Design Of Inspection Robot For Senior Residential District

With the continuous development of robot technology,in order to effectively play and auxiliary community the development of artificial intelligence technology,efficient management,to fully demonstrate the main function of intellectual property,guarantee the quality of residential area residents’ daily life,suitable for small and medium upscale community use intellectual property become one of the important direction of the development of the intelligence community,At present,the development of smart property has not formed a unified standard.In China,the development direction of smart property is still in a groping stage.Intelligent security is an important part of intelligent property.This paper designs an intelligent inspection robot for intelligent security system in intelligent property to realize the inspection and alarm functions in the community,so as to replace manual inspection scheme and reduce manual labor intensity.Design a patrol robot,based on the electric flat car design.The construction of lidar point cloud map provides information for robot positioning,terrain estimation and site environment analysis,which is an important part of inspection robot judgment and path planning.The deep learning algorithm is used to realize target identification and inspection tasks.Most of the current network map in digital map is not in conformity with the structured environments such as the condition of the complex task or environment,and the existence of civil digital map update cycle is long and low accuracy of roads surrounding tag shortcomings,unable to meet the needs of the future development of the inspection robot,so the design is suitable for the efficient map building method of inspection robot.For most inspection robots,the use of manual extraction and drawing severely restricts the large-scale application of robots,so it is necessary to extract regional targets from the constructed point cloud map to improve the positioning and detection ability of inspection robots in the residential environment.At the same time,deep learning technology will provide a new scheme for target discrimination in constant speed driving,and realize the patrol of abnormal situations of pedestrians,vehicles and buildings in the community environment.In order to solve this problem,the purpose of this paper is to design a high-grade residential area of the inspection robot,cooperate with various sensors by robots as the main study way,the collocation of sensor map model is set up at the same time,the inspection robot makes can be built on the basis of laser radar map provides a priori information pose computation task,the final combined with deep learning algorithm checking recognition in the community,The specific research contents are as follows:1.Designed the control of the robot based on the existing electric flat car.The method of steering control and power control by PWM signal is proposed.Firstly,the steering mechanism is modified.Shorten the length of steering rod and use high torque steering gear to control steering;The steering shaft and steering shaft are connected by flange coupling.For power control,Arduino is used to convert signals instead of Hall component signals.2.Study the application of Lidar and sensor combined with SLAM to realize the mapping function of robot.Carry the load platform Livox laser radar,the use of radar batch point cloud information,and use the LOAM algorithm implementation of roads in the area of map building,combining the feature points location information with which to start rebuilding,get the initial point cloud,and incrementally to join the rest of the point cloud data to get all the point cloud radar pose the sparse point cloud of geographical space as well as the district.The depth map of the image is calculated by the cell based point cloud updating method,finally the point cloud obtained by the depth map sorting.3.Based on the work,a method of extracting key targets from 2d real-time transmission images is improved by Gamma correction to realize the robot target recognition function.Firstly,target features such as personnel and vehicles in part of images are selected as training samples,and the recognition results are obtained by fitting the training samples according to the computational characteristics of the target.Secondly,the brightness processing algorithm is used to Gamma correct the image with abnormal light in order to remove the influence of brightness on the image accuracy.4.In order to verify the effectiveness of the proposed algorithm,a robot experimental platform for lidar point cloud construction and target recognition was built,and the efficiency and accuracy of the algorithm were verified comprehensively by using data sets.First platform vehicle refitting the reliability of the test load,and then show to build a wide range of maps and applied in the residential area in a specific area such as effect,is obtained by training image data set is suitable for the target recognition in the community data and analysis of the efficiency of the method of inspection,finally the proposed method is validated by the experiment on the effectiveness and stability of the community environment.