首页|基于立体视觉的高速公路收费机器人目标识别方法

基于立体视觉的高速公路收费机器人目标识别方法

扫码查看
在高速公路上,智能收费机器人的应用已经逐渐得到普及,为使其在应用中实现更加准确地进行目标识别,针对现有方法单目视觉的特性,设计基于立体视觉的高速公路收费机器人目标识别方法,以实现目标识别准确率与鲁棒性的突破.分别对双目立体视觉目标图像施灰度化处理和滤波与增强处理.采用鲁棒性好、精度较高的张氏棋盘格非线性标定法,使用Stereo Camera Calibrator工具箱对预处理后的目标图像实施双目立体视觉相机标定.基于标定结果,应用SURF算法实施双目立体视觉目标图像的特征提取与匹配.改进模板匹配算法,依据匹配度评价准则实施匹配度评价,寻找最高相似度的部分,实现双目立体视觉目标图像识别.测试结果表明,设计方法能够实现人体手部与高速ETC卡或ETC付款码的识别,不仅目标识别准确率较高,随着背景目标物个数的持续增加,其目标识别准确率没有出现大幅下降的问题,鲁棒性良好.
A Target Recognition Method for Expressway Toll Robot Based on Stereoscopic Vision
The application of intelligent toll robots on highways has gradually gained popularity.In order to achieve more accurate target recognition in applications,a stereo vision based target recognition method for highway toll robots is designed based on the char-acteristics of existing monocular vision methods,in order to achieve breakthroughs in target recognition accuracy and robustness.Ap-ply grayscale processing and filtering and enhancement processing to binocular stereo vision target images,respectively.Using the Zhang's checkerboard nonlinear calibration method with good robustness and high accuracy,the Stereo Camera Calibrator toolbox is used to calibrate the pre processed target image using binocular stereo vision cameras.Based on the calibration results,the SURF al-gorithm is applied to extract and match features of binocular stereo vision target images.Improve the template matching algorithm,im-plement matching evaluation based on matching evaluation criteria,find the part with the highest similarity,and achieve binocular stereo vision target image recognition.The test results show that the design method can achieve the recognition of human hands with high-speed ETC cards or ETC payment codes.Not only does it have high target recognition accuracy,but as the number of back-ground objects continues to increase,its target recognition accuracy does not significantly decrease,and its robustness is good.

binocular stereo visionfourier transformdual camera calibrationSURF algorithmtarget identification

王枫、张胜

展开 >

交通运输部科学研究院,北京 100029

曲靖中建曲昆高速公路投资开发有限公司,云南曲靖 655000

双目立体视觉 Fourier变换 双相机标定 SURF算法 目标识别

中央级公益性科研院所基本科研业务费专项中国建筑集团有限公司交通强国建设试点任务科研项目

20230205YLZJ2022-12

2024

自动化与仪器仪表
重庆工业自动化仪表研究所,重庆市自动化与仪器仪表学会

自动化与仪器仪表

CSTPCD
影响因子:0.327
ISSN:1001-9227
年,卷(期):2024.(4)
  • 15