Research And Implementation Of Sensor Management Framework Based On Scene Perception

Unmanned Aerial Vehicles(UAV)are playing an increasingly important role in human life,and all kinds of sensors on UAV are the main way for uUAV to obtain external information.How to manage and coordinate all kinds of sensors is the key to improve the efficiency and robustness of UAV.Different sensors have their own suitable working environment,and when the environment changes,the sensor is inefficient or even ineffective,such as that RGB sensors can not work in a dark environment.Therefore,in different environments,the amount of information and reliability provided by the sensor is varied.In order to solve the above problems,first of all,this paper proposes a kind of sensor management framework based on the class,which classifies the sensors commonly carried on the UAV at present,and endows them with the role.UAVs use frames to manage information on a wide range of sensors.They manage the properties of various types of sensors,and provide all sensors on the UAV based on sensor class management(providing interfaces for rapid replacement and configuration of different models of sensors).Secondly,this paper presents a scene perception method based on RGB channel,and a sensor group selection mechanism based on this method.The RGB channel scene perception method can detect different scenes by analyzing the RGB channel values of the obtained images.The mechanism is divided into three modules.The scene perception module perceives the scene,and the sensor selection module quickly selects,coordinates and matches various sensors according to the perceived results,forming the sensor selection decision.The sensor startup module matches the sensor's capabilities according to the above decision,turning the relevant sensor on or off.Finally,in order to verify the above framework and mechanism,we implement a sensor selection prototype system based on scene perception.The system can perceive light intensity and internal and external environment,and then select the appropriate sensor group.We run this prototype system on a Turtlebot equipped with Kinect sensor and RPLIDAR,and the experimental results show that under the framework of sensor management,the mechanism can activate the corresponding sensor group according to the results of the scene detection module,and enhance the robustness of UAV in complex environment.