3D documentation of historical sites and buildings for interdisciplinary works
The continued use of existing structures is of great importance because the built environment is a huge economic and political asset, growing larger every year. The assessment of existing structures is now a major engineering task. The structural engineer is increasingly called upon to devise ways for extending the life of structures whilst observing tight cost constraints. Historical immovables which are also existing structures should be assessed elaboratively in order to preserve cultural heritage and repair historical structures.
The main purpose of this paper is to explain the steps carried out to provide 3D visual input data for designing a semantic 3D model for part(s) of Seddülbahir Fortress from laser scanning data.
The fortress of Seddülbahir is a large site; encompassing a total landscape of nearly 24.000m2 and containing a building mass of approximately 4.200m2. The fortress is scanned using a Leica HDS 3000 scanner, and for the registration of the point clouds Leica Cyclone Register has been used. The entire site is scanned with a point distance of 5mm. Because the distance between the points in the point clouds are small the amount of data gathered is really huge (8GB for the entire site), which makes it hard to work with the data. On the other hand this dense data gives us the opportunity to evaluate the structural, archeological and architectural situations of the entire site.
This paper will focus on laser scanning data of the North Tower and the work carried out for detection of stones from the point clouds. Since the North Tower is built up of mortar and stones, that are seriously damaged because of the weather conditions, vandalism and location of the fortress (very close to the sea and lies close to the geological fault line that passes through Sarköy) , it is really hard to detect the stones and the mortar channel. Resulting data will be used as input for a 3D CityGML model of North Tower for different groups of interest like structural engineers, archeologists and architects.
Three-dimensional surface topography based on digital fringe projection
The reconstruction of three dimensional objects is an increasingly important topic in many fields. To achieve this, variety of techniques have been developed such as coordinate measuring machine (CMM), stereo vision, laser scanning and so on. Nevertheless, these methods have some limitations as follows.
The main drawback of CMM is that it is time consuming and usually requires a human operator.
Furthermore its probe makes contact with the surface which may be undesirable for fragile surface. In addition, CMMs and laser scanning systems can't carry out a whole-field measurement. Although stereo vision carries out a whole field measurement but it has complicated computation and needs two cameras to achieve 3D profilometry of surface.
To overcome this limitation, we use an optical technique, which is fast, non-contact, non-destructive, feasible and provides whole-field information.
This article presents a simple but very efficient technique based on the projection of lines pattern on a test surface made by a video projector and captured directly with a digital camera. In the measurement process, a video projector projects a lines pattern on a test surface, and a digital camera capture the image of lines pattern on it. The height distribution of surface deforms the projected line pattern and modulates them in phase domain. By applying Fourier transform analysis and phase unwrapping algorithms, the 3D profile of test surface can be reconstructed. The purposes of this paper are digital reconstruction of complex object like human face and comparison of execution time among three different phase unwrapping algorithms applied to FTP method. In addition the accuracy of measurement based on this method is compared with rightful measurement.
Large bulk-yard 3D measurement based on videogrammetry and projected contour aiding
Fast and accurate 3D reconstruction of large stack-yard is an important job in material load-and-unload and logistics management. Stack-yard owes special features as: complex and irregular shape, poor surface texture and low material reflectivity, thus its 3D reconstruction is quite difficult to be realized by non-contacting methods, such as LiDAR and photogrammetry. Light-section is good at the reconstruction of small bulk-flow but not very suitable for large-range stack-yard yet. In the paper, an improved method based on stereo cameras and laser-line projector is proposed. Key points of the method include: corresponding point of projected-contour edge matching in stereo imagery based on relationship of image gradient and epipolar-line, 3D point-set calculating for stereo imagery projected-contour edge with least square adjustment and forward intersection, contour reconstructing from 3D point-set of single contour edge based on RANSAC and contour spatial features. In this way, stack-yard surface can be scanned freely by the laser-line projector, and certain region's 3D surface can be reconstructed automatically by stereo cameras on one observing state. Experiment is fulfilled with terrain sand model and calibration frame indoor. Based on 2mega-pixel stereo imagery, 3D contours can be obtained more than 5 per second and average length of reconstructed part is near 70% to the length of projected-contour covered, distance deviation of 3D point-set to a standard plane is 0.6mm, and angle error is limited in 0.9 degree for two perpendicular-crossed planes of calibration frame. That is to say, stack-yard 3D terrain could be reconstructed and measured fast and reliably, and a new useful technology for large stack-yard 3D reconstruction and measurement would be developed with authors' following arrangements and comprehensive analysis.
