Research On Positioning Technology Of Visual Simultaneous Localization And Mapping In Nuclear Radiation Scene

The nuclear radiation scene is full of radioactive particles,and not only humans can not directly operate and even various sensors such as cameras will be affected.The development of image processing has become more and more mature.It can remove noise from image acquisition from the time and frequency domains,but cannot integrate environmental information.SLAM,as a robot that locates itself in an unknown environment and builds incremental maps,realizes autonomous navigation and positioning of the robot.It is widely used in logistics and service industries.The vision SLAM mapping technology with cameras as the main sensors is becoming more mature.This paper combines image processing and visual SLAM technology with traditional mobile robot technology,and uses visual detection technology to realize mobile robot relocation in nuclear environment.This article first introduces the latest developments in image denoising of nuclear radiation environment and SLAM,expounds the basic principle of monocular camera imaging,analyzes and resolves camera distortion.Aiming at the environmental impact of the SLAM system in the nuclear radiation environment,a set of algorithms including image preprocessing,SLAM mapping,and global relocation were proposed.Finally,some deficiencies and difficulties in the current visual SLAM recognition methods are proposed.In image processing,the causes of image noise and nuclear radiation noise are analyzed first;secondly,filtering methods such as adaptive filtering,median filtering,and wavelet transform denoising are introduced from the time and frequency domain filtering,and the comparison of operation speed and accuracy is performed Finally,a new histogram filtering algorithm based on three RGB channels is proposed to achieve the basic elimination of nuclear radiation noise.In terms of SLAM mapping,first a new visual odometer based on the traditional ORB-SLAM is proposed,which includes the feature method and the direct method;then the key frame method is used to complete the construction of the global map;the dictionary and the bag of words model are used to implement closed-loopdetection;Based on the traditional BA method,a new incremental BA method is proposed to optimize the camera pose and feature point spatial position.In global relocation,the Pn P method is used to optimize the matching results between the currently obtained feature points and bag-of-words feature points to achieve robot relocation in a nuclear radiation environment.Through experiments and tests,the entire algorithm proposed in this paper can effectively achieve the effect of global relocation in the data set and nuclear radiation environment.In the data set verification,the minimum root mean square error of global map construction can reach 0.0070 m,the minimum relocation error is 0.0039 m,the real-time performance is met and the maximum speed is 65 frames.