基于三维图像的铁路扣件缺陷自动识别算法
发布时间：2018-08-30 11:39
【摘要】：针对当前铁路扣件状态自动识别准确率和稳定性不高等问题,利用直射式激光三角测量法原理研发扣件检测系统,采集不受环境光影响的高质量轨道三维数据。提出基于三维图像的扣件区域定位方法,并利用先验知识验证扣件位置以保证扣件定位的准确性;基于弹条的高度规律信息提取弹条,采用HGOH作为特征描述算子;根据特征向量的模是否等于零可识别出缺失扣件,将模不为零的特征向量送入已训练的SVM分类器,从而识别断裂扣件和完整扣件。室内试验研究结果表明,采用本文提出的扣件缺陷自动检测算法,识别准确率可达98.0%,能满足扣件缺陷自动化检测的需要。
[Abstract]:In order to solve the problem of low accuracy and stability in automatic identification of railway fasteners, a fastener detection system based on direct laser triangulation is developed to collect high-quality 3D orbital data unaffected by ambient light. Ensure the accuracy of fastener positioning; extract the bullet strip based on the height rule information of the bullet strip, and use HGOH as the feature descriptor operator; identify missing fasteners according to whether the modulus of the feature vector is equal to zero, and send the modulus of non-zero feature vector to the trained SVM classifier to identify broken fasteners and complete fasteners. The results show that the recognition accuracy can reach 98.0% by using the proposed algorithm, which can meet the needs of automatic detection of fastener defects.
【作者单位】： 西南交通大学土木工程学院;西南交通大学道路工程四川省重点实验室;俄克拉荷马州立大学土木与环境工程学院;西南交通大学高速铁路线路工程教育部重点实验室;
【基金】：国家自然科学基金(51478398,51308477,U1534203) 中央高校基本科研业务费专项资金(2682015CX091)
【分类号】：TP391.41;U216.3
，
本文编号：2212957
[Abstract]:In order to solve the problem of low accuracy and stability in automatic identification of railway fasteners, a fastener detection system based on direct laser triangulation is developed to collect high-quality 3D orbital data unaffected by ambient light. Ensure the accuracy of fastener positioning; extract the bullet strip based on the height rule information of the bullet strip, and use HGOH as the feature descriptor operator; identify missing fasteners according to whether the modulus of the feature vector is equal to zero, and send the modulus of non-zero feature vector to the trained SVM classifier to identify broken fasteners and complete fasteners. The results show that the recognition accuracy can reach 98.0% by using the proposed algorithm, which can meet the needs of automatic detection of fastener defects.
【作者单位】： 西南交通大学土木工程学院;西南交通大学道路工程四川省重点实验室;俄克拉荷马州立大学土木与环境工程学院;西南交通大学高速铁路线路工程教育部重点实验室;
【基金】：国家自然科学基金(51478398,51308477,U1534203) 中央高校基本科研业务费专项资金(2682015CX091)
【分类号】：TP391.41;U216.3
，
本文编号：2212957
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2212957.html
