一种旋转不变的目标识别算法  

A rotation-invariant target recognition algorithm

作　　者：陈国玺 姜振东 杨贝贝 张姗姗 CHEN Guoxi;JIANG Zhendong;YANG Beibei;ZHANG Shanshan(School of Mechanical Engineering,Shangqiu Institute of Technology,Shangqiu 476000,Henan,China)

机构地区：[1]商丘工学院机械工程学院,河南商丘476000

出　　处：《农业装备与车辆工程》2025年第1期129-133,共5页Agricultural Equipment & Vehicle Engineering

摘　　要：为了解决目标图像旋转角度造成的误匹配问题并提高算法的运行速度,提出了一种旋转不变的目标识别算法。该算法融合了圆投影和小波变换的优点,通过圆环面投影方法对图像进行预处理,保证特征的旋转不变性;采用小波变换对目标图像和模板图像进行降维,加快算法的运行速度;选择基于方差结构相似性算法作为图像特征量进行目标区域识别,提高目标区域的识别精度。将不同背景环境下的图像与传统目标识别算法以及其他旋转不变算法相比较,结果表明,所提算法在目标识别过程中不仅能够解决目标图像的旋转问题,而且能够提高算法的运行速度,识别准确率超过94%,适用于较复杂背景下目标图像的识别。In order to solve the problem of mismatching caused by the rotation angle of the target image,and improve the running speed of the algorithm,a rotation-invariant target recognition algorithm was proposed,which combined the advantages of circular projection and wavelet transform.The algorithm preprocessed the image by the circular projection method to ensure the rotation invariance of the feature;and used the wavelet transform to reduce the dimension of the target image and the template image,so as to speed up the running speed of the algorithm.Then the variance-based structural similarity index was selected as the image feature quantity to identify the target area,and the recognition accuracy of the target area was improved.By experimenting with images in different background environments,compared with traditional target recognition algorithms and other rotation-invariant algorithms,the results showed that the algorithm could not only solve the problem of target image rotation in the target recognition process,but also improve the operation of the algorithm speed,suitable for the recognition of target images under more complex backgrounds.Moreover,the recognition precision was up to 94%.

关 键 词：目标识别 旋转不变 圆环投影 小波变换 基于最小方差的结构相似性算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]