说明：  利用矩特征可以从二维图像识别三维目标,4阶矩的图像峰值反映了目标的形状,为了使它具有平移、比例和旋转不变性,作者对其进行了归一化处理。
(use Moments from the three-dimensional two-dimensional image recognition, four bands peak moments of images reflect the shape of the target, In order to give it scale, translation and rotation invariant, the author of its normalized treatment.)