摘要
为减少弹载图像传感器在生产、储存以及工作过程中产生的坏点,本文分析了坏点的成因和固有特性,提出了一种改进FAST角点检测的坏点检测校正算法.算法通过降低检测核半径和设定动态判断阈值等方法提高坏点检测的准确率,减少角点误检数量,并提高对连续坏点及边缘坏点的适应性,能在校正坏点的同时保留图像中的角点和边缘特征.利用Vivado HLS高层次综合工具,实现坏点检测校正算法的并行流水线设计,使资源消耗和时序达到最优;将综合生成的坏点检测校正IP核移植到PL端实现,实现弹载相机坏点的实时矫正.
In order to reduce the bad points during the production,storage and operation of the image sensor,this paper analyzes the causes and inherent characteristics of the dead pixels,and proposes an improved FAST corner point detection for dead pixels correction.The algorithm can improve the accuracy of dead pixels detection by reducing the detection radius and setting dynamic judgment threshold,which can reduce the number of corner point misdetection,and improve the adaptability to continuous dead pixels and edge dead pixels.It can retain the corner point and edge features of the image while correcting the dead pixels.Vivado HLS tool is used to realize the parallel pipeline design of the algorithm,and achieve the optimal resource consumption and timing.The IP core of the dead pixels detection and correction is transplanted to the PL terminal to realize the real-time correction of the dead pixels of the missile camera.
作者
田宗浩
刘桢
陈栋
TIAN Zong-hao;LIU Zhen;CHEN Dong(Army Academy of Artillery and Air Defense,Hefei 230031,Anhui,China)
出处
《微电子学与计算机》
2021年第2期56-61,共6页
Microelectronics & Computer
基金
军队“十三五”预研基金项目(301070103)。
关键词
弹载相机
坏点
FAST算法
HLS
Missile camera
Dead pixels
Features from accelerated segment test
High level synthesis
