增强型轻量级网络的红外小目标检测  

Infrared small target detection based on enhanced lightweight network

作　　者：刘坚[1] 江峰[1] LIU Jian;JIANG Feng(Pingxiang University,Pingxiang Jiangxi 337000,China)

机构地区：[1]萍乡学院,江西萍乡337000

出　　处：《激光杂志》2023年第4期249-253,共5页Laser Journal

基　　金：江西省科技厅自然科学基金(No.20202BABL212005)。

摘　　要：当前红外小目标检测方法存在一定的缺陷,如检测精度低,易受外界环境干扰。为了实现高精度红外小目标检测,提出基于增强型轻量级网络的红外小目标智能检测方法。该方法采集红外小目标的红外图像,采用小波变换对红外图像进行预见处理,然后提取红外小目标检测特征,并采用增强型轻量级网络技术实现红外小目标智能检测。经测试分析证明:该方法的红外小目标采集数量正确比重为1左右,受外界环境干扰较小,频谱有效带宽信号频率集中在275 kHz附近;红外小目标检测边界相对误差均为1%左右,边界绝对误差低于1 mm,检测精度高于98.2%,检测时间低于5.9 ms,提高了检测效果。The current infrared small target detection methods have certain defects,such as low detection accuracy and easy interference from the external environment.In order to achieve high-precision infrared small target detection,an intelligent detection method based on enhanced lightweight network is proposed.This method collects the infrared image of infrared small target,predicts the infrared image by wavelet transform,extracts the detection characteristics of infrared small target,and realizes the intelligent detection of infrared small target by using enhanced lightweight network technology.Test and analysis show that the correct proportion of infrared small target acquisition quantity of this method is about 1,the interference from the external environment is small,and the signal frequency of spectral effective band is concentrated near 275 kHz;The relative error of infrared small target detection boundary is about 1%,the absolute error of boundary is less than 1 mm,the detection accuracy is higher than 98.2%,and the detection time is less than 5.9 ms,which improves the detection effect.

关 键 词：增强型轻量级网络 红外小目标 检测提取 仿真测试 

分 类 号：TN292[电子电信—物理电子学]