Digital photogrammetry for building measurements and reverse-engineering/Skaitmenine fotogrametrija matuojant bei atkuriant pastatus.

Suziedelyte-Visockiene, Jurate ; Brucas, Domantas

1. Introduction

For many purposes geometric information about existing buildings or the ones under construction in 2D plans or three dimensional CAD models is necessary for control, conservation or reconstruction. If the data are still available from the construction process, they may have insufficient actuality. In a modern environment the 3D CAD data are preferred, showing the actual state of the building. These data may be acquired by a manual measurement using geodetic survey instruments, by theodolites or tachometers, or by photogrammetry using digital images (Luhman, Tecklenberg 2001; Suziedelyte Visockiene 2007).

The project of the building under construction dimensions control is described further in this paper. Since for a higher precision the measurements were performed using 2 methods (these are measurement by photogrammetric principles and the measurements by the laser distance-meter) it was possible to compare these two principles determine the accuracy achieved.

2. Determination of the construction site dimensions

The primary objective of the investigation was to determine the actual dimensions of windows of the building under construction (Fig. 1). For that purpose it was decided to implement the photogrammetric methods, namely use the PhotoMod photogrammetric software (Boroumand, Doost 2006; Serebryakov, Nepeina 2006).

For receiving the unambiguous tie points of images for further processing and obtaining the scale factor a large number of marks were placed all over the building (marked as white circles in Fig. 1). The precise 3D position of each mark was determined using the digital tacheometer and those data were transferred to the photogrammetric Photo-Mod project. Dimensions and the position of the windows were determined inside the mentioned project.

[FIGURE 1 OMITTED]

The processing of images in the photogrammetric project with the dimensions evaluation is described in the further chapters of this paper.

3. Building reconstruction using photogrammetric method

In photogrammetry we usually measure the 2D image coordinates of points in 2 or more images and calculate the dinates of each point in the images are usually determined by a special high precision (and expensive) photogrammetric instrument, like PlaniComp P2 (Leica) in the classical photogrammetry. In digital photogrammetry this is done on screen of PC using the pixel coordinate system defined by rows and columns. If analogue image sources are to be used, they must be scanned by an image scanner. The image acquisition and processing are shown in Fig. 2 (Boroumand, Doost 2006).

[FIGURE 2 OMITTED]

The first part of the photogrammetric process is the acquisition of the images with analogue photographic (UMK camera) or digital cameras. The analogue imaging devices can be metric, semi-metric or amateur cameras. During the last years some digital cameras have been developed using a CCD (Charge-Couple-Device) sensors (arrays). But whereas the low budget cameras have an insuf- ficient resolution, cameras with larger CCD arrays are still expensive. But a development towards cheaper and better digital cameras can be anticipated in the near future. This will allow a fast digital data flow, avoiding the time consuming wet photographic processes.

Each point, which is required for the complete restitution of the object, has to be displayed on at least 2 images from different points of view. If it is desired to view the object in a stereoscopic manner, the images have to be taken according to the so called normal case of photogrammetry with nearly parallel directions of view from two points on a horizontal base, perpendicular to the viewing directions. Such an arrangement is similar to the arrangement of the human eyes.

If the pictures are acquired especially for photogrammetric purposes, the imaging team will survey a few control points by geodetic techniques (Skeivalas 2008). Control points are points with 3D coordinates in a geodetic coordinate system, which can be identified and measured in at least one image. They are necessary for the orientation to follow, which is the second step of the photogrammetric process.

The orientation procedure consists of the reconstruction of the interior orientation, which describes the geometry of the ray beam in the camera and the exterior orientation.

The interior orientation is needed for the calculation of parameters determining the position of coordinate system of image and orientation regarding the coordinate system of the digital view.

If the camera with the known camera calibration parameters is used, the transformation of image to the geodesy coordinate system is calculated according to the equation (Software PhotoMod 4.4 AT):

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (1)

where x, y--coordinates of a point in the geodesy coordinate system; [x.sub.c], [y.sub.c]--coordinates of the points in digital image coordinate system; [x.'sub.c][y'.sub.c]--null-point of the digital image coordinate system in the geodesy coordinate system; [phi]--the rotation angle of the image coordinate system in the geodesy coordinate system; [k.sub.c], [k.sub.c]--the coefficients describing an image deformations along the x, y axes.

The interior orientation can be performed in automated or manual mode using the special PC softwares (Software PhotoMod 4.4 AT).

If the used camera is calibrated, interior image orientation may be done by transforming the measured coordinates into a calibration system, defined by fiducial marks or residual crosses.

If a non-calibrated camera has been used, an independent set of parameters of the interior orientation is necessary for each image. In frame cameras independent parameters of the interior orientation are only required if the zoom factor or the focus of the camera has been changed during the image acquisition.

The exterior orientation describes the position and the viewing direction of the camera in a superior object coordinate system and is today usually calculated by a bundle adjustment. This is the simultaneous calculation of the data of the exterior orientation, the data of interior orientation (if required) and the 3D coordinates of the points, of which 2 or more pairs of image coordinates are available. For this purpose at least three control points and 5-10 tie points per image are required. Tie points have to be identified and measured in at least 2 images. For each image at least 6 unknowns (3 coordinates for the position, 3 rotations and if required parameters of the interior orientation) and for each object point 3 coordinates have to be estimated. The unknowns can be calculated by a least squares adjustment if more observations than unknowns are available. The more control and tie points are available, the better results of the orientation process in terms of accuracy and reliability can be obtained.

4. Practical measurements

As was mentioned before, the main task of the work was to measure the dimensions of the windows of the building being under construction (Fig. 1). The marks placed on the outer walls of the building were used for both exterior and interior orientation of the images taken by the Canon EOS 350D digital camera, and the exact coordinates of the points (marks) were determined using Leica total station (ISO 5725-1:1994).

The original images were digitally corrected by the TCC software (developed by the Photogrammetric Institute of University of Bonn), according to the camera calibration results (Suziedelyte-Visockiene 2007). To obtain the needed measures, the images were processed using the PhotoMod photogrammetric software. Since only the actual dimensions of the windows were required, only the measurements of the areas of interest were performed. The results of windows dimensions measurements (of a certain small area of the building) are shown in Table and graphically (with their actual positions) in Fig. 3. As can be seen from the figure, it was possible to measure only limited number of the windows--those which were clearly visible, the windows at the bottom floors were covered by the scaffolds and it was impossible to perform measuring due to the lack of the visual contact with the areas to be measured.

[FIGURE 3 OMITTED]

Since the digital images of the object (building) used for measuring were taken at quite large distances and those were impossible (or extremely difficult) to decrease, the accuracy of measurements based on the measurements of pixels position in images was also quite limited. Due to that, to obtain some additional data on the measures and the photogrammetric measurements themselves it was decided to perform additional measurements of windows dimensions using the instrument of a higher and well-known accuracy. The instrument used was the laser distance-meter Leica Disto TM Plus, having the standard deviation of measurements [s.sub.d] = 1,5 mm. The windows of the entire building were measured using the mentioned instrument at several positions (Fig. 4). The results of windows dimensions measurements performed by the laser distance-meter are shown in Table 1.

Considering the measures obtained by Distomat as being the reference ones (since the accuracy was known and it did not change through the entire measurement process), it was possible to compare and calculate the deviations of measures obtained by means of photogrammetry. The deviation of the photogrammetric measurements:

d[[DELTA].sub.ph] = [[DELTA].sub.ph] - [[DELTA].sub.d], (2)

where [[DELTA].sub.ph]--measures obtained by means of photogrammetry; [[DELTA].sub.d]--reference measures obtained by laser distance-meter.

[FIGURE 4 OMITTED]

The calculated deviations for each floor are shown in Table 1 and in Fig. 5.

[FIGURE 5 OMITTED]

Having the deviations of photogrammetric measurements calculated the standard deviations of mentioned measurements for each floor of the building could be found (ISO 17123-1:2002):

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (3)

where n--number of measurements; d[[DELTA].sub.phn]--deviation of each measure.

Standard deviations for each floor are listed in Table 1 and graphically shown in Fig. 6.

From Fig. 6 it is clearly visible that the standard deviation of measurements depends on the floor of the building, i.e. it depends on the distance from the camera to the measured object. The higher the floor of the building (larger distance), the bigger standard deviation of photogrammetric measurements appears. The dependency of the standard deviation of photogrammetric measurements upon the distance to the measured object is graphically shown in Fig. 7.

