基于机器学习的低信噪比图像序列小目标检测  

Small Object Detection in Image Sequences with Low Signal-to-Noise Ratio Based on Machine Learning

作　　者：马丽娟[1] 黄勇[1] 李艳翠 MA Li-juan;HUANG Yong;LI Yan-cui(School of Information Engineering,Henan Institute of Science and Technology,Xinxiang Henan 453000,China;College of Computer and Information,Henan Normal University,Xinxiang Henan 453007,China)

机构地区：[1]河南科技学院信息工程学院,河南新乡453000 [2]河南师范大学计算机与信息工程学院,河南新乡453007

出　　处：《计算机仿真》2023年第10期219-223,共5页Computer Simulation

基　　金：2022年度河南省重点研发与推广专项(科技攻关)项目(222102210020)。

摘　　要：低信噪比图像中的噪声会对图像序列小目标检测结果造成影响,为了提升检测精度,提出基于机器学习的低噪声比图像序列小目标检测方法。针对低信噪比图像中的噪声,对图像实行预处理,以此降低外界因素影响,提升检测效果;根据预处理结果采用Harris角点检测方法提取图像中的全部候选目标区域,将HOG算法与机器学习中的支持向量分类器结合,对低信噪比图像序列内的小目标展开检测,从中检测出真实小目标,实现最终检测。实验结果表明,通过对上述方法实行小目标检测结果与实际结果对比测试、目标丢失率测试,验证了上述方法的检测可靠性高、实用性强。Noise in the images with low-signal-to-noise ratio will affect the detection results of small targets in image sequences.In order to improve the detection accuracy,this article put forward a method of detecting small targets in the image sequence with low noise ratio based on machine learning.For the noise,we preprocessed the image with low signal-to-noise ratio at first,and thus reducing the influence of external factors and improving the detection effect.According to the preprocessing results,we used the Harris corner detection method to extract all candidate regions in the image.After that,we combined the HOG algorithm with the support vector classifier in machine learning to detect the small targets in low signal-to-noise image sequence,thus finding out real small targets.Finally,the detection was completed.Experimental results show that the method has high detection reliability and strong practicability by comparing the small target detection results with the actual results and testing the target loss rate.

关 键 词：机器学习 低信噪比 图像序列 小目标检测 支持向量分类器 角点检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]