欢迎访问《汽车安全与节能学报》，
汽车安全与节能学报

汽车安全与节能学报 ›› 2014, Vol. 5 ›› Issue (01): 38-46.DOI: 10.3969/j.issn.1674-8484.2014.01.004

• 汽车安全 • 上一篇    下一篇

面向车位识别的相似度数据融合算法

朴昌浩1，温球良1，苏 岭1,2，张 强2，禄 盛1*   

  1. 1. 重庆邮电大学，重庆 400065，中国；
    2. 重庆长安汽车股份有限公司，重庆 400023，中国
  • 收稿日期:2013-12-18 出版日期:2014-03-25 发布日期:2014-04-08
  • 作者简介:朴昌浩(1978 － )，男( 朝鲜)，延边，教授。E-mail: piaoch@cqupte.edu.cn
  • 基金资助:

    国家自然科学基金资助项目（11247325) ；国家重点实验室资助项目（Q10-111893）；重庆市自然科学基金资助项目（CSTC2013YYKFC60005，CSTC2011BB4145）

Recognition algorithm based on similarity data fusion for vehicle parking space

PIAO Changhao1*, WEN Qiuliang1, SU Ling1,2, ZHANG Qiang2, Lu Sheng1   

  1. 1. Chongqing University of Posts and Telecommunications, Chongqing 400065, China;
    2. Chongqing Changan Automobile CO., LTD., Chongqing 40023, China
  • Received:2013-12-18 Online:2014-03-25 Published:2014-04-08

PDF

1556

摘要： 为提高平行泊车车位识别率，提出了基于多个超声波传感器的车位检测、智能识别系统。
确定边缘阈值参数，利用基于相似度的数据融合算法，对多传感器测量结果进行融合。分析车速和
横向距离的试验结果。为修正误差，回归了一种多元线性模型。对比了传统单探头方法、双探头平均
值法，和本文的双探头数据融合等方法的识别率。结果表明：在模拟车位环境下和在实际泊车环境下，
修正后的本方法的识别率分别为95.24% 和90%，均高于上述两种传统方法的成功率。

关键词: 交通运输, 人工智能, 车位识别, 相似度数据融合, 多元线性回归

Abstract: A recognition algorithm based on similarity data fusion was built with a multi-ultrasonic sensor
detection model to improve the recognition rate in artificial intelligent parallel vehicle parking system. Measured
data from multi-ultrasonic sensors were fused by using the method after threshold parameters of space edges
were determined. A multiple linear regression model was established to analyze the influence of vehicle speed
and transverse distance. A new method using data fusion with double sensors, a traditional method with single
sensor, and a mean value method with double sensors were compared to estimate recognition success rate.
The results show that the recognition rate with the new algorithm is 95.4% in simulated parking environment and
is 90% in actual environment, these rates are better than the rate by the two traditional methods.

中图分类号: