一种基于Pseudo-Zernike矩和SVM的目标识别快速算法

Fast target recognition algorithm based on Pseudo-Zernike moment and support vector machine

为减少Pseudo-Zernike矩的计算复杂度,将系数迭代算法与核函数的对称性相结合,提出了一种新的混合快速算法。与现有的两种快速算法相比较,新算法有更快的计算速度。将其应用到遥感飞机图像库识别中,首先提取图像的Pseudo-Zernike矩特征,然后将其作为支持向量机分类器的输入。理论分析和实验证明,新算法在保持识别准确率的情况下提高了识别速度。

To reduce the computational complexity of the Pseudo-Zernike moment,this paper proposed a fast hybrid algorithm based on the coefficient＇s iterative method and symmetry. Compared with two existing fast algorithms,reduced the hybrid method＇s calculation cost greatly. It used the new algorithm in the experiments of airplane recognition. Firstly,extracted the Pseudo-Zernike moment feature from each image,then employed the support vector machine as classifier. Experiment and theoretical analysis shows that the new recognition algorithm performs efficiently,and the calculation speed is improved with the same accurate recognition accuracy.