Point Clouds and Hull Surfaces

Background

Back in 2013 I had the opportunity to process some Point Cloud data into a hull surface. I had a look round to find tools that would support this approach but was surprised to find that none really provided the capability to deal with survey's taken in the marine environment and generate surfaces suitable for marine engineering and Naval Architecture. I developed my own solution and this is the result, including an ICCAS Paper [1] and presentation. I'd be interested to see what you think of my approach. Check out the videos at the end of the page to get a feel for the experience and if you choose to try out PolyCAD there is a detailed explanation of the setup here.

Emerging Technology

Optical surveying techniques such as laser scanning are becoming more accessible to projects as the cost of equipment reduces and expertise is developed. These techniques use electromagnetic waves projected from a base station to measure the distances of surfaces away from the sensor, much like a radar. Development of the technology means that a greater variety of different sensors are becoming available from typical survey type systems, hand held devices to vehicle mounted units. The development of mobile technology means that many more low-cost systems based on stereoscopic images captured with two cameras are emerging with most of the effort of reconstructing the captured view being performed by software. All of these techniques create a survey dataset consisting of points often numbering in the millions.

Laser scan point cloud and 3D modelling of the boiler room of HMS Belfast in AVEVA Everything3D.

These survey techniques capture far more geometric data more safely than previous methods. An engineer can now bring qualified real world data to their desktop and use it to develop solutions in their present project. For the Naval Architect, this approach offers the opportunity to scan ships externally and internally capturing a depth of information never before feasible. The challenge of not having a hull definition because the Lines Plan is missing or not trusted is now resolvable (assuming that there is access survey the vessel out of the water).

Geometry from Point Clouds

Processing a dense Point Cloud to create a facet surface model for visualisation or a mathematical surface for modelling and analysis is straight forward if the dataset was originally surveyed in a clean and clear working environment and the intention is to completely convert dataset into geometry.  Algorithms used to make the conversion have been in the public domain for years and can be easily implemented once understood.  However, in real engineering scenarios its necessary to account for the context in which the survey is taken and how geometry will be extracted from the scene given that a great deal of extraneous information may be captured, information that may also obscure the artefact of interest. In this scenario the Point Cloud cannot be simply processed by an algorithm to generate a surface representation.  The relevant parts of the dataset need to be identified and prepared for fitting and consideration needs to be given to geometry that could not be captured if it was obscured. 

For the Maritime context there is significant challenge from the outset. Floating vessels are intended to spend their time at sea where the geometry we're interested in is hidden from view in a moving environment. The only opportunity to measure, presently at least, is when these vessels come to shore. When a vessel is removed from the water they have to be supported in dry dock or on hard standing in an unnatural way which obscures certain areas from view and therefore the survey. Removing a ship from its operational state is expensive although necessary for annual survey and refits and means that surveying must adapt to other activities that may be going on at the same time. 

Drydocks and Hard standing can be challenging areas to scan a hull.

The Dry dock is a challenging environment to survey in. The ship must be correctly supported introducing structures and obscuring the keel and any parts of the vessel close to the dry dock floor. Work is taking place, people may be moving and rotating machinery may cause vibrations. The scanner may capture unusual information such a birds flying through and reflections of the ship in any puddles. Additional scanned artefacts can be removed from the survey during post-processing but any missing information must be accepted.

A typical scan may contain artifacts that shouldn't be included in a remodelled surface and it may have other area that are poorly captured.

Once the survey has been captured and post-processed it will be available as a large database of points. This may be converted into a facet based surface which is good for display but poor as an engineering model. The size of the data will be large and it will be challenging to modify. The Naval Architect will prefer to represent the hull form as a surface. This format is strongly supported by many software tools because the data size is small and modification is possible. The challenge is that while many algorithms exist to fit surfaces to points sets they generally assume that the data is perfect. Furthermore, the shape of hull forms is often challenging to represent using a single surface. It's necessary to distort or trim it to fit. This reduces the Naval Architect's ability to manipulate the surface once created due to the additional definition and detail added by this process. 

A draped surface fitted to a Point Cloud needs to be trimmed to shape. This approach may not pick
up specific features of the hull, such as the Rabbit-Line in this case., Turn Tall Ship 3D Scan Data into a CAD Model

When scanning a hull out of the water some areas will poorly captured particularly those areas where supports touch the hull such as along the keel or the dock blocks. There may also be artefacts in the scan that should not be represented in the surface such as any appendages (Rudders, Propellers, Stabilizers, etc.). It could be said that that Naval Architects work with idealised representations of the hull form and this is somewhat true, but we gain the realism by adding features as separate components rather than representing the complete geometry in a single representation. 

