Research On Indoor 3D Thermal Modeling And Perception Technology In Service Robots

Affected by the aging of the population,the indoor mobile service robot strategy has been promoted to a national strategy.It is considered to be a powerful assistant for the elderly and accompanying work,but its intelligence level can't meet the needs of people.The environment-perception technology of robots is one of the key technologies to realize its intelligence,so the perceptual technology has always been a hot issue in the field of service robots.At the same time,SLAM(Simultaneous Localization and Mapping)technology has made great progress,but with the widespread and long-term use of home robots,robots' perceived ability of the environmental and security is far from meeting the target of operation,the need for services such as environmental fire judgment and security inspection functions.Therefore,this paper uses RGB,depth and thermal infrared sensors for the research about mobile robot environment sensing technology,and enhances the intelligence and depth capabilities of the robot environment perception.The specific contents are as follows:Research on the heterogeneous sensor registration technology of mobile robots to realize the multi-dimensional information perception consistency of robots.The RGB-D camera and the thermal imaging sensor are combined to form the vision system of the robot,and the calibration template is designed to complete the intrinsic calibration of the RGB camera and the thermal imaging sensor simultaneously,the extrinsic calibration of the two cameras is realized by the line feature,and the image fusion technology is used to display the four-dimensional information(R,G,B,T)by three channels.Based on the ORB-SLAM2 algorithm,the RGB-DT SLAM framework is constructed,which realizes the construction of the 3D environment thermal field map,and displays the multi-level information such as the ambient temperature field and the clear shape and texture of the object.In order to compensate for the poor stability of SLAM based on RGB image extracting ORB feature,the new scheme of robot pose estimation is realized by using thermal infrared image and depth image which are less affected by light to improve robot positioning stability.In order to improve the understanding t of service robots for the environment,this paper uses semantic segmentation technology to achieve the location and recognition of target objects.Firstly,the convolutional neural network is used to realize the segmentation and localization of the target object in the two-dimensional image,and then the semantic information of the object is correlated with the temperature information to display the subject temperature of the object,and then the pose obtained by the SLAM algorithm will have temperature semantics.The two-dimensional image of the information is projected to obtain a three-dimensional temperature semantic map,which improves the intelligence degree of the robot and the depth of the environment perception.Finally,this paper uses omni-directional mobile robot as an experimental platform to verify the stability of RGB-D and thermal imaging sensor registration technology and the accuracy and real-time performance of 3D environmental thermal field map construction,and verify the use of thermal infrared and depth images.The accuracy of robot positioning is finally carried out,and finally the feasibility of target location recognition technology is verified by constructing a temperature semantic map.