Digital Video Scenes Recognition using Mijn-EA and Learning Vector Quantization Network


This paper presents a hybrid method for digital video scenes recognition which uses recognition to group pixels into known object for each frame. The chromaticity is used as data source for the method because it is normalized. The recognition is carried out by means of an Evolutionary Algorithm from type Mijn-EA, which is employed to obtain the best clusters that represent each frame and number of seeds in each frame. Then, conversion the color space for each cluster from RGB to HSV to fined the textural features of co-occurrence matrix. After that the pixels for each cluster recognition according to the identified classes. The number of classes is a priori unknown and the Evolutionary Algorithm that implements the Mijn-EA is used to determine the main clusters. The detection of the classes in the LVQ3 is done using a texture features. The obtained results substantiate the feasibility of the method and the accuracy.