TV Corner Mark Recognition Algorithm Based on Feature Fusion

This paper proposed a fusion of color features, shape features and SURF image recognition algorithm. Traditional SURF algorithm take greyscale image as input to extract local extreme value points as characteristics points, ignoring color and shape information. The paper introduced the global color histogram information to make up the lack of color information loss, at the same time introduced shape context descriptor to describe shape feature of the image. Then merge three feature descriptors to form joint feature descriptor, and gives different weights to different characteristics. The experimental results show that the algorithm has good recognition ability.