Design and construction of advertising culture visual communication recognition system based on deep learning neural network

  • Qiyue Chen


As a traditional and classic advertising model, billboard is characterized by wide audience, convenient delivery, high mobility of viewers, strong pertinence of occupation and crowd, etc., so it has always been an important way of advertisement. When online payment is popular, how to keep the classical vitality of billboard has become the focus of advertisers and supervision departments. In this paper, based on the convolutional neural network technology, taking the need of an enterprise billboard recognition as the background, and on the basis of traditional algorithm as the core in the early period, the deep learning algorithm model of the convolutional neural network was introduced to conduct the research. First of all, in this paper, various technologies adopted by corresponding systems as well as the basic process owned by image recognition were established. Secondly, this study analyzed the specific requirements for billboard recognition on a mobile platform and the adaptation to a complex application environment and real-time accuracy. Thirdly, based on the above research and analysis, it proposed a recognition system scheme which divided the deep learning neural network model and then assign to the client and server. Then a certain layer of convolution operation was carried out in the client-side, which avoided the massive data from uploading. Therefore, compared with traditional algorithms and non-partition neural network algorithms, the execution efficiency was improved to some extent, which not only achieved the real-time requirement performance but also could deal with complex shooting environment.