| Displacement is an important index for structural status and safety assessment and geohazard early warning,and the development of low-cost,high-efficiency and high-precision displacement monitoring technology is an important basis for building a universal and specialized engineering safety and geohazard early warning system.Monocular visual displacement monitoring is a new monitoring technology that extracts displacement by tracking the motion of deformers in video or images,demonstrating great superiority in automated,high-precision,non-contact and safety displacement monitoring.With the rapid development of computer hardware processing capability and graphic image processing algorithms,monocular visual displacement monitoring has become an important supplement to modern displacement monitoring technology.However,influenced by the low accuracy of measurement marker detection and localization algorithm,thermal deformation and unstable motion of the visual system,atmospheric turbulence,and camera hardware conditions,the existing monocular visual displacement monitoring technology and method cannot meet the application requirements of submillimeter accuracy,wide-range and multi-point,and long-distance monitoring.To this end,this paper aims to improve the accuracy and engineering practicality of monocular visual displacement monitoring,and on the basis of in-depth analysis of the main influencing factors of monocular visual displacement monitoring,research on high-precision monocular visual displacement monitoring techniques and methods,and develop and realize the high-performance monocular visual displacement monitoring system that can be applied in engineering,so as to serve the needs of high-precision and practical displacement monitoring applications.The specific research work and results of this paper are as follows.(1)Through theoretical analysis and experimental verification,this paper summarizes four main factors that limit the development and application of monocular visual displacement monitoring technology: the detection and localization error of measurement markers,the thermal deformation and unstable motion error of the visual system,the atmospheric turbulence error,and the imaging limitation of the camera.Further detailed and in-depth analysis shows that:(1)The degradation of the imaging quality of measurement markers is mainly manifested as image noise,blurring,distortion and shadow occlusion,which can seriously affect the detection efficiency and localization accuracy of measurement markers;(2)The thermal deformation of the visual system is the result of the coaction of temperature changes and camera pose changes,and the camera pose is unknown and uncontrollable in practical application,so it is difficult to achieve robust and high-precision thermal deformation error correction by using the temperature compensation model.(3)The visual system may still produce creep deformation due to thermal expansion and contraction of materials and self-weight even though it is firmly fixed,and such deformation is especially obvious during long-term measurement.(4)The spatial distribution of image distortion and displacement measurement errors under the influence of atmospheric turbulence have obvious randomness,so it is difficult to construct mathematical models to correct for atmospheric turbulence errors.(5)The limitations of the field of view,depth of field and resolution of conventional cameras make it difficult to achieve wide-range,high-resolution imaging,resulting in the efficient data acquisition and multi-point simultaneous monitoring performance of existing monocular vision systems is greatly reduced.(2)A high-precision detection and localization method of crossshaped markers based on local Radon transform is proposed,which effectively weakens the influence of image noise,blurring,distortion deformation and uneven illumination on marker detection and localization,and achieves reliable detection and precise localization of measurement markers under complex imaging conditions.Cross-shaped markers are commonly used as artificial markers in the field of vision measurement because of their simple patterns,clear center position and eccentricitydifference-free.To this end,this paper proposes to apply the local Radon transform to the saliency enhancement of the graphic features of crossshaped markers such as central symmetry and high contrast of the blackand-white sector of cross-shaped markers,and establish the mapping relationship from the central corner of the cross-shaped markers of the original image to the peak spot of the saliency response map,and the optimal localization is determined with the help of the surface peak fitting algorithm.The method is validated with computer simulated images,ground station and UAV images respectively,and the results show that the localization accuracy of the method is better than 0.2 pixels under the complex imaging conditions,which is suitable for high-precision displacement monitoring applications.