Sports Videos Classification Based on Color Texture and SVM
Wu Pen · Journal of Fujian Normal University · 2014
The existed methods of the sports video classification cannot improve classification accuracy using a single feature or simple combination,a method is proposed to extract the nonlinear characteristics of color and texture,and then design a classifier based on SVM. The main works include: color space conversion,definition of color index matrix,color co-occurrence matrix,color texture extraction and classification algorithm. The experimental results show that the discrimination degree of category using nonlinear fusion method is higher than single color,single texture or simple combination of the color and texture in sports video classification. The average accuracy increases9. 94%,8. 66%,6. 90% respectively.