Smart Advertising And Effect Evaluation System

With the development of commerce, the businessperson pays more and more attention to the effect of advertising, advertising effectiveness assessment and intelligent play are two important parts for research. This paper designes a smart advertising and effect evaluation system first by collecting customers'face images, then detecting faces, tracking and identifying these detected faces to distinguish consumer groups.The purpose of this paper is to design an embedded processing terminal, which can realize the functional features above automatically. ADI Blackfin series DSP is applied as hardware platforms, the algorithms are mainly designed to analyze face images, and take different actions according to the analysis results, associated algorithms have been developed and debugged on PC, methods of migrating these algorithms to DSP are given at last. Based on previous research, this paper studies mainly on three aspects-face detection, tracking and recognition:1. Based on the AdaBoost algorithm for face detection, the algorithm of the color matching test is proposed. AdaBoost algorithm usually uses human face gray image information. In the detection process, it is easy to mistake the human face as non-human face. The faulty results can be removed by using color matching strategy in the following-up treatment in order to reduce the false detection rate.2. As for particle filter theory, the tracking algorithm based on particle filter is discussed. Introduce the implementation of tracking algorithm based on the following aspects: target motion model, target observation model and partical resampling. And, a strategy of Mean Shift algorithm combining particle sampling is proposed to reduce the complexity of particle filter and achieve better tracking performance.3. Classify the human face using SVM. Based on SVM theory, the face will eventually devided into two categories by LIBSVM parameter designing and positive/negative samples training, and a satisfactory result was achieved.