Development and Calibration of an Image Assisted Total Station
Published in January 2005
Book
Paperback/Softback
168 pages
978-3-906467-54-2 (ISBN)
Unfortunately, price unknown
Article not available
Description
There exists an increasing demand for higher accuracy, faster processing and ease of use of modern total stations. The purpose of my work is to combine the strength of traditional user controlled surveying with the power of modern data processing to satisfy the needs. The combination of the user's experience and a higher degree of automation retains the efficiency of a manually operated theodolite and enhances the reliability and accuracy of measurements through automation.
The user identifies his targets mainly by their 'structure', which he usually interprets as simple geometrical shapes. Such 'primitive' features, however, can be handled effectively by algorithms to either identify and measure single points or to guide the instrument to areas of interest.
Thus the main goal is to find the 3D coordinates of a non-cooperative but structured target by using a theodolite together with an imaging sensor. The surveyor no longer has to rely on active or cooperative targets like prisms, and this new freedom facilitates his work tremendously. However, the integration of 2D image sensors requires additional calibration effort. My thesis presents a prototype of such an "Image Assisted Total Station" (IATS), models the imaging process and outlines the calibration procedures. Image assisted measurements of artificial markers are compared with traditional measurements. The main effort, however, is focused on applications with natural objects: I try to assess the precision in terms of repeatability, the usability and the comfort of semi-automatic measurements.
A Leica Total Station of the TPS1100 Professional Series is modified into a prototype of an IATS. A 2D CCD sensor is placed in the intermediate focus plane of the objective lens, replacing the eyepiece and the reticle, and an autofocus unit to drive the focus lens is implanted. The image data from the sensor are transferred to a PC using a synchronized frame grabber. To maintain the mechanical stability, the connecting cables transmitting the video signals are guided through the hollow tilting axis. The pixel size of 9.8 µm (Hz) × 6.3 µm (V) corresponds to viewing angles of 2.7 mgon (Hz) × 1.8 mgon (V). To fulfill the specified precision requirements of 0.5 mgon, a resolution of better than 0.2 pixels is required.
Traditional optical total stations measure 'on-axis' objects, i.e. determine both pointing angles of the reticle crosshair. In case of an IATS viewing angles can be assigned to all CCD pixels inside the optical field of view. To describe the relation between sensors pixel coordinates and the angular viewing angles in the object space, a mathematical model is needed, which describes the optics used, and which specifies the contributions of various sources to the overall error budget. In particular, the optical mapping model has to include the theodolites tilting axis errors, the collimation error, the pointing error of the optical axis, and the vertical-index error. Further errors result from a displacement of the projection center from the intersection of the standing and tilting axis and from the optical distortions of field points.
The semi-automated measurement process is based on a permanent interaction between user and instrument. The user supervises the measurement sequence while the IATS executes the measurements. For example, the surveyor proposes a pattern - a geometrical 'primitive' - which adequately represents the object of interest. The processing software estimates the position of the object by local and global template matching. This estimate is used to point the range finder to the selected target to get a valid estimate for the third dimension (depth, distance).
Since the required coordinates of an object point are deduced from the theodolite pointing angles, target distance and the image point location on the CCD, all sensors must be calibrated. It turned out to be useful to perform first a temperature calibration, then determine the exact value of the camera constant with respect to the distance and finally extend the geometrical calibration to all pixels in the optical field of view.
Temperature calibration is similar to the calibration of an optical tacheometer. Using its image processing capabilities, the IATS can automatically drive to measurement positions in both faces, which increases the reliability of the test campaign at different temperatures. The theodolite is positioned, that the object resides at the same sensor position within one pixel for all measurements. This allows us to ignore the influence of deformations caused by optical distortions and mechanical assembly during the calibration, because it is constant for all measurements.
The transformation of the pixel position into viewing angles depends on the camera constant c of the optical system. Its value is a function of the focus lens position, which is monitored by an encoder. During calibration we measure the encoder values at the best focus position for different target distances, using the autofocus option. Then c is determined from the optomechanical construction model.
The geometrical transformation for field pixels outside the optical axis (crosshair) depends on the image deformation and on axis errors of the theodolite. Scanning a stationary object with the theodolite performs the 'off-axis' calibration. For different theodolite positions the CCD images are recorded and a "least squares template matching" algorithm is applied to increase the mapping accuracy. The scanning is done in both theodolite faces with different objects. The transformation parameters are calculated using the horizontal and vertical theodolite angles and the measured pixel locations.
In order to assess IATS capabilities and to check the calibration, a benchmark is used. Limits of operation are tested with the aid of reference markers of circular shape whose positions are known. Furthermore, the capability to measure non-cooperative targets is outlined. Finally, two field tests are performed by measuring a historic building, the Löwenhof in Rheineck/Switzerland, and by measuring the six degrees of freedom of a workpiece at different spatial positions.
