未知环境有害气体烟羽跟踪问题研究

项目名称： 未知环境有害气体烟羽跟踪问题研究

项目编号： No.61201081

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 电子学与信息系统

项目作者： 贾云伟

作者单位： 天津理工大学

项目金额： 25万元

中文摘要： 有害气体不仅破坏生态环境、危害人体健康，还易引发爆炸、火灾等重特大事故，研究未知环境有害气体烟羽跟踪问题，探索未知环境感知规律，可为实际工况有害气体监测寻踪提供依据，为机器人在未知环境下作业提供指导，具有重要的科学与社会意义。首先，针对实际工况中湍流、风速、视觉等信息均未知的情况提出分层渐进跟踪思想；其次，基于SIFT算法进行分层多尺度特征点提取，通过研究特征点参数对特征分区的影响，揭示未知环境视觉感知规律；再次，为提高跟踪效率，创建嗅觉信息与视觉信息的深度融合模型，仿真分析各种因素对融合结果的影响，探索未知环境嗅视觉感知规律；最后，构建烟羽跟踪策略进行未知环境有害气体烟羽跟踪实验研究，通过分析不同跟踪策略的实验结果，总结各种因素对烟羽跟踪的影响，验证和完善嗅视觉信息深度融合模型，进而推理未知环境感知规律。

中文关键词： 气体传感；烟羽跟踪；环境感知；机器视觉；融合

英文摘要： It is important to explore the law of environmental perception by applying harmful gas plume tracing method in unknown environment, because harmful gases not only have adverse effect on human life and the environment, but also often cause serious accident. This study will contribute to effective gas monitoring in working condition and provide reference for mobile robot which works in unknown environment. Firstly, the layered tracing idea is proposed for unknown environment. Secondly, the hierarchical multi-scale feature points are detected by SIFT algorithm, and the relationship between environmental segmentation and feature extraction is investigated so as to exposit the law of visual environmental perception. Thirdly, the depth fusion model of the olfactory information and the visual information is constructed, the influence factors are analyzed so as to improve the gas tracing efficiency and explore the law of olfactory and visual fusion environmental perception. Finally, the experiments are carried out by the layered plume tracing method, and the results are compared with the other data that are made by other tracing algorithms. Then, the depth fusion model is modified and the law of environmental perception is deduced.

英文关键词： Gas Sensing；Plume Tracing；Environmental Perception；Machine Vision；Fusion