FAST AND ACCURATE CLOSE RANGE 3D MODELLING BY LASER SCANNING SYSTEM
暂无分享,去创建一个
Completeness, speed, accuracy are some aspects of the laser scanning system for the acquisition of complex structures and sites. Complete geometry of exposed surface is remotely captured in minutes in the form of dense, accurate “3D point clouds”, ready for immediate use. This technique is used for architecture, virtual reality, heritage preservation and some other engineering and civil applications. Laser scanning technology offers many advantages over traditional surveying and photogrammetric methods: better quality results, improved safety during data capture, no interference with construction and operations activities, no time consuming, simplicity and easiness in learning. Furthermore in many cases, it can provide significant cost saving in both capturing surface geometry and in generating CAD models or otherwise using the gathered data. We applied the laser scanner Callidus Precision System to digitise the shape of the three-dimensional small temple inside the Mole Vanvitelliana in Ancona to build a 3D model. It is a complicated task, made harder by the unusually large size of the data sets. We processed the data by several TIN methods to obtain CAD meshes and realize an efficient 3D rendered virtual object close to the reality. 1 CLOSE-RANGE LASER-SCANNER TECHNOLOGY For more than three decades the measurements of distances by means of laser have been operational in everyday surveying. Only recently the advancements in computer technology enabled the automatic collection and processing of large volumes of laser range data. There are a lot of such systems on the market and they differ from each other (Greco, 2000). The choice depends on the application. The basis of this technology is a scanning laser range finder: the distance from sensor to arbitrary points on the object surface is calculated from the pulse travel time. Some laser scanners measure distances with the flight time and some others use a phase-difference method. By computing the angles we can only obtain the coordinates of the point in the space. According to this method, the continuous laser is based on the contrary on the emission of a bundle. This bundle is divided in two wraps: a reference bundle that immediately hits the system for the measurement of the phase; a measure bundle, that hits the object, comes back and arrives to the phase system measurement. By measuring the phase-difference between two wraps it is possible to go back to the distance of the object. Also, the intensity of the reflected laser pulse is often recorded, providing an indication of the reflection characteristics of the surface. In some systems a fourth dimension, intensity feedback, provides an extra input as to the different materials and colours of an object surface. The 3D laser scanner is also called an active remote sensing system because no additional personnel are needed to hold a range pole or to place targets for measuring surfaces. Combining a pulsed laser with high speed scanning optics we can get detailed and accurate 3D models of industrial object, works of art, buildings or inaccessible structures (Lemmens, 2001). Once tuning parameters have been set, like horizontal and vertical range and angular steps increment, the creation of the initial point clouds of 3D data is done automatically, thousands of points are scanned every second. To efficiently use this powerful and innovative technology, it is necessary to plan the survey correctly. A complex object demands the acquisition of the scan data from different positions in the space and the production of several range maps. It can be too large to be acquired in a single step. It can happen also that all the zones are not visible from the same direction of scanning (for instance in case of 3D objects with great elevation). So it is important that the first phase of the scanning process is the plan of all the scans. Furthermore in case of objects with a complex structure, it is necessary to acquire a high number of range maps, partly overlapped, employed for the meshes generation. It is, moreover, advisable to add to these range maps other range maps, in order to acquire particular or small articulated parts of the object. Take into account that the several range maps will have then to be connected in order to reconstruct the total shape. Such a connection has to be carried out manually. It is advisable, to scan together, at least, three flat surfaces, according to the aligning procedure of several software, used as references in the orientation. 