Deep feature of image screened by improved clustering algorithm cascaded with genetic algorithm

Feature extracting and screening get more important and necessary because of data analysis will become very slow and difficult with the increasing of data dimension. To reduce the dimension of features, we propose a new way of feature screening in this paper. The improved clustering algorithm is employed to screen the features preliminarily, and then the genetic algorithm synergistically combined with the random forest is cascaded to screen the features deeply. To validate the way feasible, 1588 tobacco leaves belonging to 41 grades are used to be classified in the experiments. The results show that both the recognition rate and the speed can be improved. This demonstrates that the presented cascaded screening approach can raise not only the recognition rate but also the speed because the feature dimension is decreasing effectively.