基于SIFT-SVM的发动机主轴承盖识别与分类  

Classification of engine main bearing cap parts using SIFT-SVM method

作　　者：石志良[1] 张鹏飞 李晓垚 SHI Zhi-liang;ZHANG Peng-fei;LI Xiao-yao(Mechanical and Electrical Engineering,Wuhan University of Technology,Wuhan Hubei 430070,China)

机构地区：[1]武汉理工大学机电工程学院,湖北武汉430070

出　　处：《图学学报》2020年第3期382-389,共8页Journal of Graphics

摘　　要：机械零部件的识别与分类任务是制造业自动化生产线的关键环节。针对发动机主轴承盖混合清洗后的分类,通过分析发动机主轴承盖零件的实际特征,提出基于SIFT-SVM的主轴承盖分类识别方法。该方法首先提取训练数据集图像的所有尺度不变(SIFT)特征向量,采用K-means聚类方法,将所有的特征向量聚类成K个分类,并将其代入词袋模型(BoW)中,使用K个"词汇"来描述每一张训练图像,从而得到图像的BoW描述。且以每张图像的BoW描述作为训练输入,使用支持向量机(SVM)训练主轴承盖的分类模型。实验结果表明:在标定的照明条件下,主轴承盖零件的识别率可达100%,单个零件识别时间为0.6 s,验证了该算法的有效性和高效性。The recognition and classification of mechanical components is a key process on the manufacturing automation line. In terms of the classification of the mixed and cleaned engine main bearing cap, through the analysis of the actual characteristics of the main bearing cap parts, the classification and recognition method for the main bearing cap based on SIFT-SVM was proposed. The method first extracted all scale-invariant feature transform(SIFT) feature vectors of the training dataset image, then employed the K-means clustering method to cluster all feature vectors into K classifications, and substituted the obtained K clustering results into the bag of word model(BoW). "Vocabulary" was utilized to describe each training image, thereby obtaining a BoW description of the image. The BoW description of each image served as a training input, and the classification model for the main bearing cap was trained using a support vector machine(SVM). The experimental results show that under the calibrated lighting conditions, the recognition rate of the main bearing cap parts can reach 100%, and the recognition time for a single part was 0.6 seconds, which verified the effectiveness and efficiency of the algorithm.

关 键 词：零件识别与分类 机器视觉 SIFT 词袋模型 支持向量机分类器 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]