Abstract: To meet the requirements of cavity feature recognition for corner box parts， an automatic cavity feature recognition method was proposed based on improved graph matching. The cavity features of corner box parts were analyzed and the common template attributes of feature simplified model were extracted. The raster height point cloud data of the model was obtained， and the median height was taken as the threshold value to transform it into a 0-1 matrix. The projection eigenvalues of the cavity surfaces were extracted from the 0-1 matrix to separate and recognize the cavity wall and edge surfaces. Face adjacency attribute was used to search the side wall surface of the cavity， and the complete machining features of the cavities were finally identified， including the bottom surface set， the wall side surface set and the side wall surface set. The results of example verification show that the proposed method is more efficient and robust than those of the fixed template matching method to identify the machining features of the cavity of corner box parts. 

Key words: attributed adjacency graph, graph matching, corner box part, feature recognition

摘要： 为满足角盒类零件型腔特征识别需求，提出了一种基于改进图匹配的型腔特征自动识别方法。分析了角盒类零件型腔特征，提取了特征简化模型的共性模板属性。获取了模型栅格高度点云数据，以高度中值作为阈值并将其转换为0-1特征值矩阵，提取了型腔面投影特征值以分离识别型腔壁边面。采用共性模板面邻接属性判定搜索型腔侧壁面，最终识别出了包含底板面集、壁边面集和侧壁面集的完整型腔加工特征。实例验证结果表明，相较于固定模板匹配的特征识别方式，采用所提方法识别角盒类零件型腔加工特征时具有更高的效率和更好的鲁棒性。

关键词: 属性邻接图, 图匹配, 角盒类零件, 特征识别