首页|融合旋转不变约束的哈希分类器的设计与应用

融合旋转不变约束的哈希分类器的设计与应用

扫码查看
分类器是目标识别、目标检测的重要部分,而针对遥感图像的目标检测在交通、军事、农业等方面具有重要的应用价值.但随着遥感图像的分辨率急剧增加,如何提高遥感图像目标检测的效率成为一个挑战.利用遥感目标具有的旋转不变特性以及哈希学习具有的快速分类能力,设计并实现了融合旋转不变约束的哈希分类器,其目的是使遥感目标在旋转前后具有相似的二进制哈希码.通过实验表明该分类器能在大幅提高检测速度的同时提高检测的准确性,并且可以拓展到其他哈希学习方法上.
Design and Application of Hash Classifier Incorporating Rotation Invariant Constraint
Classifier is an important part of target recognition and detection,and target detection for remote sensing images has im-portant application value in transportation,military,agriculture,and other fields.But with the sharp increase in resolution of remote sensing images,how to improve the efficiency of target detection in remote sensing images has become a challenge.This article uti-lizes the rotation invariance characteristics of remote sensing targets and the fast classification ability of hash learning to design and implement a hash classifier that integrates rotation invariance constraints.The aim is to make remote sensing targets have similar bi-nary hash codes before and after rotation.Through experiments,it has been shown that this classifier can significantly improve the detection speed while also improving the accuracy of detection,and can be extended to other hash learning methods.

classifierhash learningrotation invariantremote sensing imagesobject detectionimage processing

徐晖

展开 >

南京理工大学紫金学院,江苏 南京 210023

分类器 哈希学习 旋转不变 遥感图像 目标检测 图像处理

2024

电脑与电信
广东省对外科技交流中心

电脑与电信

影响因子:0.117
ISSN:1008-6609
年,卷(期):2024.(4)
  • 4