As it can be seen from Fig. 7 the dependency of the standard deviation of measurements on the distance is almost linear despite some obvious faults falling out of the more or less linear shape.

[FIGURE 6 OMITTED]

[FIGURE 7 OMITTED]

Overall standard deviation of measurements including the standard deviation of reference measures (laser distance-meter) can be calculated (ISO 5725-1:1994):

[S.sub.phf] [square root of ([S.sup.2.sub.ph] + [s.sup.2.sub.d]). (4)

The calculated full standard deviation of measurements is [S.sub.phf] = 51.03 mm.

5. Conclusions

1. Photogrammetric methods can be implemented in the reverse-engineering of buildings (for heritage reconstruction or dimension control) quite successfully allowing obtaining the needed dimensions from the images long after the conditions of object has changed.

2. Photogrammetric method of dimension evaluation produces quite large errors of measurement heavily depending on the distance to the measured object and the resolution of camera.

3. Depending on the distance to the measured object, standard deviations of measurement in project described varied from 39 mm (at 30 m) to 64 mm (at 39 m).

4. General standard deviation of 9-floor building measurements was estimated as 51 mm, which is not quite acceptable in many cases.

5. Despite the accuracy of not high enough (at large distances), the photogrammetric object measurements allow estimating not only the dimensions of the object itself but also its position (coordinates), which is sometimes quite difficult to accomplish by other means of measurement.

doi: 10.3846/1392-1541.2009.35.61-65

Received 01 06 2008; accepted 21 03 2009

References

Boroumand, M.; Doost, N. 2006. Report on Processing Digital Images Taken by UltraCam In PHOTOMOD Version 4.0, in VI International PHOTOMOD User Conference, Becici, Budva, Montenegro. 46 p.

Skeivalas, J. 2008. GPS tinklu teorija ir praktika [Theory and practice of GPS networks]. Vilnius: Technika. 288 p.

ISO 17123-1:2002 Optics and optical instruments--Field procedures for testing geodetic and surveying instruments--Part 1: Theory. Geneva: ISO. 10 p.

ISO 5725-1:1994 (Trueness and Precision) of Measurement Methods and Results--Part 1: General Principles and De- finitions. Geneva: ISO. 22 p.

Leica DISTO. Available from Internet: <http://www.leica-geosystems.com>.

Luhman, T.; Tecklenberg, W. 2001. Hibride photogrammetric and geodetic surveillance of historical building for urban tunnel construction. Available from Internet: <http://www.fh-oow.de/institute/iapg/paper/Tunnel%20Hemelingen.pdf>.

Serebryakov, S. V.; Nepeina, N. N. 2006. Practical experience of digital topographic mapping and orthophotos creation using PHOTOMOD software system, in VI International PHOTOMOD User Conference, Becici, Budva, Montenegro, September 19-22, 15.

Suziedelyte-Visockiene, J. 2007. Skaitmenines matuojamosios fotokameros kalibravimo parametru itaka nuotraukas transformuojant i plokstuma [Definisions influence for Digital Photocamera parameters of calibration on transformation of photos in the plane], Geodezija ir kartografija [Geodesy and Cartography] 33(1): 26 -30.

Software PhotoMod 4.4 module Aerial Triangulation user manual. Racurs, Moscow. Available from Internet: <http://www2.racurs.ru/docs/en/at.pdf>.

Jurate SUZIEDELYTE-VISOCKIENE. Assoc. Prof. Doctor. Vilnius Gediminas Technical University, Dept of Geodesy and Cadastre, Ph +370 5 2744703, fax +370 5 2744705.

Doctor of technical sciences of Vilnius Gediminas Technical University, 2003.

Research interests: digital photogrammetry, land management.

Domantas BRUCAS, Doctor. Vilnius Gediminas Technical University, Dept of Geodesy and Cadastre, Ph +370 5 2744 703, fax +370 5 2744705.

Doctor of technical sciences of Vilnius Gediminas Technical University, 2008.

Research interests: development of comparator for angular measurements, automation of the processing of measurement results.

