基于多模感知和移动互联网协作的导盲系统关键问题研究

The traffic situation in urban transportation system is quite complex and changeable. For the visually impaired with the traditional blind guiding tool that could not provide adequate information, it is very dangerous to go outside. This project aims to resolve several key issues of advanced mobile blind guiding system by effectively leveraging multimode awareness, computer vision and crowd sensing as well as artificial intelligence. Firstly, according to complicated and volatile traffic status, the navigation for blind road will be realized via real-time information capturing, obstacle detection and analysis by using multimode awareness sensors of smart phones and artificial intelligence to detect and analyze obstacles in road. Furthermore, the navigation for special road will be realized via crossroad recognition, Zebra crossings detection and veering rectification based on computer version, sensor technology as well as ergonomics. Then, the crowd sensing technology will be also incorporated, and under the assistance of the mobile internet, it can sense and transmit obstacle information, transportation density and emergency accidents in time. This contributes into the real-time traffic situation updating and path planning and thus can greatly increase the robustness and intelligence of the blind guiding system. Finally, via building non-intrusive and modularized experimental platforms, the GPS or BEIDOU based mobile blind guiding system will be constructed and tested. This system can provide guiding service for the visually impaired at normal, blind and special roads, respectively, aiming at resolving significant navigation problems for the visually impaired with a precision approaching ten even one meter. This project possesses significant theoretical research value, social care meaning and competitive application prospects.

城市交通实时路况复杂多变，视障人士依靠传统导盲方式出行，获取的信息有限，安全性难以得到有效保障。本项目将多模感知、计算机视觉、群智感知和人工智能相融合，开展移动导盲系统中的关键问题研究：1）针对复杂多变的路面状况，利用智能手机多模态感知器和人工智能对路况进行实时捕捉、障碍检测，提出规避方法，解决盲道路面导航问题。2）利用计算机视觉、传感器技术和人体工程学原理，进行路口识别、斑马线检测和偏移校正，解决特殊路段导航问题。3）通过移动互联网协作，引入群智感知，实时感知和传送障碍信息、交通密度和突发状况，进行实时路况更新和路径规划，提升系统的鲁棒性和智能化程度。4）建立非侵入式、模块化实验平台，搭建基于GPS或北斗导航的导盲系统，提供普通路段、盲道和特殊路段导航服务，为视障人士解决最后10米甚至最后1米的精确导航问题，具有重要的理论研究价值、社会关怀意义和极具竞争力的应用前景。

视觉障碍者由于视觉辨识能力的缺失，往往只能借助导盲基建和辅助设备进行户外活动。传统辅助设备信息获取有限，视障人士的出行和自由被大大限制，人身安全也不能得到有效保障。因此本项目将多模感知、计算机视觉、群智感知和人工智能相融合，开展移动导盲系统中的关键问题研究：1）针对复杂多变的路面状况，利用智能手机多模态感知器和人工智能对路况进行实时捕捉、障碍检测，提出规避方法，解决盲道路面导航问题。2）利用计算机视觉、传感器技术和人体工程学原理，进行路口识别、斑马线检测和偏移校正，解决特殊路段导航问题。3）通过移动互联网协作，引入群智感知，实时感知和传送障碍信息、交通密度和突发状况，进行实时路况更新和路径规划，提升系统的鲁棒性和智能化程度。4）建立非侵入式、模块化实验平台，搭建基于GPS或北斗导航的导盲系统，提供普通路段、盲道和特殊路段导航服务，为视障人士解决最后10米甚至最后1米的精确导航问题，具有重要的理论研究价值、社会关怀意义和极具竞争力的应用前景。其次该项目也致力于人工智能及信息器件，在非线性系统稳定性、检测与分析、新型信息器件和智能信息处理等领域取得了一些重要的研究成果。建立了新型纳米级传感器和感知系统，发展了非线性函数重构准则和卡尔曼滤波器，实现了实时检测、智能分析与高效处理。设计了时滞、多涡卷混沌电路，实现了保密通信和伪随机数序列产生方案。建立基于纳米器件的人工神经网络，实现了多值存储和智能控制，发展了类脑智能信息处理模型。将忆阻器与控制理论相结合，构建性能更优的忆阻控制器，为类脑智能系统提供控制与算法支撑。项目负责人获得“万人计划”人才称号、2019年中国产学研合作创新成果奖一等奖等奖励；团队在国际国内杂志上发表相关学术论文65篇，其中SCI检索41篇，EI检索56篇，申请专利25项，其中授权专利15项；项目组积极推动产学研合作，获得了较大的经济和社会效益。