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。