Research On Multi-headed Camera Extrinsic Calibration Of Camera Series In Videometrics

In the construction of large-scale projects in water conservancy and hydropower,transportation,high-rise building,mining and other fields,there are settlement changes in the foundation of such basic facilities.Serious roadbed subsidence not only affects the construction and operation safety of construction projects,which is more likely to cause irreparable damage.The demand for the real-time automatic monitoring of roadbed subsidence is in abundance,and the displacement-relay camera network videometrics is one of the effective ways to solve the problem.High precision extrinsic camera calibration of multi-headed cameras of displacement-relay camera network is a necessary task for the accurate roadbed subsidence monitoring.This paper focuses on classifying and comparing the existing methods and comprehensively evaluates the performance of the methods from many aspects by reviewing a large number of domestic and foreign literatures and systematically investigating the historical development and research status of the calibration method of extrinsic parameters of the multi camera system.The vital problem to be handled in multi-headed camera extrinsic calibration is discussed.This paper discusses the source of errors in the subgrade settlement measurement system with focus on studying the model error caused by misalignment of the double-headed cameras' optical axes and analyzing the influence of such misalignment on measurement accuracy by simulation experiments.The simulation results show that the marker image point changes caused by that misalignment are small in amount,but they can enlarge the measurement error along with the increase of the transmission series.The accuracy of the subgrade settlement is generally required to be in the sub-millimeter,for a very slight error will affect the accuracy of settlement measurement.Therefore,it is very necessary to calibrate the fix connection between double-headed cameras and use the relative relationship between them to eliminate such error.In order to solve the problem that there is non-overlapping field of view between multi-headed cameras in the displacement-relay camera chain,this paper studies two methods respectively by robot-based hand-eye calibration and using a planar mirror to reflect the auxiliary visual field.Due to the constraints of fixed connection between double-headed cameras,there is a limited range for the camera rig to move in the calibration process,which produces great limitations on the accuracy of hand-eye calibration.Compared with the traditional hand-eye calibration,the method presented in this paper and using conjugate rotation averaging is not sensitive to noise,and enjoys higher accuracy in unsatisfactory setting conditions.To make the multi-headed cameras calibration easily operating,this paper studies a multi-headed cameras calibration method with a mirror based on pose estimating by using multiple reflections.The knowledge on Cramer-Rao low bound is used to deduce the theoretical error bounds of the estimated lower pose of the auxiliary plane mirror visual field,with analysis of the impact of image point errors,number of images and the plane mirror parameters on the estimated pose accuracy.Two different calibration schemes by using planar mirror reflections are designed.The feasibility and accuracy of two calibration methods with multi-headed cameras are verified by simulation experiments and real-world experiments.In the camera measurement of the subgrade settlement,the calibration of fixed connection of multi-headed cameras is extremely important and basic.It has important research meanings to solve the problem of calibration of multi-headed cameras for effectively reducing multistage error accumulation and improving the settlement monitoring system in the subsidence displacement-relay.