论文部分内容阅读
介绍一种利用径向矢量提取形状特征的方法，着重于分析二维图形的不变性，提出一种带有方向因子的径向矢量描述．该矢量以图形边界弧长为自变量，完整地刻画了图形特点，克服了非凸图形识别中存在的多义性．以归一化的特征矢量为输入，采用三层神经网络为分类器，在字符及军事目标不变性识别中取得了令人满意的结果
A method of using radial vectors to extract shape features is introduced, focusing on the analysis of the invariance of two-dimensional graphics, a radial vector description with directional factors is proposed. This vector, taking the arc length of graphics boundary as an independent variable, completely characterizes the graphics and overcomes the ambiguity existing in non-convex graphics recognition. Taking the normalized feature vector as input and the three-layer neural network as the classifier, satisfactory results are obtained in the invariance of characters and military targets