Research On Visual SLAM Technology Based On NI FlexRIO

With the continuous development of science and technology,the scope of interests between countries has also developed from the traditional territory,territorial sea and airspace to a deeper level of space.Exploring the space above the blue sky has also become a national strategic task.The exploration of the resources contained in each planet in space is also particularly important,which requires the space probe to have good performance.Therefore,each space probe must go through a large number of rigorous and repeated ground experiments to verify its reliability before launching.As a reliable visual measurement method,SLAM technology can realize high-precision real-time positioning and measurement of its own motion trajectory through sensors in unknown environment,which can be well adapted to the ground experiment of space detectors.This paper mainly studies the visual SLAM technology based on the NI Flex RIO platform.Aiming at the time-consuming feature extraction and feature matching part of SLAM technology,we use the excellent hardware processing ability of NI Flex RIO platform to achieve algorithm acceleration.On this basis,in order to improve the extraction speed and adaptability of image feature points,an adaptive ORB feature point to extraction method based on dynamic local threshold is proposed based on the ORB feature algorithm.The number of image feature points extracted by the adaptive ORB algorithm is not too many,which is easy to extract,and has strong robustness.The same feature point can be extracted in different directions of the image,and the threshold value of feature points is calculated by dynamic threshold method,so that the extraction results will be the same under different illumination,and the adaptability to the environment is better than other algorithms.In addition,this paper also proposes a BRIEF matching method suitable for hardware computing,which greatly simplifies the matching process and reduces the time consumed by feature matching.The feature point information and matching results calculated by the hardware platform will be transmitted to the upper computer.Through processing this part of data,the upper computer will complete a series of work such as pose estimation,pose optimization,closed-loop detection,and finally get the pose information and trajectory drawing of the moving camera.In this paper,a large number of experiments are carried out with the open data sets of TUM,which verify that the feature points extracted by the adaptive ORB feature algorithm have higher stability and robustness than the traditional algorithm,and the feature points with good quality will also play an important role in feature matching.In addition,using the evaluation tool of the data set of the TUM,the camera track datas obtained from the SLAM system based on the adaptive ORB algorithm is analyzed,and the absolute track error is analyzed with the real track data provided by the TUM,which verifies the feasibility of the system at the software level.Finally,the hardware design based on NI Flex RIO platform is completed throughLab VIEW FPGA module,and the SLAM system design based on NI Flex RIO is completed with the upper computer.