基于纹理基元与颜色的室外自然场景分类

为解决算法生成纹理地图时时间耗费量大的问题,提出采用KD-tree算法对数据结构进行划分、减小KNN算法搜索复杂度、提高搜索速度的方法.针对基于纹理基元的分类算法无法准确检测室外某些纹理相似性较高的自然场景,提出加入颜色特征、设置相应权值构建混合模型的方法.实验结果表明,基于KD-tree的KNN算法可缩短分类时间、满足实时性的要求,基于纹理基元与颜色的分类算法在室外自然场景中能够获得较高的分类精确度.

Algorithm for Outdoor Natural Scene Classification Based on Texture-Element and Color

For the problem of time-consuming in generating texture map, KD-tree algorithm is used to divide the data structure and reduce the complexity of KNN search algorithm. To solve the difficulty of image classification based on texture algorithm that can not accurately detect some textures with high similarity in outdoor natural scene, an improved method is proposed by adding color feature, setting the corresponding weight, and building the hybrid model. Experimental results show that the improved KNN algorithm based on KD-tree can shorten the classification time and meet the real-time requirements. The improved classification algorithm based on texture and color can obtain higher classification accuracy for the application of outdoor natural scene.