针对现有算法对稀疏图像识别率不高的问题,提出了基于目标顶部特征的BP神经网络目标识别算法.该算法通过分析模拟目标的线阵扫描数据,应用统计算法提取目标顶部轮廓数据特征,建立目标三维特征图模型,采用最小生成树Prim算法得到目标顶部轮廓空间分布特征,在此基础上设计了对目标特征分类的BP神经网络.仿真结果表明,该算法简单易于实现,并且在预测误差│error│≤0.015时,该算法对目标稀疏图像有较好的识别效果.%Aiming at the low recognition rate for sparse images in using existing methods, an target recognition algorithm of neural network based on target top profile features was presented in the paper. The method analyzed the linear array scanning data of target, and extracted the target's contour data characteristics with statistical methods, then the minimal spanning tree prim algorithm was used to get the spatial distribution characteristics of target top profile on the basis of 3-D feature graph model about target's top contour. Based on which, a BP neural network algorithm for classification of target characteristics was designed. The simulation result showed that the arithmetic is simple and easy to implement, and had preferable effect in discerning the sparse image of targets.
展开▼
