融合视觉信息的语义导航实体搜索算法

为解决家庭服务机器人在实体搜索缺乏语义信息的问题,研究融合视觉信息的语义导航实体搜索算法。首先,提出基于Places205-AlexNet的场景自适应识别算法,以促进机器人对场景标注的自适应性;其次,设计实体在语义地图上的位置映射方法,帮助机器人能以人的生活习惯理解人类居住环境,具备室内物体识别能力;然后,为了获取场景信息和物体语义信息,提出了面向语义导航的实体语义知识库构建方法,使机器人进行实体搜索时能更容易定位到场景;最后,在给出基于面向语义导航的实体语义知识库构建方法的基础上,提出了融合视觉信息的语义导航实体搜索算法。实验数据表明,机器人执行实体搜索的导航准确率均在0.75以上,这证明所提出的算法能够获得较准确的场景信息和实体语义信息。

Semantic Navigation Entity Search Algorithm Integrating Visual Information

In order to solve the problem that the home service robot lacks semantic information in entity search,this paper studies the semantic navigation entity search algorithm which integrates visual information.Firstly,an adaptive scene recognition algorithm based on Places205-AlexNet was proposed to promote the robot's adaptability to scene labeling.Secondly,the position mapping method of the entity on the semantic map is designed to help the robot to understand the human environment based on people's living habits and have the ability of indoor object recognition.Then,in order to obtain scene information and object semantic information,a semantic navigation-oriented entity semantic knowledge base construction method was proposed to make it easier for the robot to locate the scene during entity search.Finally,based on the construction method of entity semantic knowledge base based on semantic navigation,a semantic navigation entity search algorithm combining visual information is proposed.Experimental data show that the navigation accuracy of the robot in entity search is above 0.75,which proves that the proposed algorithm can obtain more accurate scene information and entity semantic information.