The system described in this thesis can be profitably employed wherever today's theodolite measurement systems or close-range photogrammetric systems are deployed: Surveying, vehicle
construction, surveillance, industrial measurement and forensic.
The user identifies his targets mainly by their 'structure', which he usually interprets as simple geometrical shapes. Such 'primitive' features, however, can be handled effectively by algorithms to either identify and measure single points or to guide the instrument to areas of interest.
Thus the main goal is to find the 3D coordinates of a non-cooperative but structured target by using a theodolite together with an imaging sensor. The surveyor no longer has to rely on active or cooperative targets like prisms, and this new freedom facilitates his work tremendously. However, the integration of 2D image sensors requires additional calibration effort. My thesis presents a prototype of such an "Image Assisted Total Station" (IATS), models the imaging process and outlines the calibration procedures. Image assisted measurements of artificial markers are compared with traditional measurements. The main effort, however, is focused on applications with natural objects: I try to assess the precision in terms of repeatability, the usability and the comfort of semi-automatic measurements.
A Leica Total Station of the TPS1100 Professional Series is modified into a prototype of an IATS. A 2D CCD sensor is placed in the intermediate focus plane of the objective lens, replacing the eyepiece and the reticle, and an autofocus unit to drive the focus lens is implanted. The image data from the sensor are transferred to a PC using a synchronized frame grabber. To maintain the mechanical stability, the connecting cables transmitting the video signals are guided through the hollow tilting axis. The pixel size of 9.8 µm (Hz) × 6.3 µm (V) corresponds to viewing angles of 2.7 mgon (Hz) × 1.8 mgon (V). To fulfill the specified precision requirements of 0.5 mgon, a resolution of better than 0.2 pixels is required.
Traditional optical total stations measure 'on-axis' objects, i.e. determine both pointing angles of the reticle crosshair. In case of an IATS viewing angles can be assigned to all CCD pixels inside the optical field of view. To describe the relation between sensors pixel coordinates and the angular viewing angles in the object space, a mathematical model is needed, which describes the optics used, and which specifies the contributions of various sources to the overall error budget. In particular, the optical mapping model has to include the theodolites tilting axis errors, the collimation error, the pointing error of the optical axis, and the vertical-index error. Further errors result from a displacement of the projection center from the intersection of the standing and tilting axis and from the optical distortions of field points.
The semi-automated measurement process is based on a permanent interaction between user and instrument. The user supervises the measurement sequence while the IATS executes the measurements. For example, the surveyor proposes a pattern - a geometrical 'primitive' - which adequately represents the object of interest. The processing software estimates the position of the object by local and global template matching. This estimate is used to point the range finder to the selected target to get a valid estimate for the third dimension (depth, distance).
Since the required coordinates of an object point are deduced from the theodolite pointing angles, target distance and the image point location on the CCD, all sensors must be calibrated. It turned out to be useful to perform first a temperature calibration, then determine the exact value of the camera constant with respect to the distance and finally extend the geometrical calibration to all pixels in the optical field of view.
Temperature calibration is similar to the calibration of an optical tacheometer. Using its image processing capabilities, the IATS can automatically drive to measurement positions in both faces, which increases the reliability of the test campaign at different temperatures. The theodolite is positioned, that the object resides at the same sensor position within one pixel for all measurements. This allows us to ignore the influence of deformations caused by optical distortions and mechanical assembly during the calibration, because it is constant for all measurements.
The transformation of the pixel position into viewing angles depends on the camera constant c of the optical system. Its value is a function of the focus lens position, which is monitored by an encoder. During calibration we measure the encoder values at the best focus position for different target distances, using the autofocus option. Then c is determined from the optomechanical construction model.
The geometrical transformation for field pixels outside the optical axis (crosshair) depends on the image deformation and on axis errors of the theodolite. Scanning a stationary object with the theodolite performs the 'off-axis' calibration. For different theodolite positions the CCD images are recorded and a "least squares template matching" algorithm is applied to increase the mapping accuracy. The scanning is done in both theodolite faces with different objects. The transformation parameters are calculated using the horizontal and vertical theodolite angles and the measured pixel locations.
In order to assess IATS capabilities and to check the calibration, a benchmark is used. Limits of operation are tested with the aid of reference markers of circular shape whose positions are known. Furthermore, the capability to measure non-cooperative targets is outlined. Finally, two field tests are performed by measuring a historic building, the Löwenhof in Rheineck/Switzerland, and by measuring the six degrees of freedom of a workpiece at different spatial positions.
The system described in this thesis can be profitably employed wherever today's theodolite measurement systems or close-range photogrammetric systems are deployed: Surveying, vehicle
construction, surveillance, industrial measurement and forensic.