【作者】 阳明；

【导师】 唐新蓬；

【作者基本信息】 华中科技大学 ， 车辆工程， 2006， 硕士

【摘要】 随着汽车安全性要求的日益提高,车载传感器的数目和种类也越来越多。这些相同或不同类型的传感器所提供的局部观测信息间必然存在冗余和矛盾,因此有必要通过信息融合,将多个传感器提供的信息加以分析、综合,利用信息互补,形成对环境相对完整一致的感知描述。本文借鉴生物免疫学的基本原理,设计了一种应用于车辆多传感器信息融合的人工免疫模型。解决装有CCD摄像机和激光雷达的车辆在高速公路上行驶时的信息融合问题。即通过融合车载CCD摄像机和激光雷达实时探测得到的高速公路上本车前方的各种信息,去除干扰和冗余,最终得到前方各车所在车道、与本车车距、车速、车型等信息。根据此种信息融合所具有的特点,提出了一种基于克隆选择和阴性选择算法的两阶段融合模型。以传感器预处理信息作为抗原,最终得到的外部环境模式作为抗体。在系统接收传感器实时信息抗原后,比较抗原间的匹配度,判断两者的准确程度,如果达到准确阈值则直接将抗原基因段组合为抗体;否则,利用对已有抗体进行克隆得到的新抗体组来对抗原进行识别。抗体基因段间信息多样性则通过相似度的阴性选择计算来保证。使用MATLAB软件编写了仿真程序,并参照实际情况编写了抗原输入样本。通过人工模仿可能出现的各种情况在抗原样本中施加干扰,得到融合结果后通过比较抗原和抗体的误差率和误检测率来判断算法的优劣。仿真结果证实了算法在解决车载多传感器信息融合问题上的有效性。文章最后指出人工免疫算法在多传感器信息融合领域应用的前景,以及进一步的研究方向。更多还原

【Abstract】 With the requirement of automotive safety was getting stricter, more and more on-board sensors were used to form a system to detect environmental information. However, redundant and contradictive information from these sensors might cause the system ineffective. Thus we are trying to find an effective way to utilize the different sensors information, reduce the errors from every single sensor and form a comprehensive and exact result.Based on the fundamental principle of immunology, we designed an artificial immune model for solving the freeway-running vehicle’s multi-sensor fusion problem, which called for a method to fuse the data from CCD camera and Radar sensor, wipe out the disturbance and redundancy, and obtained the important navigating information like the line, distance and velocity of frontal automotives.For specialty of the problem, we bring forward a two-phase fusion model of the artificial immune algorithm based on the clone selection theory. We set the sensors’pre-treated data as antigens and the fusion result as antibodies. As soon as the antigens invaded the system, it reacted to calculate the match between antigens and estimate if their information were accurate enough to directly recombine the gene together from both antigens then form an antibody. If not, we utilized the clone selection arithmetic to clone a new set of antibodies and calculate the fitness between antigens and them to decide a most suitable antibody. The diversity between the gene sectors is ensured by negative selection theory.We used MATLAB software to write a simulation program and set the samples of antigen. By simulating the different errors which would happen in the sensors’input data, and analyzing the outputs’error and misdetection ratio, we proved the validity of using the artificial immune arithmetic to solve the data fusion problem.In the end, we indicated the foreground of artificial immune theory in multi-sensor fusion area and the farther potential research directions.更多还原