Design And Implementation Of Intelligent Video Advertising System Based On Jetson TX1

In recent years,the emergence of various new media has brought tremendous impact on traditional offline video advertising.Since the traditional offline advertising format is very rough,and is mostly circulated on the basis of local media advertising libraries.Moreover,it only focuses on the number and frequency of delivery but does not care whether advertisements are of concern of concern to consumers.Therefore,advertising has not achieved the expected results.In view of the above problems,this thesis presents an intelligent advertisement display system based on the Jetson TX1 visual development platform.In the absence of human decision-making,the system can analyze the surrounding people's information through the camera,and push the ad that is most suitable for the current audience through the decision.Experimental results show that the smart advertisement display system achieves better advertisement push.The main contents of this thesis are as follows:Demand analysis was conducted on the business functions of the intelligent video advertisement display system,expounding the overall design scheme of the smart video advertisement display system based on Jetson TX1,and setting up the hardware and software development environment.(2)Using the hash algorithm to compare the similarity of the feature maps extracted from the gender and age recognition process,a YOLO-based gender age detection network is implemented.The network can simultaneously perform gender identification and age estimation.The experimental results show that the network improves the operating efficiency approximately twice on the basis of ensuring better recognition accuracy.(3)Using the GTK+ graphical interface toolkit and the Gstreamer multimedia framework,a media advertisement player is designed and implemented,including audio and video decoding module,graphical user interface and system control module.Based on the bus and message processing mechanism,the message transfer and status control between the various modules is realized.The player can support the playback of video files in a variety of formats,and has a good man-machine interface.It can also switch current advertisements in time according to detection and recognition results.(4)In order to realize the overall function of the ad-push system,an ID3 decision tree algorithm is used to classify ads,and a method for statistical classification detection and recognition results is proposed.Based on the above-mentioned advertisement classification and classification methods,an ad-push software is designed to realize intelligent advertising push.The kernel function in the detection and recognition algorithm is optimized,and the test results show that the running time of the algorithm is reduced by 12% compared with the original.