Tomaž Kunst (2012) Application for analysis of promotions based on computer vision. EngD thesis.
In our work we present product based on computer vision for automatic analysis of persons in the vicinity of a promotion point. For execution we need a computer with camera and program for capturing real time video. We can use camera that is plugged in via USB port or remote network enabled camera. Firstly, we use Viola-Jones method for face detection.Later, we classify them based on gender. Support vector machine (SVM) in combination with principle component analysis (PCA) is used for classification and face features recognition. The classifier is built using image database of students pictures from University of Essex (GB). Achieved gender classification score is 84.75%. When faces cannot be detected, we use Lucas-Kanade method for following people. This approach helps us gain accurate data for number of persons who walked by. We monitor customer feedback on promotion. Final report includes data of predefined customer responses and classified data about them.
Actions (login required)