(3)An instability error correction method for monocular vision systems that takes into account both unstable motion and thermal deformation is proposed to significantly improve the measurement accuracy and stability of monocular vision systems under complex working conditions.In this paper,an imaging strategy that adopts the principle of collimated light was introduced into the monocular vision system,and a parallel light-monocular and dual-view imaging model is constructed with the help of the biprism.The model uses the homography constraint to estimate the image deformation,which can effectively eliminate the displacement measurement errors caused by unstable motion and thermal deformation.The results of indoor and outdoor experiments show that the system can reduce the instability error caused by temperature variation and strong wind from 2.13 mm to 0.1 mm.The correction rate of instability error can reach more than 90%.(4)A high-precision correction method of atmospheric turbulence error based on multi-view observation of microlens array and spatial multipath displacement averaging is proposed.The spatial distribution of image distortion caused by atmospheric turbulence is random,so "spatial averaging" instead of "temporal averaging" can effectively avoid the influence of time-domain or frequency-domain methods on the time-series displacement results.Firstly,this paper introduce the principle of multiview imaging by microlens array,construct a monocular vision multi-view imaging model based on frontal microlens array to realize the multi-path segmentation of the imaging beam in space,which provides the necessary theoretical basis to fully utilize the spatial random characteristics of the image deformation distribution under the influence of atmospheric turbulence;secondly,with the multi-path displacement results obtained by single-aperture multi-view imaging,the atmospheric turbulence error is corrected by multi-path displacement averaging.As shown by the experimental analysis of indoor and outdoor displacement measurements,the correction rate of the method for the atmospheric turbulence error can be more than 50 %.Compared with the time-series sliding window averaging method,the method is less likely to lose the detailed information of the time-series displacement data and is more suitable for the processing of high-frequency displacement data.Compared with the frequency domain filtering method,the method can effectively deal with the displacement signals mixed with different frequencies,and has better applicability and flexibility.(5)A wide-range and multi-point displacement monitoring method based on monocular Scheimpflug imaging is proposed,which breaks the depth of field and field of view limitations of conventional monocular vision systems,and improves the performance of monocular vision for wide-range,multi-point displacement monitoring.Firstly,this paper introduces the principle of the Scheimpflug imaging to optimize the conventional monocular imaging,which can greatly expand the imaging depth of field without reducing the image magnification,thus making the monocular vision have the imaging characteristics of both high resolution and large depth of field;secondly,to address the problem of tedious and time-consuming adjustment process of Scheimpflug imaging,a precision Scheimpflug imaging model is derived,and an installation method of Scheimpflug camera based on this model is proposed,which effectively improves the convenience of using the Scheimpflug camera,and the quantitative installation process also helps to improve the standardization of monocular Scheimpflug measurement scheme;finally,based on the imaging characteristics that the image plane of the Scheimpflug is not perpendicular to the optical axis of the lens,the translation and rotation model of the Scheimpflug camera is derived,and a high-precision calculation method of the displacement of multiple measurement points is proposed.The experimental validation and result analysis show that the monocular axis-shifting imaging system developed in this paper has a root mean square error(RMSE)of displacement better than 0.54 mm at a measurement distance of 160 m,and also has excellent performance of wide-range and multi-point monitoring.(6)For the purpose of engineering practicality,a high-performance monocular visual displacement monitoring system based on the research results of this paper is designed and implemented,and a bridge multi-point displacement monitoring scheme based on this system is designed.The hardware of the system adopts a modular design idea,in which the parallel light imaging for correcting the system instability error,the single-aperture multi-view imaging for correcting the atmospheric turbulence error,and the Scheimpflug imaging for wide-range and multi-point measurement are packaged into separate modules,and each module can be used in combination or independently.The system software is designed and developed around the key functions such as camera remote control and wireless data transmission,and the overall architecture based on the userside software and camera-side software is built by combining the highprecision detection and localization methods for cross-shaped markers proposed in this paper.The Yueliangdao highway bridge in Changsha City is selected for field test,and the results show that under complex working conditions,the system can accurately detect the multi-point dynamic response of small and medium span bridges with small displacement and high frequency,high accuracy,high sensitivity and multi-point simultaneous monitoring are its main advantages. |
PDF Full Text Request