基于嵌入式的多机制识别技术在ITS中的应用研究
发布时间：2018-08-24 15:02
【摘要】：随着城市道路交通管控的需要，道路交通信息的相对匮乏正促使着交通信息全面感知技术的快速发展。在自由流状态下，对当前某一特定车辆的精准识别是交通感知领域的一个重大研究课题，如何借鉴人类大脑的认知和融合机理，利用多机制传感信息，有效融合并识别当前指定车辆，从而达到精准无误的效果，已成为车辆自动识别的一个研究热点。 本文在充分研究和总结车辆视频识别、射频识别、电感传感识别、模糊理论及多源信息融合的研究现状和基本理论的基础上，提出了基于嵌入式双核架构的车辆多机制识别应用模型，并对车辆多机制融合识别算法、复杂情况下的车牌定位算法及嵌入式QT平台的车牌识别系统设计与实现等关键问题进行了深入研究。主要研究工作包括： （1）提出了基于证据理论的多特征融合车牌定位算法。该算法通过对多个候选车牌区域分别提取宽高比、纹理密度和色调主值等显著性特征，，通过证据理论融合各特征置信度，识别出真正的车牌区域，最终实现车牌定位，更加适应复杂情况下的车牌定位。 （2）提出了基于模糊理论和证据理论的多机制传感信息融合车辆识别算法。该算法通过提取由视频、射频及电感传感器识别的车辆特征信息，采用基于贴近度的模糊识别算法求出待识车辆对标准模式车辆的相似度，并将其作为各识别机制的基本概率分配，送往信息融合中心，通过证据理论进一步融合，最终输出融合后的目标车辆识别结果，克服了单一车辆识别机制存在的固有缺陷。 （3）完成了基于嵌入式QT的车牌识别系统的设计。以Linux为平台，以QT为开发工具，实现了ARM嵌入式平台抓拍图像的读取显示，灰度变换，二值化、膨胀腐蚀、平滑滤波、边缘提取等相关图像处理操作，最终实现车牌定位功能。 仿真实验表明，所提出的算法是可行的。基于证据理论的多特征融合车牌定位算法明显优于单特征定位效果，提高了车牌定位准确率；基于模糊理论和证据理论的多机制传感信息融合车辆识别算法能够给出可信度更高的识别结果。
[Abstract]:With the need of urban road traffic control, the relative lack of road traffic information is promoting the rapid development of comprehensive traffic information perception technology. In the condition of free flow, the accurate recognition of a particular vehicle is an important research topic in the field of traffic perception. How to use the mechanism of cognition and fusion of human brain for reference and use multi-mechanism sensing information for reference is an important research topic in the field of traffic perception. Effective fusion and recognition of current designated vehicles, so as to achieve accurate results, has become a research hotspot in automatic vehicle recognition. In this paper, the research status and basic theories of vehicle video identification, radio frequency identification, inductance sensor identification, fuzzy theory and multi-source information fusion are fully studied and summarized. An application model of vehicle multi-mechanism recognition based on embedded dual-core architecture is proposed, and the vehicle multi-mechanism fusion recognition algorithm is proposed. The key problems such as the algorithm of license plate location and the design and implementation of the vehicle license plate recognition system based on embedded QT platform are studied in detail. The main research work includes: (1) A multi-feature fusion license plate location algorithm based on evidence theory is proposed. The algorithm extracts salient features such as ratio of width to height texture density and color principal value from several candidate license plate regions. Through the evidence theory fusion of each feature confidence the true license plate region can be recognized and finally the license plate location can be realized. It is more suitable for vehicle license plate location in complex situations. (2) A multi-mechanism sensor information fusion vehicle recognition algorithm based on fuzzy theory and evidence theory is proposed. By extracting the vehicle feature information identified by video, radio frequency and inductance sensors, the fuzzy recognition algorithm based on closeness degree is used to obtain the similarity of vehicles to standard mode vehicles. As the basic probability distribution of each recognition mechanism, it is sent to the information fusion center, and finally the target vehicle recognition result is output through the further fusion of the evidence theory. It overcomes the inherent defects of single vehicle recognition mechanism. (3) the design of vehicle license plate recognition system based on embedded QT is completed. With Linux as the platform and QT as the development tool, the image reading and displaying, gray level transformation, binarization, expansion corrosion, smoothing filtering, edge extraction and other related image processing operations are realized on the ARM embedded platform. Finally, the license plate location function is realized. Simulation results show that the proposed algorithm is feasible. The multi-feature fusion license plate location algorithm based on evidence theory is obviously superior to the single feature location algorithm, and improves the accuracy of license plate location. The multi-mechanism sensor information fusion vehicle recognition algorithm based on fuzzy theory and evidence theory can give a more reliable recognition result.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP391.41;TP368.1
本文编号：2201193
[Abstract]:With the need of urban road traffic control, the relative lack of road traffic information is promoting the rapid development of comprehensive traffic information perception technology. In the condition of free flow, the accurate recognition of a particular vehicle is an important research topic in the field of traffic perception. How to use the mechanism of cognition and fusion of human brain for reference and use multi-mechanism sensing information for reference is an important research topic in the field of traffic perception. Effective fusion and recognition of current designated vehicles, so as to achieve accurate results, has become a research hotspot in automatic vehicle recognition. In this paper, the research status and basic theories of vehicle video identification, radio frequency identification, inductance sensor identification, fuzzy theory and multi-source information fusion are fully studied and summarized. An application model of vehicle multi-mechanism recognition based on embedded dual-core architecture is proposed, and the vehicle multi-mechanism fusion recognition algorithm is proposed. The key problems such as the algorithm of license plate location and the design and implementation of the vehicle license plate recognition system based on embedded QT platform are studied in detail. The main research work includes: (1) A multi-feature fusion license plate location algorithm based on evidence theory is proposed. The algorithm extracts salient features such as ratio of width to height texture density and color principal value from several candidate license plate regions. Through the evidence theory fusion of each feature confidence the true license plate region can be recognized and finally the license plate location can be realized. It is more suitable for vehicle license plate location in complex situations. (2) A multi-mechanism sensor information fusion vehicle recognition algorithm based on fuzzy theory and evidence theory is proposed. By extracting the vehicle feature information identified by video, radio frequency and inductance sensors, the fuzzy recognition algorithm based on closeness degree is used to obtain the similarity of vehicles to standard mode vehicles. As the basic probability distribution of each recognition mechanism, it is sent to the information fusion center, and finally the target vehicle recognition result is output through the further fusion of the evidence theory. It overcomes the inherent defects of single vehicle recognition mechanism. (3) the design of vehicle license plate recognition system based on embedded QT is completed. With Linux as the platform and QT as the development tool, the image reading and displaying, gray level transformation, binarization, expansion corrosion, smoothing filtering, edge extraction and other related image processing operations are realized on the ARM embedded platform. Finally, the license plate location function is realized. Simulation results show that the proposed algorithm is feasible. The multi-feature fusion license plate location algorithm based on evidence theory is obviously superior to the single feature location algorithm, and improves the accuracy of license plate location. The multi-mechanism sensor information fusion vehicle recognition algorithm based on fuzzy theory and evidence theory can give a more reliable recognition result.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP391.41;TP368.1
【引证文献】
相关硕士学位论文 前1条
1 秦悦;BTM测试管理系统与测试关键技术的研究[D];北京交通大学;2013年
本文编号：2201193
本文链接：https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/2201193.html
