室内空间视觉移动目标特征提取方法仿真  

Simulation of Moving Target Feature Extraction Method in Indoor Space Vision

作　　者：吴一非[1] 杜尚广[2] WU Yi-fei;DU Shang-guang(Jiangxi university of Technology,Jiangxi Nanchang 330000,China;School of Life Sciences,Nanchang University,Jiangxi Nanchang 330000,China)

机构地区：[1]江西科技学院,江西南昌330000 [2]南昌大学生命科学学院,江西南昌330000

出　　处：《计算机仿真》2022年第10期234-238,共5页Computer Simulation

基　　金：江西省教学改革课题(JXJG-19-24-9)。

摘　　要：目前室内空间视觉移动目标特征提取方法没有对原始数据进行预处理,存在阈值对错误率影响程度高、重复率高和特征点完整性低的问题。提出基于VR的室内空间视觉移动目标特征提取方法,首先对原始移动目标数据进行归一化和背景分割的预处理,其次在主成分分析法的帮助下利用预处理后的数据计算出移动目标中的主成分,并在其中添加KPCA思想提高提取性能,最终根据权重得出信息量最多的特征点,并和阈值进行比较,得到最终结果,实现室内空间视觉移动目标特征提取。实验结果表明,所提方法的阈值对错误率影响程度低、重复率低和特征点完整性高。Due to the fact that the present the feature extraction methods for moving objects in indoor space vision do not preprocess the original data, there are problems such as high impact of threshold on error rate, high repetition rate and low integrity of feature points. Based on VR,a method to extract features of visual moving target in interior space was proposed. Firstly, the original moving target data were preprocessed by normalization and background segmentation. Secondly, with the help of the principal component analysis method, the principal components of the moving target were calculated using the preprocessed data, and the KPCA idea was added to improve the extraction performance. Finally, the feature point with the most information was obtained according to the weight, and compared with the threshold value to obtain the final result, which realized the feature extraction of moving targets in indoor space vision. Experimental results prove that the threshold of the proposed method has low influence on the error rate, low repetition rate and high feature point integrity.

关 键 词：虚拟现实技术 预处理 归一化 特征提取 主成分分析 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]