基于RDK的目标识别算法研究

Research on Target Recognition Algorithm Based on RDK

摘要: 目前，目标检测算法的应用场景不易灵活变动，机器人在执行复杂问题时，面临较高的计算和存储资源的问题，以及机器人的可扩展性较低的问题。针对机器人进行目标检测方面的扩展，提出用基于RDK云服务机器人实现多场景式的目标检测，基于YOLOv5优化的目标检测算法对环境中的多种物体进行识别。云机器人服务的分布式部署可提高资源的利用率。通过编辑蓝图调用算法模型和基于Blender设计的动作，Smart Voice语音系统可以调用整个蓝图事件，完成数据采集、目标导航、物体识别、机器人运动等多种工作，达到实现基础服务功能的目标。

Abstract: At present, the application scenarios of target detection algorithms are not flexible, and robots face the problems of high computing and storage resources and low scalability when executing complex problems. Aiming at the expansion of robot target detection, this paper proposes to use RDK-based cloud service robots to achieve multi-scene target detection, and recognize multiple objects in the environment based on YOLOv5 optimized target detection algorithm. Distributed deployment of cloud robot services can improve resource utilization. By editing the blueprint calls the algorithm model and actions based on Blender design, Smart Voice voice system can call the whole blueprint event to complete data collection, target navigation, object recognition, robot movement and other work, and achieve the goal of realizing basic service functions.