基于LabVIEW实现传统目标对象分类的方法  

Method of Traditional Object Classification Based on LabVIEW

作　　者：王娟 郭辉 吴龙飞 何多政 WANG Juan;GUO Hui;WU Longfei;HE Duozheng(AVIC Landing Gear Advanced Manufacturing Corp.,Changsha 410000,China)

机构地区：[1]中航飞机起落架有限责任公司,长沙410000

出　　处：《机械工程师》2025年第4期85-88,共4页Mechanical Engineer

摘　　要：目标对象分类是通过目标训练和目标分类,根据图像特征向量进行匹配识别,达到判定样品类别归属的一种方法。LabVIEW在目标对象分类方面提供了4种视觉颗粒分类训练方法。文中对Vision Particle Classification Training Interface工具配置颗粒分类器文件和借助Vision and Motion工具包搭建生成颗粒分类器文件的程序方法,进行了详细阐述,便于工程人员快速掌握和工程应用。在传统目标对象分类无法满足要求时,可基于深度学习的目标检测算法,在LabVIEW 64位环境中调用OpenVINO、TensorFLow进行对象训练和分类,进一步降低背景误检率,提高程序通用性。Target object classification is a method to determine the classification of sample through target training and target classification,matching and recognizing according to image feature vector.LabVIEW provides 4 visual particle classification training methods for target object classification.This paper describes the program method of configuring particle classifier file with Vision Particle Classification Training Interface tool and building and generating particle classifier file with Vision and Motion toolkit in detail.It is easy for engineers to grasp and apply in engineering quickly.When traditional object classification fails to meet the requirements,object detection algorithms based on deep learning can be used to call OpenVINO and TensorFLow in LabVIEW 64-bit environment for object training and classification,further reducing the background false detection rate and improving the versatility of the program.

关 键 词：目标训练 目标分类 机器视觉 图像采集 分类器 灰度图像 

分 类 号：TH12[机械工程—机械设计及理论]