一种FPGA实现的红外小目标检测算法  

An infrared small target detection algorithm realized on FPGA

作　　者：王岳环[1,2] 张清顺 WANG Yuehuan;ZHANG Qingshun(School of Artificial Intelligence and Automation,Huazhong University of Science and Technology,Wuhan 430074,China;National Key Laboratory of Multispectral Information Processing Technology,Huazhong University of Science and Technology,Wuhan 430074,China)

机构地区：[1]华中科技大学人工智能与自动化学院,湖北武汉430074 [2]华中科技大学多谱信息处理技术国家级重点实验室,湖北武汉430074

出　　处：《华中科技大学学报（自然科学版）》2020年第1期20-25,共6页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家重点预研计划资助项目(41415020402)

摘　　要：针对环形去最大中值滤波和帧间轨迹关联红外小目标检测算法流程与现场可编程门阵列(FPGA)并行不兼容的问题,对原有算法进行了并行化改进.主要从简化阈值计算、采用上一帧统计值进行自适应阈值分割和多帧检测循环三个方面进行了优化.在对检测性能影响较小的前提下,通过对算法结构进行优化,消除了算法中全局计算依赖,使其更适于FPGA进行算法实现.以FPGA实现优化后算法时,在80 MHz工作时钟下,处理2 800×2 800像素图像能够达到10帧/s的速度,满足大部分实时红外探测系统的要求.To solve the incompatibility between the algorithm flow and the parallelism of the FPGA(field programmable gate array),the algorithm based on the ring maximum median filter and using inter-frame trajectory correlation to detect small infrared targets was optimized in parallel.The improvement included simplifying threshold calculation,adaptive threshold segmentation based on last frame statistics,and loop unrolling for multi-frame detection.On the premise that the detection was less affected,the global computing dependence of the algorithm was eliminated by optimizing the structure of the algorithm,which was more suitable for the implementation on the FPGA.When the optimized algorithm was realized on the FPGA,the processing speed of 2 800×2 800 images can reach 10 frames/s under 80 MHz working clock,which could meet the requirements of most real-time infrared detection systems.

关 键 词：红外图像 目标检测 现场可编程门阵列(FPGA) 实时红外探测系统 

分 类 号：TP391.7[自动化与计算机技术—计算机应用技术]