基于SVM图像分割方法的自动导盲系统  

An Automatic Navigation System Based on SVM Image Segmentation for the Blind

作　　者：田亚男[1] 雷红玮[1] 王旭[1] 

机构地区：[1]东北大学信息科学与工程学院,辽宁沈阳110004

出　　处：《东北大学学报（自然科学版）》2010年第12期1681-1684,共4页Journal of Northeastern University(Natural Science)

基　　金：国家自然科学基金资助项目(60970157)

摘　　要：传统的自动导盲系统直接将图像映射为声音模式,使得声音模式中包含很多背景信息,干扰了盲人对图像的理解.针对这一问题,提出了一种基于SVM图像分割方法的自动导盲系统.首先使用SVM算法将摄像头采集到的环境图像分割为目标和背景两部分,并用灰度变换增强目标消除背景.然后将目标映射为声音模式.实验表明,经过训练,盲人能够理解系统合成的声音模式,从而形成对目标形状、尺寸以及位置的感知.In the conventional automatic navigation systems,images are directly mapped into those auditory representations which cover a lot of background information interfering in the blind＇s understanding of images.An automatic navigation system is therefore presented for the blind based on SVM image segmentation,where the images taken by a video camera are all segmented into two parts via SVM,i.e.object and background,and the former is enhanced with the latter removed by gray level transformation.Then,the object is mapped into auditory representation.Experimental results showed that the blind is able to understand the auditory display and form the cognition for the shape,size and position of an object after training.

关 键 词：电子行走辅助 视觉代替 图像可听化 听觉显示 支持向量机 图像分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]