复杂网络下机械零件三维轻量级识别方法

Three Dimensional Lightweight Recognition Method of Mechanical Parts Under Complex Network

为精准获取机械零件尺寸,明确产品质量是否达到安全标准,提出一种复杂网络下机械零件三维轻量级识别方法。将机械零件三维信息转换为复杂网络下加工数据,利用邻接矩阵描述网络节点与边的内在关联,创建机械零件三维模型;把零件数据点拟作局部多项式曲面,推算各点在曲面内的投影残差,计算特征点相关指数,使用折线生长技术提取模型特征线;将三维模型变换成体素矩阵,确立不同三维特征间的差异,通过二维典型视图训练卷积神经网络,在正、反向传播运算后输出三维轻量级识别结果。仿真结果表明,所提方法有效提升了机械零件三维模型识别精度与效率,鲁棒性强,为机械零件的高质量生产提供可靠数据支持。

In order to accurately obtain the size of mechanical parts and determine whether the product quality meets the safety standard,a three-dimensional lightweight recognition method of mechanical parts under complex network was proposed.The three-dimensional information of mechanical parts was transformed into machining data under complex network,and the internal relationship between network nodes and edges was described by adjacency matrix to create the three-dimensional model of mechanical parts.The part data points were modeled as a local polynomial surface,the projection residual of each point in the surface was calculated,the correlation index of feature points was calculated,and the broken line growth technology was used to extract the feature line of the model.The three-dimensional model was transformed into voxel matrix,the differences between different three-dimensional features were established,the convolution neural network was trained through two-dimensional typical view,and the three-dimensional lightweight recognition results were output after forward and back propagation operation.Simulation results show that the proposed method effectively improves the accuracy and eficiency of 3D model recognition of mechanical parts,has strong robustness,and provides reliable data support for high-quality production of mechanical parts.