In step 2.4.2), the pixels in the red-marked area contain multiple pixels. When the pixel overlap of the red-marked areas in two different images or valid frames exceeds 40%, they are considered to be the same red-marked pixel range. If the number of times the same red-marked pixel range appears in different images or valid frames exceeds the set threshold, it is considered to have exceeded the time threshold.

Images