Infrared digital holography for 3D display
Current and newly-developed 3D displays have the disadvantage that they either force the user to wear special eyewear, limit the number of simultaneous viewers, discard completely certain depth cues (such as blurring) thus causing fatigue, or else encode only a small number of distinct different views of the 3D scene. There is only one known family of techniques that can capture a full 3D scene in a single shot, including phase information, and re-project that light field perfectly thus overcoming all of the above disadvantages: the family of holographic techniques. All other methods are only 3D under a whole lot of conditions. However, unfortunately, holographic techniques for the time being present also a number of drawbacks. Conventional holograms are not dynamic. By replacing the conventional holographic plate with a digital camera and an optoelectronic 2D screen, we can capture and display only holographic images. In order to overcome some other disadvantages, we capture 3D scenes at infra-red wavelengths. IR digital holograms have some plus compared to those ones recorded with a visible radiation source. Firstly, thanks to the long wavelength used, the distance between the camera and the object can be reduced of about a fourth with respect to visible holograms. Moreover, the high output power of IR laser sources and the lesser sensitivity to seismic noise, when compared to the setup exploiting the visible spectrum, make them suitable for the recording of human-size objects. In this paper we present a sequence of holograms which are recorded rotating the object with a fixed angular step. Then they are optically reconstructed using a spatial light modulator using an illumination wavelength of 532 nm. The final reconstruction is obtained from different holograms captured for different observation angles. This kind of reconstruction allows to obtain a 3D imaging of the object. Moreover, using holograms of different objects, we can synthesize numerically 3D scenes to be displayed.
Robust sharp features infer from point clouds
A novel sharp features extraction method is proposed in this paper. First, we calculate the displacement between the point and its local weighted average position and we label the point with salient this value as the candidate sharp feature points and we estimate the normal direction of those candidate sharp feature points by means of local PCA methods. Then we refine the normal estimated by inferring the orientation of the points near the candidate sharp feature region and bilateral filtering in the normal field of point clouds. At last we project the displacement between point and its local weighted average position along the direction of normal .We use value of this projection as the criteria of whether a point can be labeled as sharp feature. The extracted discrete sharp feature points are represented in the form of piece-wised B-Spline lines. Experiment on both real scanner point clouds and synthesized point clouds show that our method of sharp features extraction are simple to be implemented and efficient for both space and time overhead as well as it robust to the noise ,outlier and un even sample witch are inherent in the point clouds.
Longitudinal resolution improving of 3D range imaging lidar through redundant detection and intensity distribution analysis
To improve the poor longitudinal resolution of 3D range imaging Lidar system, the redundant detection is employed. The intensity distribution curves of light reflected by detected objects can be fitted by the intensities in neighboring images, and according to the curves, the objects can be located more precisely. By this method, the longitudinal resolution can be improved from 1.2m to about 0.15m.
3D detecting is one important application of Lidar. Currently, the mature technology is by 3D scanning, and products have already been applied. However, the scanning unit will make the whole system heavier, and the scanning precision will get worse because of mechanical abrasion. Moreover, it has strict demands on laser, such as high frequency, large pulse energy, small divergence angle and narrow pulse width, and this will increase the cost.
To avoid the disadvantages above, the method of 3D range imaging Lidar is put forward, and the experimental prototype is designed. Through control of the gating timing, the system can detect objects in different layers, and in association with the push-broom of airborne platform, it can detect objects in 3D space. It does not need scanning unit, and the demands for Laser are much lower. In addition, every time an image is taken a large area is detected, so it can detect more efficiently than 3D scanning Lidar. 3D rang imaging Lidar is a promising technology to develop.