Jurate Suziedelyte-Visockiene (1), Domantas Brucas (2) Department of Geodesy and Cartography, Vilnius Gediminas Technical University, Sauletekio al. 11, LT-10223 Vilnius, Lithuania E-mail: (1) j_visockiene@hotmail.com; (2) domka@ktv.lt

Table 1. Measures of the windows taken by different methods with
the deviations

Floor 1 Floor 2

 Photogram- Photogram-
Distomat metry Deviation Distomat metry Deviation
(mm) (mm) (mm) (mm) (mm) (mm)

3367 3371 4 4116 4084 -32
3385 3304 -81 4137 4082 -55
2508 2551 43 2575 2585 10
2510 2613 103 2567 2574 7
2686 2702 16 2985 3078 93
2679 2707 28 2986 3110 124
1789 1784 -5 1817 1817 0
1789 1797 8 1815 1796 -19
411 403 -8 1166 1121 -45
425 387 -38 1168 1115 -53
1579 1576 -3 1846 1899 53
1566 1628 62 1844 1875 31
1782 1788 6 1819 1814 -5
1773 1786 13 1652 1702 50
3781 3794 13 1664 1697 33
3776 3790 14 1784 1785 1
1790 1794 4 1786 1826 40
1806 1797 -9 824 810 -14
 811 805 -6
 3367 3362 -5
 3363 3399 36
 1798 1800 2
 1788 1788 0

St. deviation 39.25 St. deviation 44.70

Average distance from camera to object (m)

30.064 31.037

Floor 3 Floor 4

 Photogram- Photogram-
Distomat metry Deviation Distomat metry Deviationon
(mm) (mm) (mm) (mm) (mm) (mm)

4095 4064 -31 3334 3247 -87
4111 4043 -68 3289 3296 7
2580 2590 10 2506 2516 10
2597 2580 -17 2513 2533 20
2257 2284 27 2965 2957 -8
2261 2288 27 2954 2951 -3
1777 1693 -84 1782 1761 -21
1772 1723 -49 1790 1816 26
449 401 -48 697 745 48
440 406 -34 704 740 36
2443 2487 44 1738 1695 -43
2444 2465 21 1736 1700 -36
1763 1692 -71 1791 1799 8
1757 1703 -54 1794 1794 0
1100 1146 46 1472 1543 71
1085 1089 4 1479 1432 -47
1764 1776 12 1767 1761 -6
1764 1820 56 1766 1779 13
578 569 -9 957 923 -34
611 609 -2 956 968 12
4199 4151 -48 2667 2721 54
4133 4211 78 2656 2717 61
1768 1790 22 1787 1784 -3
1762 1811 49 1771 1791 20

St. deviation 45.42 St. deviation 37.25

Average distance from camera to object (m)

32.289 33.725

Floor 5 Floor 6

 Photogram- Photogram-
Distomat metry Deviation Distomat metry Deviation
(mm) (mm) (mm) (mm) (mm) (mm)

3994 4050 56 2410 2480 70
4026 4035 9 2416 2424 8
2454 2566 112 2530 2498 -32
2457 2607 150 2533 2559 26
2167 2274 107 6056 5957 -99
2155 2270 115 3139 3105 -34
1758 1778 20 976 921 -55
1765 1754 -11 1934 1882 -52
615 530 -85 1763 1830 67
631 489 -142 1794 1818 24
2529 2438 -91 6786 6868 82
2499 2480 -19 6795 6882 87
1774 1800 26 5753 5673 -80
1772 1822 50 2564 2580 16
1883 1881 -2 1955 2001 46
1883 1865 -18
1781 1823 42
1772 1798 26
555 556 1
555 575 20
2675 2750 75
2685 2737 52
1767 1796 29
1791 1772 -19

St. deviation 70.82 St. deviation 60.64

Average distance from camera to object (m)

35.367 37.214

Floor 7

 Photogram-
Distomat metry Deviation
(mm) (mm) (mm)

6843 6756 -87
6843 6767 -76
2893 2826 -67
2898 2846 -52
2050 2066 16
2050 2122 72
1763 1750 -13
1779 1761 -18
6749 6869 120
6792 6833 41
2324 2292 -32
2349 2390 41

St. deviation 64.14

Average distance from camera to object (m)

38.876