Research On Navigating Technology Of Indoor Inspection Robot Based On Vision

In nuclear power plants,steel mills,chemical plants,oil depots,substations and other dangerous operating environment,to ensure safe and stable operation of all equipments,regular and uninterrupted inspection is essential.At present,the main inspection method is manual inspection.Manual inspection is inefficient,at the same time,long-term work in high-risk environment,the human body can easily cause harm.Therefore,there is an urgent need to develop an indoor inspection robot under dangerous working conditions to assist or replace manual inspection.For indoor inspection robots,autonomous walking is the prerequisite for the completion of inspection tasks.The research and development of the navigation system is of great significance for improving the intelligent level of the indoor inspection robot and improving the efficiency of the inspection.Aiming at the requirements of indoor navigation system,we design the overall design requirements and schemes,determine the software and hardware frame structure of the robot body and distributed vision system,and complete the design of the robot control system,communication system,perception system,power system,drive system and decision system.According to the design,we complete the selection of hardware devices,including global cameras,inspection cameras,motors,drives,infrared obstacle avoidance sensors,routers,switches,the underlying drive,serial port Ethernet port module,and complete the installation and debugging of these devices.At the same time,we completed the construction of the software system,including the design and compilation of the host computer interface software and the underlying driver driver.Finally,a vision-based indoor inspection robot navigation system is designed,which includes an indoor inspection robot,a distributed vision system,an upper computer software.Aiming at the problem of global path planning of indoor inspection robots,the genetic algorithms,the clonal selection algorithm and the ant colony algorithm are mainly studied.First of all,we elaborate on the principle of the above bionic intelligent algorithm and analyze the flow of the algorithm.Then we compare the performance of the three algorithms with the classical test functions of Shekel's Foxholes,Shubert,Hansen,Bohachevsky,Ackley,Sphere Model and so on.Function test results show that ant colony optimization algorithm has the strongest searching ability,the best convergence performance and the most stable algorithm performance.Then,in order to further verify the performance of the three algorithms in the robot path planning,we compare and test the three algorithms by TSP.TSP test results show that ant colony algorithm has the strongest ability to obtain the shortest path.Through analyzing simulation experiments and comparing the advantages and disadvantages of each algorithm,we finally determine to use ant colony algorithm for indoor inspection robot global path planning.Considering the complexity and difficulty of parameter selection of ant colony algorithm and the coupling characteristics of each parameter,we propose a parameter selection of ant colony algorithm method based on bacterial foraging algorithm.First,we map the parameters of the ant colony algorithm to a multidimensional space,and use the chemotaxis operator to approximate each parameter to the optimal value,so as to accelerate the convergence speed of each parameter.Then,we speed up the optimization of the whole set of parameters by the propagating operator.Finally,we use the migration operator to strengthen the global optimization of parameters,and avoid the parameters falling into the local optimal solution.In order to verify the effectiveness of the proposed method,the robot path planning problem is simulated and compared with the ant colony algorithm parameter selection method based on genetic algorithm and particle swarm optimization algorithm.The results show that the optimal parameters of ant colony algorithm based on bacterial foraging algorithm can find the best combination of parameters quickly and accurately,which shows the effectiveness and superiority of this method.Aiming at the key technologies of image processing of indoor inspection robots vision system,this paper has carried out in-depth research on the related algorithms and carried out experimental verification.First,we complete the calibration of the camera and correct the camera distortion image.Then,according to the needs of work space for indoor inspection robots and the parameters of the global surveillance camera,we obtain the number of cameras required for the layout and complete the construction of a distributed vision system.And,with the square target,we finished the monocular camera range.Then,by introducing SIFT,SURF,ORB algorithm,we use the image mosaic technology to fuse the images of multiple cameras to get the global map and compare the performance of the three algorithms.The results of image stitching test show that the ORB algorithm has the fastest stitching speed and the best effect.After the preprocessing of the global image,we use the findContours()function in OpenCV to find out the outline of the image and draw the outline using the drawContours()function.Then,we inflate and fill the image to get a binary image with "0-1",which is converted into a raster map for robot path planning.Finally,using the image template matching technology,taking the robot body image as the template image and the image acquired by the global surveillance camera as the target image,we obtain the position of the robot in the target image and realize the positioning and tracking of the mobile robot.In order to verify the performance of the indoor inspection robot navigation system,we tested the robot navigation system in a laboratory environment.First,we completed the exercise performance test of indoor inspection robots,including straight-line driving deviation,driving speed,turning radius and turning speed in situ.By testing the performance of the robot,we test the stability of the underlying robot driver and understand the robot's motion performance to better control the robot.This is the precondition and foundation of the robot's navigation test.In the actual navigation process,in order to improve the efficiency of the algorithm,we didn't stitching all the images together,but we used two cameras as a group to do related operations.In this way,we can not only improve the efficiency of the algorithm,but also avoid the accumulation of robot motion errors increasing,so as to improve the efficiency of autonomous inspection of indoor inspection robots.We test the communication system of the navigation system and the results show that the information exchange between the host computer and the underlying controller is rapid and the video transmission does not suffer from the phenomena of stuttering and delay.We test the software performance of the host computer interface of the navigation system and complete the operation of video image acquisition,image splicing,obstacle detection and identification,grid map acquisition,path planning,path generation and execution,and robot positioning and tracking.The results show that the performance of the software is stable.In the laboratory test environment,the software of the interface software of the upper computer can finish the operation instructions better and the indoor inspection robot can realize the autonomous navigation inspection.