基于差异演化概率神经网络的电气构件分类研究

摘要

不变矩特征具有较好的几何图形物理含义,提出了一种对安装工程电气构件利用神经网络分类识别的方法.通过对安装工程图分割不同的电气构件,计算电气构件的不变矩特征,对特征进行矢量归一处理,引入差异演化算法优化平滑因子,改进概率神经网络作为分类器对电气构件进行分类.实验结果证明方法的有效性和准确性,对于安装工程量计算自动化具有一定的实用性和参考价值.%Moment invariant feature has good geometric and physical meaning, proposed a method for electrical components of installation engineering using neural network to classification and recognition. Through the installation engineering of graph partitioning different electrical components, Calculation of the electrical component of moment invariant feature, normalization processed feature vector, using differentia evolution algorithm to optimize the smooth factor, modified probabilistic neural network as classifier of electrical components. Experimental results show that the method is effective and accurate ,for installation engineering quantity calculation of automation has practicability and reference value.