利用空间金字塔分块与PLSA的场景分类方法

Scene Classification Using Spatial Pyramid Blocks and PLSA

提出一种基于空间金字塔分块与PLSA方法相结合的场景分类方法.该方法首先通过空间金字塔分块的方式来构建图像区域集合,然后利用概率潜在语义分析(PLSA)从图像的区域集合中发现潜在语义模型,最后根据潜在语义模型找出所有图像区域中潜在语义出现概率来构建区域潜在语义特征,并使用该特征构建SVM模型进行场景分类.在13类场景图像上的试验表明,和其他方法相比,该方法中不需要进行大量的手工标注,而且具有更高的分类准确率.

Presented a novel scene classification method based on Spatial Pyramid Blocks and PLSA. Spatial pyramid blocks collection is first achieved by dividing image, and then Probabilistie Latent Semantic Analysis （PLSA）, a generative model from the statistical text literature is applied to a bag of visual words representation for block regions. Finally, latent semantic feature is constructed by applying PLSA model to each spatial pyramid block region in an image. The resulting feature retains sufficient discriminative information for accurate scene classification. Experiment results show that this method has satisfactory classification performances on a large set of 13 categories of complex scenes.