面向FPGA的连通域快速标记方法

FPGA-Oriented Fast Connected Component Labeling Method

针对遥感领域检测海面舰船目标的应用中,连通域标记算法消耗更多资源与时间的问题,提出一种逐像素扫描的流水线式的连通域快速标记方法及其FPGA(Field Programmable Gate Array)硬件设计。该方法通过对图像进行一次逐像素扫描,统计各行的游程信息并进行整理,之后通过遍历已统计的游程即可得到各个连通域的统计信息,从而提取特征。通过优化设计,该方法可以充分发挥FPGA的并行处理优势,只进行一次像素扫描,不产生标记表,使得方法的运行时间减少,资源消耗变小,满足实时性要求。

In the application of detecting ship on the surface of sea in the remote sensing field,the connected component labeling algorithm consumes lots of resources and time.A pixel-by-pixel scanning pipelined fast connected component labeling method is presented,and its hardware based on Field Programmable Gate Array(FPGA)is designed.The method performs a pixel-by-pixel scan on the image,and collects and edits the runs information of each row,and then obtains statistical information of each connected component by scanning the calculated runs,thereby extracting features.By optimizing the design,this method takes full advantage of the parallel processing of FPGA.The pixel-by-pixel scan is performed once,and no markup table is generated,so that the running time of the method is reduced,and the resource consumption is reduced.This method satisfies the requirement of real-time.