模糊图像中的飞机识别方法

针对模糊图像中飞机识别问题,提出一种基于角点和核聚类的飞机识别方法。该方法提取图像的角点特征,对角点进行核聚类,根据聚类结果识别飞机的角点,完成飞机的定位识别。在核聚类算法中,引入有效性函数,能自适应的确定聚类数目,解决了现有核聚类算法需要事先确定聚类数的弱点。实验表明,角点特征具有位移、旋转、尺度不变性,自适应核聚类算法能准确识别模糊图像中的飞机,具有较好的鲁棒性。

Airplane Identification Method in Blur Images

Aiming at the airplane identification problem in blur image, a novel airplane identification method based on corner and kernel clustering is presented. The new method extracts corners of the image, classifies corners using kernel clustering, and then detects airplane corners based on clustering result and finishes identification of airplanes. In the kernel clustering algorithm, by introducing validity measure function, the clustering algorithm can get the sort number self-adaptively, and overcomes the weakness of existing kernel clustering algorithm which must set the sort number before clustering. Experiment results show that the corner feature is invariant, the self-adaptive kernel clustering algorithm can recognize airplanes correctly in blur images and is applicable.