| Mobile mapping systems integrated multiple sensors such as lidar,integrated navigation system,and camera sensor are being widely used in various fields such as surveying and mapping,intelligent driving,etc.The diversification of sensors has greatly promoted the rapid development of mobile measurement technology.The heterogeneous data collected by different sensors has brought serious challenges to the integrated processing of data by MMS.How to realize the effective fusion of the heterogeneous data acquired by the sensors in the MMS is the key question facing the development of mobile measurement technology.Aiming at the problem about high-precision fusion of heterogeneous data(3D laser point cloud,2D image)in MMS,in this dissertation,relevant research is carried out from the two key aspects of system calibration and heterogeneous data fusion.Based on the basic theory and the high-precision calibration algorithm of the MMS,the integrated data model of MMS is established through the high-precision fusion of heterogeneous data to realize the data effective integration and information shared.The main research contents and innovations points of the dissertation are as follows:(1)A boresight self-calibration model MMS based on multi-feature constraints is constructed,which solves the problems of poor adaptability of single-feature and low degree of automation of the calibration process,and enhances the robustness of the calibration model According to the influence law of boresight error,starting from the calibration model with single-feature constraint,the boresight calibration model based on the reference plane constraint and the boresight calibration model based on spherical feature are proposed.The experimental results show that the single-feature constrained calibration model can effectively solve the boresight error and obtain a higher calibration accuracy,but this calibration method has poor universality for MMS integrating different types of lidar sensors.In order to improve the universality and stability of the calibration model,combined with automatic feature extraction algorithm,the paper constructed a self-calibration model of boresight error based on multi-feature constraints.When constructing the calibration model,multiple features are used for common constraints,and the arrangement method is introduced to complete the constraint matching between the features to ensure the equivalence of the observations on the calibration results.The calibration model is verified by using four types of MMS that integrate sensors with different accuracy.Experimental results show that this multi-feature self-calibration methord can effectively eliminate the boresight error,and has good adaptability and reliability.Finally,according to the law of covariance propagation,it is theoretically verified that the reliability of the multi-feature constrained calibration model is better than the single-feature constrained calibration model,with higher accuracy and better adaptability;(2)A robust camera internal parameter calibration method based on total least squares is proposed,and a calibration device is designed to overcome the problem of precise extraction of corresponding features between the image(2D)and the laser point cloud(3D).To solve the problem of local optimal solution caused by too many parameters in the internal parameter calibration method,and there are errors in the observed values and coefficient arrays,the direct linear transformation method and the spatial resection method are combined to effectively solve the local optimal solution problem;By constructing the totall least squares and using the standard deviation of the pixel distance residuals as the threshold to eliminate gross errors in the solution process,a reliable estimate of the camera parameters is obtained.Based on the accurate calibration results of the the internal parameters of the camera,in order to realize the integrated data model of MMS,three-dimensional point cloud data as a reference,and the bevel is the connecting bridge.To combination point cloud automatic algorithm and image recognition algorithm to solves the extraction of corresponding features between point clouds(3D)and image(2D).Finally,the MMS with integrated single camera and panoramic camera were used for experimental verification,the results showed that the standard deviation in the calibration of the external camera parameters was better than 1 pixel;(3)A method of image data fusion based on time series is proposed,and the problem of lack of true color point cloud caused by mismatch of camera field of view is solved by neighborhood image repair technology,and high precision fusion of laser point cloud and image data is realized.According to the time synchronization of the mobile measurement system and the camera pulse trigger mechanism,based on the GNSS time triggered by the image,combined with the continuity of laser point cloud acquisition and the interval of image acquisition,through ime-domain segmentation method,high-precision fusion of laser point cloud and image is realized.The MMS with integrated single camera and panoramic camera were used for experimental verification,the results showed that the proposed fusion method is feasible and reliable.Based on the high-precision camera external parameter calibration technology,the neighborhood image repair mechanism is designed to solve the problem of lack of true color point cloud information caused by the mismatch of the sensor's field of view,effectively improving the completeness of the true color point cloud.(4)A panoramic stereo measurement model with collinear relationship in spherical coordinate system is established.Combining the principle of depth camera,a single panoramic measurement model based on distance image is proposed.Based on the calibration of the mobile measurement system and high-precision fusion results,a panoramic stereo measurement model based on the collinear relationship under the spherical system is constructed for the panoramic image measurement problem,combined with the traditional photogrammetric measurement theory.Through experimental analysis,the standard deviation of the panoramic stereo measurement model within a range of 15 meters is 2.3 cm.Considering the complex interaction of panoramic stereo measurement and the influence of intersection angle on the measurement accuracy,combined with the depth camera principle and the support of point cloud reverse fusion,a single panoramic measurement model based on distance image is proposed.This method uses reverse fusion of three-dimensional laser point clouds to provide distance information for image data,and realizes sublimation of image data from 2D to 3D dimensions.Through experimental evaluation,the single panoramic measurement model is theoretically feasible and reliable,and the standard deviation of the measurement is 3.8cm,which can fully meet the needs of panoramic measurement in surveying and mapping. |
PDF Full Text Request