|
Research objectives:
One way to determine the limitations of a sensor is to characterize its performance. Characterization is the process of experimentally and/or theoretically determining the factors that affect the quality of the sensory data. By identifying and quantifying these factors, viable applications for the range sensor can be decided. Also, this process precedes the development of a correction model for the sensor that can be used for correcting the degrading factors. The identification process and the resulting correction model are the focus of this research work. |
 |
This thesis presents an error characterization scheme with application to a pulsed time-of-flight laser rangefinder called the AccuRange 3000. Characterizations of range-imaging sensors are important because they determine the limitations of the sensor. Also, characterizations are the first step in developing a correction model for range data and they provide information about the amount of errors present in the data to aid in post-processing techniques. The combination of all these factors provides a basis for making a decision on which type of sensor to use for a particular application. The entire system is simulated using Telegrip simulation software. An error characterization of the laser rangefinder, utilized as a point laser, is given. This includes errors associated with temperature, sampling rates, mixed pixels, varying values of reflectance at incremental distances, and varying values of reflectance at incremental angles of incidence. Selected information gathered during these experiments is combined to form a correction model for the laser system. This correction model is based on the range and intensity data output from the laser rangefinder. A comparison of corrected and uncorrected range values is given for several data sets. A confidence metric is also derived and given for each corrected range value.
This work was conducted by R. Collier while at IRIS lab under supervision of R. Whitaker (Thesis Chair) and M. A. Abidi. This work was supported by DOE's University Research Program in Robotics under grant DOE-DE-FG02-86NE37968.