激光点云解算可配置阵列设计与实现

Design and implementation of reconfigurable array for laser point cloud resolving

激光雷达通过点云解算处理获取扫描目标的三维点云图,点云解算算法复杂,SoC以其优越的处理性能为点云解算的实时处理提供了解决方案,但其应用功能设计与输入数据特性是紧耦合的,输入数据的变化将导致需重新进行FPGA设计,来应对资源消耗和布线情况的改变。为降低系统与输入数据的耦合性并提高系统可扩展性,对点云解算处理各步骤采用流水线设计,利用AXI-4协议设计了基于可配置路由的并行阵列架构。实验结果表明,该设计方法能够通过简单的参数配置,实现灵活高效的点云解算多阵列处理设计。

Lidar obtains the 3D point cloud image of the scanned target through point cloud solution pro⁃cessing.The point cloud solution algorithm is complicated.SoC provides a solution for the real⁃time process⁃ing of point cloud solution with its superior processing performance,but its application function design and input data characteristics are tightly coupled,and changes in input data will lead to the need to re⁃design the FPGA to cope with changes in resource consumption and wiring conditions.In order to reduce the coupling between the system and the input data and improve the scalability of the system,a pipeline design is adopted for each step of the point cloud solution processing,and a parallel array architecture based on configurable routing is designed using the AXI-4 protocol.Experimental results show that this design method can realize a flexible and efficient point cloud solution multi⁃array processing design through simple parameter configuration.