基于类脑仿生的环境感知技术  

Environmental Perception Technology Based on Brain Like Bionics

作　　者：高昊 王仁茂 GAO Hao;WANG Renmao(The 723 Institute of CSSC,Yangzhou 225101,China;Military Respresentative Office of Shanghai Bureau of Naval Armament Department in Yangzhou,Yangzhou 225101,China)

机构地区：[1]中国船舶集团有限公司第七二三研究所,江苏扬州225101 [2]海装上海局驻扬州地区军事代表室,江苏扬州225101

出　　处：《舰船电子对抗》2024年第5期42-46,55,共6页Shipboard Electronic Countermeasure

摘　　要：针对机器人即时定位与地图构建(SLAM)领域环境感知中场景图像表征难的问题,仿照人脑视觉系统感知环境的过程,提出了一种类脑仿生的环境感知方法,将场景图像压缩成视觉模板以表征场景。首先模拟大脑先验知识,利用尺度不变特征变换(SIFT)算法提取通用场景数据集中图像的特征,并通过K均值聚类(K-Means)算法来构建视觉特征库,然后模拟视觉注意机制,提取目标场景图像的SIFT特征并将其与视觉特征库进行匹配,完成场景表征。最后通过实验验证了方法的可信性。In response to the problem of difficult scene image representation in the field of robot simultaneous localization and mapping(SLAM)environment perception,this paper proposes a brain like environment perception method by simulating the process of human visual system percering the environment.The method compresses scene images into visual templates to represent the scene.Firstly,this paper simulates prior knowledge of the brain,use scale-invatiant feature transform(SIFT)algorithm to extract features of images in a general scene dataset,and constructs a visual feature library through K-Means algorithm;then simulates visual attention mechanism,extracts SIFT features of target scene images and matches them with visual feature library to complete scene representation.Finally,the reliability of the method is verified through experiments.

关 键 词：即时定位与地图构建(SLAM)环境感知 类脑仿生 尺度不变特征变换(SIFT) 场景表征 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TP18[自动化与计算机技术—计算机科学与技术]