基于嵌入式图像处理及路径规划的智能导盲仪  

Intelligent Blind Guide Device Based on Embedded Image Processing and Path Planning

作　　者：杨馥语 陈娟[1] YANG Fuyu;CHEN Juan(School of Information Science and Engineering,Xinjiang University,Urumqi 830017,Xinjiang,China)

机构地区：[1]新疆大学信息科学与工程学院,新疆乌鲁木齐830017

出　　处：《流体测量与控制》2022年第3期65-68,共4页Fluid Measurement & Control

基　　金：国家级大学生创新创业训练资助项目(202110755113)。

摘　　要：当前导盲仪存在功能单一、不能动态规划路径等问题,因此研发了一款基于嵌入式的智能导盲仪。选用语音识别模块,将语音输入的目的地信息转换成文字,选用高德地图API进行开发,并结合语音合成(TTS)芯片,对实时规划的路径进行语音播报;基于车辆、行人等可移动式目标,展开基于YOLOv5的目标检测,训练适合所述场景的权重,用于监测复杂路况;选取改进的Deeplabv3+语义分割算法,通过对图像像素点阈值进行判断,检测斑马线、车道线等交通标识物,用于引导盲人定位并通过人行横道;采用HC-SR04超声波测距模块,该模块提供2~400 cm的非接触式距离感测功能,可以近距离地帮助盲人躲避危险;选取树莓派4B作为主控核心,通过将各模块搭载到linux系统中进行运算,实现导盲仪的总体功能。The current blind guide device has problems such as single function and inability to dynamically plan the path. In this paper,an embedded intelligent blind guide device is developed. The speech recognition module is used to convert the destination information entered by the voice into text,and the AutoNavi map API is used for development,and combined with the speech synthesis chip,the real-time planned path is broadcasted by voice;based on movable targets such as vehicles and pedestrians,the development based on YOLOv5 target detection,train weights suitable for the scene described in the article to monitor complex road conditions;select the improved Deeplabv3+ semantic segmentation algorithm to detect zebra crossings,lane lines and other traffic signs by judging the threshold of image pixel points,which are used for guide the blind to locate and pass the pedestrian crossing;adopt HC-SR04 ultrasonic ranging module,which provides 2-400 cm non-contact distance sensing function,which can help blind people avoid danger at close range;choose Raspberry Pi 4b as the main control core,by loading each module into the linux system for operation,the overall function of the guide for the blind is realized.

关 键 词：导盲仪 实时路径规划 斑马线识别 超声波避障 语音识别与合成 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]