| With the development of automobile technology,the computer revolution and the widespread application of artificial intelligence technology,the research work of self-driving car technology has also developed,and environment perception technology is considered to be the foundation of modern self-driving car technology.However,the composition of traffic participants in the campus scenario is different from that of urban roads,which makes autonomous vehicles face certain potential risks when driving on campus.In terms of obstacle detection in autonomous driving,lidar technology is regarded as one of the most important information collection devices due to its advantages of high measurement accuracy and strong adaptability.In this paper,combined with the current environmental perception technology of self-driving cars,a method is proposed to quickly and effectively detect pedestrians and cars in the process of self-driving cars driving on campus.The main work and contributions of this thesis include:(1)In the process of ground segmentation,an improved method of ground segmentation is proposed.Divide the point cloud map into two-dimensional grids with the center of the vehicle body as the origin of coordinates,and pack them into the corresponding grids according to the X and Y coordinates of the radar reflection points;divide the parts according to the maximum height value of the reflection points in the grid Ground grid;calculate the maximum difference of the heights of the reflection points in the remaining grids to divide some obstacle grids;divide the ground,obstacles and longitudinal slope grids according to the variance of the heights of all the reflection points in the remaining grids.The above methods are experimentally verified.The method proposed in this paper can maintain the segmentation accuracy of the original flat road section and improve the removal efficiency of the reflected wave on the slope road section.The edge stone for example)performs well and can meet the requirements of autonomous vehicles driving on campus.(2)In the process of obstacle detection,an improved clustering method is proposed based on Euclidean clustering method.Based on the Euclidean clustering method,set different area radii,set different clustering radii in different areas,first divide the areas according to the area radius to generate their respective cluster sets,and then combine the areas to calculate the cluster set center,according to The distance and angle between the centroids of adjacent cluster sets are selectively combined,and then the echo intensity of people and cars is statistically analyzed,and the type of cluster set is judged by combining the difference of echo intensity of two types of obstacle reflection points.,calculate the length,width and height of the clustering target,and finally combine all the above information to generate a Bounding Box model.The above methods are experimentally verified,and the results show that the method proposed in this paper can accurately capture the positions of pedestrians and cars,and generate box models with corresponding colors,which can meet the requirements of autonomous vehicles driving on campus. |
