A banner ads searching and counting system for sports videos

Abstract

The ads exposure frequency in a TV is very important for the pricing of the advertising. If we count this exposure frequency by person, it will be very time consuming and error-prone. In this paper, we propose an approach and design a system for searching and counting the number of banner ads appearing in sports videos. The searching proceeds by matching feature points from pre-specified sample ads using the color histogram and the SURF (Speeded-Up Robust Features) algorithm. This searching approach can robustly identify objects among scaling and partial occlusion while achieving near real-time performance. This system worked effectively in our experiments using real videos.