摘要本文毕业设计是属于社区安保服务机器人团队，团队总体工作是设计制作用于社区安保服务的机器人，本文负责的部分是使用激光雷达进行障碍物检测。论文包括三个部分：对激光雷达采集的点云数据进行预处理并基于点云特征进行初步分割、运用凝聚层次聚类算法对点云数据进一步分割、将基于信息熵的免疫遗传算法对聚类进行优化提取障碍物信息。87273
本文使用激光雷达进行了三次实验，初步试验获取的点云数据进行预处理和基于点云特征进行初步分割，实验一和实验二获取的点云数据进行聚类分割和基于信息熵的免疫遗传算法优化聚类分割。
点云数据的预处理主要包含误差处理、噪声处理以及坐标转换，对于激光雷达提取的点云数据是有序的，采用插补的方法进行误差处理，采用高斯滤波的方法进行噪声处理，可以很好的保持数据原貌；点云数据特征主要包含坐标、高程、斜率、曲率，对基于点云数据特征进行初步分割提取边界点。
本文提出聚类对障碍物信息进行更细化的提取描述，使用的是凝聚层次聚类对点云数据进行分割，利用图像显示，不同的障碍物标记不同的颜色，同一障碍物标记同一颜色，检测出清晰的障碍物信息。
因为聚类分割存在误差，所以分割效果不是很理想，本文提出将免疫遗传算法优化聚类分割，从而可以获得更好的聚类信息，由后两个实验提取的数据进行测试前后的对比，可以明显发现优化后的聚类效果比较理想，障碍物检测的效果较好。
毕业论文关键词：激光雷达；障碍物检测；点云数据；预处理；聚类算法；免疫遗传算法

Abstract   This graduation design is to belong to the community security service robot team。 Team's overall work is design for community security service robots, responsible for part of this article is to use laser radar obstacle detection。 Thesis includes three parts: preprocessing point cloud of laser radar acquisition and preliminary segmentation based on the characteristics of point cloud, condensed hierarchical clustering algorithm is used to point cloud data further segmentation, based on the information entropy of immune genetic algorithm to optimize clustering extraction of obstacles information。
   This article uses the laser radar on three experiments, the preliminary test to get point cloud data in pretreatment and preliminary segmentation based on point cloud characteristics, experiment 1 and experiment 2 get to clustering segmentation of point cloud data and the immune genetic algorithm based on information entropy optimization clustering segmentation。
   Point cloud data preprocessing mainly includes error handling, noise processing and coordinate transformation, for the extracting of lidar point cloud data is ordered, adopt the method of interpolation error in principle, adopt the method of gaussian filtering noise processing, this can be very good to keep original data; Characteristics of point cloud data mainly includes the coordinates, elevation, slope, curvature, to preliminary segmentation based on point cloud data provided to extract the boundary point。
   In this paper, clustering to extract the obstacle information more detailed description,  using the coacervate clustering segmentation of point cloud data, using the image display, different obstacles marker a different color, the same obstacles tag the same color, clearly detect obstacles。
   Because of clustering segmentation error, so the segmentation effect is not very ideal, this paper puts forward the immune genetic optimization clustering segmentation method, which can obtain better clustering information, contrast before and after the data extracted by two experiments to test, can be found that the optimized clustering effect is more ideal, obstacle detection effect is better。 聚类激光雷达的社区安保服务机器人障碍物检测:http://www.youerw.com/zidonghua/lunwen_131026.html