Calibration of a Terrestrial Laser Scanner - Institute of Geodesy and ...

108 5. Kinematic Laser Scanningonlyin a forward direction but also in a backward direction.The combination of forward and backwardfiltering leads to optimal smoothing. The smoothing algorithm are classified in three groups [Jekeli, 2001].Fixed-point smoothing estimates only the system state at a fixed epoch. Fixed-lag smoothing is applied onestimating the systemstate for a fixed, i.e. constant,time delay in the past. Fixed-interval smoothing usesall information, i.e.observations, of the interval for an estimation of the system state at each epoch.Since the Kaiman filter is applied in the post-processing, fixed-interval smoothing can be applied. Severalversions of the smoothing algorithms were developed. One of them is the filter according to [Fraser, 1967],which uses a combination of forward-in-time and backward-in-time filter.The forward-in-time filter corre¬sponds to the above discussed Kaiman filter, whereas the backward-in-time filter starts from the last epochand runs to the first epoch by using the same Kaiman filter equations with careful implementationindices. The optimal estimation is obtained by building the weighted averageof timeof the forward and backwardfilter. Further information regarding Kaiman filter and smooting algorithms can be found in [Gelb, 1974]and [Jekeli, 2001].Analogous to adjustment techniques, blunder detections can be carried out accordingto [Baarda, 1968]and [Pope, 1975]). The innovation i is appropriate as a parameterfor a so-called local test that followsa normal distribution.Unfortunately, local tests are not sensitive enough to properly detect all possibleerrors.Additionally, global tests, which use the information of several foregoing epoches instead of onlyone foregoing epoch, should be applied. One global test is introduced by [Teunissen and Salzmann, 1988].Before Kaiman introduced his filter in the 1960s, the Wiener filter technique was widelyfiltering is a method that recovers the original signal as close as possible from the received signalequation (5.14) by minimizing the mean-square error. It is designed to operate directlyused. The Wienerbased onfor each estimateon all of the data. The Kaiman filter, instead, recursively conditions the current estimate on all of the pastmeasurements.5.4 SynchronisationThe data acquired by the laser scanner and the total station have to be synchronized. For the laser scanner,a rotation time was derived, cf. Section 5.2, which defines a relative time tag for each measurement. Sincethe laser scanner and the total station are controlled by one and the same computer,the common timebase is provided by the frequency of the high-resolution performance counter of the computerused. Thefrequency cannot change while the system is running.Each received data sent from the total station islinked by a time tag derived by the performance counter of the computer. The data received bythe laserscanner are not linked with the time tag since the rotation time of the rotating mirror provides a highresolutioncounter.Only for the operation of starting the scanning process is a time tag generated. Thus,the relative time tag of the rotation time can be linked with the absolute time tag.The time point of starting the operation of the scanningmode of the laser scanner does not fitthe timepoint of the first measured point of the first profile. The reason is there is a first (short) time delay betweensending the command of starting the profile mode from the computer to the laser scanner. Then, a second(long) time delay is given by starting the profile mode and acquiring the first data point. The rotating mirrorhas to be accelerated to the angular velocity, and if the angular velocity is reached, the data are acquired.This time delay from sending the command of starting the profile mode up to measuring the first point canbe described by a constant time-offset that can be calibrated, cf. Section 5.2.2.The synchronisation of multi-sensor systems is an important aspect. The use of several sensors requiresreal-time operating systems and knowledge of programming languages. Software packages, e.g. Dasylab®,ADWin®, or LabView®, contain tools and libraries that facilitate the user in developing and programming[Glaser and Schollmeyer, 2004].