Taking all these requirements into account it is challenging to model or recapture hull surfaces from Point Cloud survey data in generic CAD tools efficiently because there are many special surface features that need to be captured in the right way.  Surface design tools specifically for hull forms provide the capability to represent these features easily with special definiton and by documenting the modelling techniques.  This highlights that there are benefits to be had through the integration of Cloud Point data into these marine specific hull design software.

An Alternative Approach

Regenerating a hull surface from a Lines Plan can be a tedious experience. In fact, the time required to generate the surface definition itself is fairly small compared with the physical aspect of measuring the coordinates and typing in the information. PolyCAD addresses this particular problem by allowing scanned images of plans to be imported into the software and traced in the graphical environment. The simple geometry used to trace the lines of the image can be used to generate the specific X-Topology Curves required to represent a complex surface using the Intersection and Curve Fitting tools . The software assists the user assemble the design by providing functions that quickly generate good quality geometry, and allow thorough review it before accepting it. Point Clouds are just another source of geometry and are exposed in the intersection tools for curve fitting as if it were another source geometric data source.

Capturing shape from a Point Cloud into a surface definition may be achieved selectively
by choosing an intersection plane and reviewing the fitting parameters to achieve the best definition.

Rather than providing specialist tools or processes that generate a hull surface from a Point Cloud, PolyCAD just extends the range of entities that can be intersected to include Point Cloud data. Hull Surfaces can therefore be built up using standard methods and incorporate any information the user has available. This extends to handling areas of the Point Cloud where data is missing or poor and the exclusion of appendages that should not be captured in the surface. Since the user select the location in the Cloud where geometry should be fitted poor or excluded areas of definition can be easily avoided. 

Basic Geometry Extraction

Taking a marine perspective when extracting information from a Point Cloud highlights some interesting opportunities that may not be apparent in generic CAD/Cloud software tools. Take for example the ability to cut section through the point cloud. When cutting a surveyed hull, the points associated with cut visually form a curve. Assembling a number of cuts can reveal the 'Lines' of the hull geometry and produce a viable visualisation tool for surface regeneration both for understanding the shape of the surface but to compare the corresponding contours generated from the surface with the Cloud.

Simply sectioning the cloud and capture scanned points near the sectioning
plane reveals the contours of the hull without any geometry being fitted.

Curves implied by the cuts through the Point Cloud may be of course be turned into geometry . Entering the Edit Mode (F2) of each Point Cloud Section allows the points of the intersection to be selected and a polyline automatically fitted to the data. This allows simple offsets or hull sections to be quickly extracted from the Point Cloud when a full hull surface definition isn't required.

Contours defined by the sections through the Point Cloud can be simply converted
to polylines if the quality of the dataset is good.

Rebuilding Surfaces using X-Topology

The process of defining a hull surface using X-Topology fitting to the Point Cloud remains the same as if you are designing the surface without reference data. First define the boundaries, then feature curves and capture the curved shape before finishing with crossing curves. Since the boundaries and features curves usually have shapes which are discontinuous it is challenging to use the fitting tools without human direction. In fact, unless these shapes can be captured directly using the Curve fitting tools it often more effective to digitise these initial shapes manually.  The Point Cloud Sections cuts described above implement snapping so it's just a case of picking up the positions highlighted on screen.

Once the basic Form Topology definition is in place the shaped areas of the surface can be defined. Start with the Primary Shape Curves, the initial set of curves which drive the shape of surface across the region. Following the recommended process for design it’s best to choose a direction across each region of shape which has the least or minimal variation in curvature. For a hull form this normally means a longitudinal direction, either waterlines or buttocks depending on the style of the portion of the hull form. Non ship hull forms will often benefit from capturing shape using diagonals. If possible, the curves across each region of shape should be a family. This means having the same number of points to maintain mathematical consistency. The Curve Fitting tool allows these parameters to be reviewed after each intersection has been processed. Once the first set of Primary Shape Curves they can be followed up with the second set of crossing curves. These are usually quickly created using the Spline Fit tool. This part of the process is shown in the video below.

The Surface Generation Experience

The Surface Definition Process from Begining to End

This video shows the complete process to capture the geometry of a hull form in a Point Cloud into an X-Topology Surface

Generating a surface from real-world Point Cloud data and all the Challenges that brings

This video uses X-Topology to generate a hull surface from a Laser Scan. The Point Cloud exhibits the usual challenges of surveying in the real environment such as obscured areas, lack of detail on the keel and dealing with a shiny hull surface. PolyCAD is able to deal with these issues by selecting the good parts of the data and using a tolerant fitting approach.

Further Information

This short article only touches the surface of the Point Cloud capabilities in PolyCAD. More information can be found here

References

  1. Regenerating Hull Surface Definition from Laser Point Clouds, M. Bole, ICCAS 2015, Bremen, Germany, 29th September - 1st October 2015.