利用高阶累积量(HOS)的性质推导出了一种根据三阶累积量求图象特征集的“不变性”算法。得出的二维特征集，不但具有对图象的平移、旋转和比例变化的不变性、对附加高斯噪声的抑制作用、不存在相位丢失问题（它是关于图象信息的全描述），而且不需要进行对数坐标的傅氏变换，因而计算简化、实时性高。仿真试验结果证明了该方法的有效性

Phenomenon of missing often occur at the description of images by corrlelation function with order 1 and 2. However,the triple cumulation of 2D images,which possess translation invariance and insensitivity to gauss noise,is a complete reprsentation of images.In this paper,a new algorithm based on the propreties of translation invariance and insensitivity to gauss noise is suggested for the extraction of feature vector from images. The feature vector can be used for target recognition for the reason that it is invariance of shift.rotation and scaling. The simulation experiments are carried out to show the effectiveness of the new method.