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

4.1 Data Processing 774.1.5 Noise ReductionNoise in laser scanning point clouds is mostly caused by the distance measurement system. The range, thecolour of the surface and the angle of incidence influence the intensity of the reflected laser beam. The lessthe intensity, the more biased the data, cf. Chapter 3. Considering the noise, which is aligned along thedistance measurement direction, processing of the noisy data can be carried out by filtering the data alongthe measuring direction.It was found that the approximation of a pyramidal frustum, with its principal axis representingthe mea¬suring direction as defined by the center of the laser scanner towards the object points, isa promisingfiltering model. The length of the principal axis, which defines the height of the pyramidal frustum, con¬siders the noise of the data. The noisier the data, the longer the principal axis. The bottom base and the topbase define the filtering interval according to the scanningresolution of the laser scanner in the horizontaland vertical directions. The larger the base of the pyramidal frustum, the larger the resulting point spacingof the noise-reduced point cloud. Naturally, the base dimensions should not be smaller than the scanningresolution of the acquired point cloud.As an example, a biased data set can be seen in Figure 4.4. The noise of the data is mostly caused by asmall angle of incidence and a dark surface colour. The data shown describe the cross section of a one laneroad section. The noise is found mostly at the dark skidmark. The indicated pyramidalfrustum is orientedalong the measuring direction. The height of the pyramidal frustum is chosen to define the amplitude orthickness of the noise. The rays defining the measuringdirection intersect in the center of the laser scanner.pyramidal frustum_^road surfacemeasuring directionFigure 4.4: Noise in laser scanning data oriented along the measuring direction according to the filtering approach.The noise reduction is made by averaging all points inside the pyramidal frustum. The averaginglow pass filtering suppressing high-frequency noise from the signal. The signalacts as ais assumed to be the meanvalue, i.e. Gaussian noise. If the points are described by spherical coordinates1, which refer to the localscanner origin (center of laser scanner), the mean value can be calculated by:n1h=-y~]hl, ~y~]vl, -'S~]sl, v = s =int=i t=i t=iin(4.5)where n is the number of points inside the pyramidal frustum. The selection of pointsthe pyramidalfrustum is carriedto be located insideout based on the differences of the spherical coordinates of each pointof the point cloud. One point of the point cloud is picked and the spherical coordinates are compared.The base of the pyramidal frustum is defined by a maximum acceptable angle difference, which has to besmaller than the scanning resolution of the point cloud. The maximum acceptable angle difference is equalin the horizontal and vertical directions since the scanningprecision and the accuracy of the angle measurement system are also equalresolution of the laser scanner as well as thein both directions. If a different1Spherical coordinates are defined by two angle values h and v in the horizontal and vertical directions as well as by a distancevalue s.