基于分割的局部Sigma语义特征点，是对场景中的语义目标进行建模。先在传统的图像分割基础上，分割出场景的前景目标，再结合像素位置、颜色、Gabor特征和LBP特征[构造出表征目标语义信息的协方差描述子，最后将其转换成欧式空间下的Sigma点特征，适用于标准SVM的场景学习和分类。(The segmentation based local Sigma semantic feature points are modeling the semantic objects in the scene. In the traditional image segmentation based on the segmented foreground scene, we combine with the pixel location, color, Gabor and LBP feature [construct covariance descr iptor represents the target semantic information, and finally converted into Euclidean space Sigma features, suitable for learning and classification standard to the scene SVM.)