Research On Vision Location Of Mobile Robot In Unstructured Environment

With the continuous development of sensor technology,computer technology and artificial intelligence technology,mobile robot research is becoming more and more independence and intelligence.In an unknown environment,mobile robots need to rely on the sensors to determine their own pose,which is the basis of mobile robots to perform tasks.The vision sensor has the advantages of small size,low power consumption,low cost,and rich information,so that the vision-based positioning method has always been a hotspot in the field of mobile robot location.As a systematic problem with the actual connection,the positioning of the mobile robot needs to take into account the characteristics of the sensor,the environment in which it is located and the specific form of the positioning algorithm.Based on the two mobile robots developed by our team,this paper focuses on the specific needs of the mobile robot in the task.Combined with the existing problems of mobile robot localization algorithm,this paper studies the problem of the hovering of the mobile robot based on vision,the visual orientation of the hydraulic quadruped robot relative to the front guide,and the visual positioning of the two mobile robots in the process of moving relative to the surrounding environment.The specific contents are as follows:Firstly,the theory of vision is described systematically.It includes camera imaging model,binocular stereo vision and polar line correction.Then,the system of mobile robot object studied in this paper is introduced,the choice of the system and related parameters related to vision,which is the theoretical basis and physical basis of the follow-up study and of great significance to the subsequent research.Secondly,a kind of autonomous hovering method of flying robot based on monocular vision is introduced.Using the camera mounted under the cage to estimate the three-dimensional position of the aircraft relative to the ground without the aid of other sensors such as barometers,inertial measurement units(IMU),and so on.The camera is installed under the cage to capture the ground image as a benchmark.The change in position is detected by calculating a change in the corresponding point between the subsequent frame and the reference image.And then processed by a Kalman filter and sent to the PID controller to control the hover of the cage aircraft.Finally,the performance of the whole system is tested experimentally.The results show that this method can make the flying robot to maintain a stable hover.Thirdly,the problem of automatic accompanying of hydraulic quadruped robot in the field environment is studied.For the wild environment,the instructor's camouflage is difficult to identify.So a special sign is designed to solve this problem.At the same time,the monocular and stereoscopic vision localization method are studied,and the overall optimal estimate of parallax is obtained by contour detection and dynamic programming.And the Newton iterative method is used to estimate the optimal pose.On this basis,the advantages and disadvantages of monocular vision localization method and binocular stereoscopic vision method in such environment are analyzed,and the fusion method of two methods is proposed.Finally,the feasibility of this method is proved by simulation and physical experiments.Fourthly,a hybrid loop detection method based on NDT is proposed to detect the loop of mobile robot in SLAM process.The algorithm first obtains the initial position by matching the RGB-D data by normal distribution transform(NDT).In the loop detection,the space-based method is used to lock the initial candidate matching object,which reduces the time-consuming of global search.And then finalizes the candidate object according to the method of appearance similarity(DBoW2),and finally selects an optimal loop.These two methods of loopback detection can achieve faster loopback detection speed,and because of the improved loop screening mechanism,it is ensured that the time of loop detection does not increase obviously with the increase of the frames.