Generic object recognition using automatic region extraction and dimensional feature integration utilizing multiple kernel learn