CAJ | 학술논문

针对场景分类方法在室内场景领域的分类精度普遍较低的问题，提出一种融合全局特征和局部特征的多特征室内场景分类的方法。首先，提取场景图像的SIFT局部特征并根据关键点位置进行聚类处理和降维，得到统一维度的SIFT特征矩阵；其次，提取场景图像的PHOG局部特征和Gist全局特征，并与SIFT特征融合在同一特征矩阵中；然后，采用SVM分类器进行场景分类的训练与识别。实验结果表明，相对于单一特征的场景图像分类方法，本文的方法具有更高的分类精度。
침대장경분류방법재실내장경영역적분류정도보편교저적문제，제출일충융합전국특정화국부특정적다특정실내장경분류적방법。수선，제취장경도상적SIFT국부특정병근거관건점위치진행취류처리화강유，득도통일유도적SIFT특정구진；기차，제취장경도상적PHOG국부특정화Gist전국특정，병여SIFT특정융합재동일특정구진중；연후，채용SVM분류기진행장경분류적훈련여식별。실험결과표명，상대우단일특정적장경도상분류방법，본문적방법구유경고적분류정도。
In order to improve the recognition accuracy of indoor scene images a method was proposed through fusing the global and local features.First, the SIFT feature of scene images are extracted and the key points of SIFT are clustered in order to obtain the same dimension feature vector.The PCA is em-ployed to reduce the dimension of feature matrix.Second, the PHOG and Gist features are extracted re-spectively and fused with the SIFT to construct feature matrix.Finally, the SVM is employed to classify the scene images'types.The experimental results show that the recognition accuracy of multi-featured fu-sion is better than those single-featured ones.