Longitudinal resolution of 3D range imaging Lidar depends on the laser pulse width and the exposure time of ICCD, both of which cannot be infinitesimal, and so compared with 3D scanning Lidar, its longitudinal resolution is lower. To fix this serious disadvantage and insure the detecting efficiency meanwhile, redundant detection is employed to make sure that one object will appear in two and just two neighboring images in time sequence. The intensities of the object in the two images are different. By using the intensities and distances corresponding to the offered gating time, the distance corresponding to the highest intensity, which is the most precise location of the object, can be determined. The mathematical principle of this method, together with necessary setting conditions, is described in detail. The method is verified by experiments, and results show that the longitudinal resolution can be improved from 1.2m to 0.15m.
A novel color coding method for structured light 3D measurement
This paper presents a novel color coding method which is applied to structured light measurement using fringe-pattern projections. This method is based on time coding method of structured light measurement, in which the encoded fringe pattern is sequentially projected to the measured object surface by a digital projector, meanwhile a camera captures the deformation fringes patterns modulated by the measuring object surface in sequence. Then the three-dimensional shape information of the measured object can be obtained by analyzing the deformation fringe pattern images. And in order to reduce the times of fringe pattern image projection maintaining the same number of fringe patterns or improve the measurement efficiency, a novel coding method of adding colors into the original fringe pattern is proposed. That is, on the basis of time coding, each of the original fringe is subdivided into several color fringes by specific colors, and the sub-fringes can be distinguished automatically from each other by those colors. As a result, the number of measured fringe patterns is increased, or the measurement efficiency is improved under the same projection times. But the introduction of colored fringes also brings a difficulty in image processing, that is color crosstalk. To address this problem, sinusoidal fringe patterns with four colors of white, red, green, blue are used to project to the measured object surface to reduce the edge effect. And with a complementary fringe pattern image corresponding to the original fringe pattern image, image subtraction and other image processing methods, the center of the fringe is extracted more correctly, thus a more accurate fringe center can be obtained. Based on the above measurement idea and methods, a three-dimensional measurement system is built, and the direct linear method of calibration is used to calibrate the internal and external parameters of the system. In addition, lens distortion of the CCD camera and the projector is corrected with the cross ratio invariance method before the calibration of the other system parameters. Then the internal and external parameters of the system are optimized by the bundle adjustment method. Finally, a standard metal hemisphere with spraying was measured, and the accuracy of the measurement system is analyzed by comparing with a standard data obtained by a coordinate measuring machine (CMM) with higher measurement accuracy. Experimental results show that with the same times of projection, the lateral measurement resolution is higher than the traditional method, and the measurement accuracy of the depth direction is about 0.25mm.
Three-dimensional contour reconstruction of push-broom range-gated lidar data: case studies
This article suggests the use of a cross-layer constraint (CLC), originated from the intrinsic relation between time slices (or range slices) of range-gated imaging Lidar, to facilitate the three-dimension contour reconstruction of the target. By using this physical-based method, both the effect of range ambiguity in the raw data and the dependence on the empirical image processing can be reduced, even when the Lidar system is operated under a defocusing condition. Three outdoor scenes are fully studied to demonstrate the validity of CLC method, by using the data released by the DeepView (the push-broom range-gated imaging Lidar system newly developed by Beijing Institute of Technology, BIT).
In the first scene with a simple building corner, we present a close investigation on the intensity of each reflecting surface flows through the time-slice stack to derive the mathematically model of CLC. Under the constraint of the CLC, the range ambiguity of each reflecting surface is resolved in two aspects: determine the most precise longitude location and determine the transverse profile from fuzzy boundaries. The derived CLC model is then applied in the second scene to reconstruct a car with enough depth contrast and semi-transparent surfaces. Many key details mixed by the effect of range ambiguity are successfully decoupled, and thus a more informative 3D contour of car is created with depth resolution about 0.7~0.9m at 60m. Last, the use of CLC in a push-broom process is obtained in the third case, where a large complicated scenario is reconstructed by aligning two small adjacent time-slice stacks. Building corner, trees, oil barrel on the top of building, and several reference targets are well discriminated and are in good agreement with their real locations, where a deviations smaller than 25cm is achieved. This proves that the CLC method is stable against reflectance variation within the scene.
Therefore, the major conclusion is that when the range-gated Imaging Lidar is extended by push-broom mode and the data is reconstructed by using the CLC method, a wide area survey is possible. It should be noted that all three experiments are conducted in a defocusing condition to demonstrate the robust of CLC. This is essential for applications with a large range of depth, or for system with constant focus length.