基于支撑向量机的空瓶智能检测方法  

Empty Bottle Intelligent Inspecting Method Based on Support Vector Machines

作　　者：刘焕军[1] 王耀南[1] 段峰[1] 

机构地区：[1]湖南大学电气与信息工程学院,长沙410082

出　　处：《控制与决策》2005年第12期1434-1437,共4页Control and Decision

基　　金：国家自然科学基金项目(60375001)

摘　　要：为实现空瓶的智能检测,对空瓶检测的机器视觉方法进行了深入研究.论述了如何进行处理区域的标定,以及提取空瓶图像特征的方法.在提取了特征之后,提出用遗传支撑向量机算法来进行分类决策,支撑向量机具有较好的推广能力.同时采用遗传算法来优化选择支撑向量机的参数,以保证支撑向量机具有优良的分类性能.实验表明,采用这种方法检测空瓶,其检测准确率可达95%以上.The empty bottle inspected methods are discussed based on machine vision for inspecting empty bottles intelligently. A method is presented to identify the processing region, and extract features from images. After extracting feature, the GA-support vector machines （SVMs） are used as classifier. The SVMs have good generalizing ability. At the same time an optimization algorithm based on genetic algorithm is proposed to confirm the parameters of the support vector machines, and it ensure that the SVMs have good classification ability. The experiments show that by using this method to inspect empty bottles, the accuracy rate may reach above 95%.

关 键 词：智能检测 支撑向量机 机器视觉 遗传算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]