Real time multi-sensor data acquisition and processing for a road mapping system

Publication Type:
Thesis
Issue Date:
2014
Full metadata record
Files in This Item:
Filename Description Size
Thumbnail01front.pdf229.58 kB
Adobe PDF
Thumbnail02whole.pdf2.9 MB
Adobe PDF
Road assets condition has a critical impact on road safety and efficiency. Accurate and efficient monitoring and management of road assets is a challenge. This research is focused on developing a cost efficient mobile surveying system to tackle this challenge. The system is equipped with LADARs (LAser Detection And Ranging) and a camera as exteroceptive sensors, and other sensors including Inertial Measurement Units (IMU), odometer and GPS (Global Positioning System).This system can acquire road assets information expeditiously in highly dynamic environments, where data collection has previously been inefficient, laborious and even dangerous. Continuous Position, Velocity and Attitude (PVA) information is obtained by the integration of IMU, GPS, camera and odometer. Then PVA information is fused with range and remission data from LADARs to achieve multiple functions for road assets surveying and management. The functions include road clearance surveying, road surface profiling, 3D structure modelling, road boundary detection and road roughness measurement. The processing results are presented in a user-friendly graphical interface and can be saved as videos for convenient data management. Two sets of GUI (graphical user interface) have been developed for data acquisition from all the sensors and data processing for the system functions. A Data Acquisition GUI is used for sensors control, data acquisition and pre-processing. It has multiple functions, including configuring LADARs scan frequency and resolution, displaying and recording data and exporting data with the required format. The Data Processing GUI includes various algorithms to perform all the data processing and management functions. The camera in the proposed system provides not only a vision reference, but also visual odometry for improving PVA estimation when GPS is unreliable. In order to obtain a robust and accurate visual odometry, a new algorithm named PURSAC (PURposive SAmple Consensus) has been purposed for model fitting, which purposely selects sample sets according to the sensitivity analysis of a model against sampling noise and other information. This in turn can improve the accuracy and robustness of fundamental matrix estimation, resulting in a more precise and efficient visual odometry. A prototype system designed for online data processing has been developed and four road tests have been successfully completed. Experimental results on a variety of roads have demonstrated the effectiveness of the proposed mobile surveying system.
Please use this identifier to cite or link to this item: