文献类型：会议论文

英文题名：Research on the Application of Semantic Segmentation of driverless vehicles in Park Scene

作者：Ren, Lijun[1];Liu, Yuansheng[2]

第一作者：Ren, Lijun

通讯作者：Ren, LJ[1]

机构：[1]Beijing Union Univ, Beijing Key Lab Informat Serv Engn, Beijing, Peoples R China;[2]Beijing Union Univ, Coll Robot, Beijing, Peoples R China

第一机构：北京联合大学北京市信息服务工程重点实验室

通讯机构：[1]corresponding author), Beijing Union Univ, Beijing Key Lab Informat Serv Engn, Beijing, Peoples R China.|[11417103]北京联合大学北京市信息服务工程重点实验室;[11417]北京联合大学;

会议论文集：13th International Symposium on Computational Intelligence and Design (ISCID)

会议日期：DEC 12-13, 2020

会议地点：Hangzhou, PEOPLES R CHINA

语种：英文

外文关键词：driverless; semantic segmentation; PFPN; model pruning

摘要：Semantic segmentation is widely used in the establishment of semantic map, but a good semantic segmentation model still has the disadvantage of insufficient real-time performance for driverless applications, especially in the park. In this paper, an improved PFPN (Panoptic Feature Pyramid Network) network model is proposed to reduce the time of semantic segmentation. In this algorithm, the instance segmentation branch function in the original PFPN is trimmed and the small target feature layer extracted from the semantic segmentation branch is pruned, so as to reduce the semantic segmentation time of the model. In order to verify the correctness and feasibility of the model, a test environment is set up in the campus scene. The experimental results show that the improved lightweight semantic segmentation model achieves better segmentation results, and the average image segmentation time is reduced by 22.2%.