3D laser scanning projects enable us to obtain a fine resolution for better modelling. In fact in order to create complete, accurate models and drawings, it is often important to capture precise edges of structures, piping flanges, etc. In addition, to achieve high accuracy results registering multiple scans together, it is necessary to determine accurate positions of at least three scan “targets” within each scan scene. These situations require a high-resolution scanner that acquires features and targets, with a small spot size laser beam. 2 THREE-DIMENSIONAL SURVEY WITH CALLIDUS The innovative system used in this survey has been the 3-Dimensional Laser Scanner Callidus (Fig. 1) (van Spanje, 2001). Thanks to the movement of the sensor (from 30° to 180° for the vertical movement and 360° for the horizontal spin), it allows to measure with extreme accuracy and speed. It is possible to set its field of view, in both horizontal and vertical directions, of 1°, 0.5° and 0.25°. Inside the scanner, two mirrors rapidly and systematically sweep the narrow, pulsing laser beam over the chosen scene. Callidus uses a flight time method. To capture complete sites or structures, the instrument can be rotated, tilted, and moved around the site. The maximum range is 150 meters. Typical measuring accuracy is ± 5 mm below 32 meters. The system’s accuracy, even at long range, is determined by many factors: high strength and narrow width of each laser pulse; short laser wavelength; high detection sensitivity; ultra high speed timing circuitry, and small laser beam spot size. Via the computer it is possible to define the scansion parameters like step angle irons, minimal and maximum range, the regulation of the camera etc., or to visualize the scans in real-time, and to acquire the measurements. A digital camera is placed sideways to the laser sensor. It allows having some images of the survey, for the survey documentation. Every scansion to 360° for million point’s heartbeats employs little less than 10 minutes. The files of every scansion are approximately of 2.000 Kb (Levoy, 2001). Fig. 1: The 3-Dimensional Laser Scanner Callidus Precision System The surveyed object was a small temple inside the Lazzaretto of Ancona, also called Mole Vanvitelliana, built between 1732-1743. This bastioned pentagonal building, made up of a double line of rooms covered with brick cross vaults, was ordered by Papa Clemente XII to Luigi Vanvitelli. In the centre of the courtyard there is a small neo-classical pentagonal temple dedicated to St. Rocco. This building shows, internally, a couple of columns for each pillar; it is considered a visible part of a system for water collection (Fig. 2). The laser scanning acquired, inside and outside the temple, 8 scans including the buildings facing the courtyard. The number of measured points considered for each scan has been approximately 800.000 at a horizontal and vertical resolution of 0.25° for every scan, at extended horizontal resolution till 2.500.000 points. The final density of the whole model of the surveyed object consisted on many millions of points taken in 45 minutes. The results are graphically displayed like detailed, painted point clouds (Fig.3). In fact for every measured point a direction vector is calculated; it is the result of the average of the normal of the triangle to the bordering measuring points. This direction vector is displayed as point colour in the presentation of point clouds and allows a visual differentiation of the represented surfaces (points of the same direction have the same colour). It can be viewed from any perspective. Every point has an accurate 3D position. The point clouds are an immediately rich data set, already a starting point for many activities. It can be directly used for 3D visualization, point-to-point measurements, or stored for subsequent use. In fact, using the “best fit” software, the clouds can be converted, partially or wholly, into 3D models, 2D drawings, contours, profiles, etc., and used with CAD and rendering applications. Fig. 2 : The neo-classical temple inside the Fig. 3: The point clouds of the laser scanning survey Mole Vanvitelliana in Ancona 3 A SERIES OF TOOLS FOR PROCESSING POINT CLOUDS After the acquisition we have a large size of data sets, which must be processed to model the 3D object surface. First of all, to convert the point cloud into 3D CAD models, we need a powerful PC and an advanced CAD modelling with algorithms of best fitting. A very important thing is how to deal with the large point files. There are few CAD programs that can handle points but most of them cannot properly deal with such information. In fact it is necessary to use appropriate procedures to clean, edit, create meshes and align them. Such laser scanners use a combined software to merge every scan in a whole 3d model. The peculiarity of the software 3D Extractor of the Callidus System is the possibility to join the scans in a unique reference system. The first procedure to carry out is the loading, in succession, of the single scans until the total composition of the object model in a unique system of reference that has origin in the centre of the laser instrument. The program requires the knowledge of three plans, not parallel (possibly orthogonal), and one of them must be horizontal. The accuracy of such aligning procedure has been estimated from a minimum of 2mm/m, joining 2 or 3 scans, to a maximum of 2
[1] Marc Levoy,et al. The digital Michelangelo project: 3D scanning of large statues , 2000, SIGGRAPH.
[2] Paul J. Besl,et al. A Method for Registration of 3-D Shapes , 1992, IEEE Trans. Pattern Anal. Mach. Intell..
[3] P. Pingi,et al. A SUITE OF TOOLS FOR THE MANAGEMENT OF 3 D SCANNED DATA , 2001 .
[4] Paolo Cignoni,et al. Metro: Measuring Error on Simplified Surfaces , 1998, Comput. Graph. Forum.