CASIA OpenIR  > 毕业生  > 硕士学位论文
无人驾驶车辆交通标志和泊车认知能力测试设计与评价方法研究
Alternative TitleStudy on Traffic Sign and Parking Cognitive Ability Tests Design and Evaluation Methods of Unmanned Vehicles
张雅如
Subtype工学硕士
Thesis Advisor汤淑明
2012-06-03
Degree Grantor中国科学院研究生院
Place of Conferral中国科学院自动化研究所
Degree Discipline控制理论与控制工程
Keyword无人驾驶车辆 认知测试 交通标志识别 自主泊车 评价方法 Unmanned Vehicle Cognitive Test Traffic Sign Recognition Autonomous Parking Evaluation Methods
Abstract无人驾驶车辆集环境感知、认知、决策规划和驾驶控制等功能于一体,能够自主、安全、可靠地在特定环境下行驶。无人驾驶车辆智能行为的测试与评价是国家自然科学基金委重大研究计划“视听觉信息的认知计算”(简称重大研究计划)的一项重要研究内容。目前,我国对无人驾驶车辆的智能驾驶行为能力的测试与评价主要是通过举办“中国智能车未来挑战赛”开展相关研究工作的。 众所周知,无人驾驶车辆的智能驾驶行为是受其感知、认知、决策、控制及相关自然环境等多种因素共同作用的结果,其中认知能力是体现无人驾驶车辆智能行为能力的一个重要方面。对无人驾驶车辆认知能力测试与评价的研究也是该重大研究计划的一项主要工作。本文拟开展无人驾驶车辆认知能力测试设计及其评价方法方面的研究工作,希望能推动无人驾驶车辆智能测试与评价及其相关技术的快速发展。 首先介绍了本论文的整体研究思路:从交通场景认知和交通现象认知两个角度设计了无人驾驶车辆认知能力的测试方案,提出了交通标志认知能力和泊车认知能力两项测试内容,选取了评价指标,并对其进行了评价。交通系统中的交通标志共有七大类,本文将其认知能力测试分为图像特征的分辨性能测试和分类算法的分类性能测试两种,每种测试进一步又分为类间交通标志识别测试和类内交通标志识别测试。对于泊车认知能力测试,本文通过改变无人驾驶车辆与泊车位之间的方向和位置关系,来测试无人驾驶车辆的泊车认知能力。在评价研究方面,基于上述两项测试内容,通过比较多种不同的评价算法,本文选取模糊TOPSIS评价方法对无人驾驶车辆认知能力进行评价,并对该方法进行了改进,将改进后的方法与机器智能测试中常用到的Sugeno模糊积分评价方法进行比较,从而验证本文改进的模糊TOPSIS评价方法在无人驾驶车辆认知能力评价方面的有效性。 其次,测试了无人驾驶车辆的交通标志认知能力。在模式识别研究中,通常采用的交通标志识别能力评价指标包括识别正确率、训练时间、分类时间等,本文设计了基于GTSB交通标志数据库的类间测试和类内测试。在类间和类内测试中,测试了基于One-Against-One SVM分类器和One-Against-All SVM分类器分别对不同的图像特征的交通标志的识别能力,并用混淆矩阵分析了基于不同图像特征的两种SVM分类器对不同交通标志的识别能力。然后在类间和类内两项测试的基础上,进一步选取了未训练情况识别正确率作为对常用指标的补充,为后续的评价提供了基础。接着还探讨了可能的10余种指标,并验证了其有效性。 再次,测试了无人驾驶车辆的泊车认知能力。采用本文提出的泊车控制算法,通过改变无人驾驶车辆与泊车位之间的相对方向和位置,测试了无人驾驶车辆的泊车认知能力。选取了三种指标(泊车时间、泊车结束后车辆中心位置与泊车位中心位置的距离、泊车轨迹的平滑程度),并验证了其有效性,为后续的评价提供了基础。 然后,评价了无人驾驶车辆的认知能力。本文对模糊TOPSIS评价方法进行了改进:在研究模糊数计算方法的基础上,对传统三角模糊数的距离计算公式进行了简化计算,并将简化后的三角模糊数计算方法引入到模糊TO...
Other AbstractUnmanned vehicle integrates environmental perception, cognition, decision making and driving control function. It can drive autonomously, safely and reliably under certain circumstances. The testing and evaluation of intelligent behavior of unmanned vehicle is a major research project of “Cognitive Computing of Visual and Auditory Information” funded by National Natural Science Foundation of China” (referred to as a major research project). By far, the testing and evaluation of intelligent behavior of unmanned vehicle is through holding “China intelligent vehicle challenge”. As we all know, the intelligent driving behavior is the combined results of perception, cognition, decision, control and various natural factors, and cognition is an important aspect that reflected intelligent behavior of unmanned vehicle. Testing and evaluation of cognitive ability of unmanned vehicle is also an important work in this major research project. Therefore, this thesis intends to carry out the research work of tests design and evaluation of cognitive ability of unmanned vehicle, hoping to promote the rapid development of the testing and evaluation of unmanned vehicles. First, the overall research ideas have been described. Traffic sign cognition test and parking cognition test have been designed from two perspectives of traffic scene cognition and traffic phenomena cognition, indicators have been selected to be used in further evaluation. Based on the fact that traffic signs can be divided into seven categories, the traffic sign cognition test is divided into image feature resolution performance test and classification algorithm performance test. Each test is further divided into inter-category test and inner-category test. The parking cognition test has been designed by changing the relative direction and position between unmanned vehicle and parking space to test parking cognitive ability of unmanned vehicle. On the basis of the above two tests, based on the study of various evaluation methods, this thesis intends to improve the fuzzy TOPSIS evaluation method to measure cognitive ability of unmanned vehicle, and compared with Sugeno fuzzy integral evaluation method which is often used in machine evaluation to validate the method proposed in this thesis. Second, traffic sign cognition ability test of unmanned vehicle has been carried out. In the study of pattern recognition, usually the common traffic sign recognition indicators include recognition accuracy, tr...
shelfnumXWLW1774
Other Identifier200928014628025
Language中文
Document Type学位论文
Identifierhttp://ir.ia.ac.cn/handle/173211/7645
Collection毕业生_硕士学位论文
Recommended Citation
GB/T 7714
张雅如. 无人驾驶车辆交通标志和泊车认知能力测试设计与评价方法研究[D]. 中国科学院自动化研究所. 中国科学院研究生院,2012.
Files in This Item:
File Name/Size DocType Version Access License
CASIA_20092801462802(3803KB) 暂不开放CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[张雅如]'s Articles
Baidu academic
Similar articles in Baidu academic
[张雅如]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[张雅如]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.