To use this site, please enable javascript

New cleaning option excels in not just 2D but also 3D sonar and laser data

11 October 2017

Sonar and laser data can include a relatively high level of noise. A wide selection of data cleaning tools is available in the EIVA NaviSuite software products to help remove the noise. These range from various manual as well as automated cleaning/data processing techniques using algorithms such as CUBE/CHRT or S-CAN.

Many of these algorithms have been developed with the purpose of ‘finding’ the seabed, and are consequently especially useful when cleaning 2D data – but less efficient when cleaning 3D data such as laser scans of structures, water column data from sonars, and the like.

As 3D data has entered the maritime industry in recent years and is now used by the majority of our customers, we decided to create a new set of cleaning tools designed to handle not just 2D but also 3D data. That is, tools that can handle noise such as the example shown in the screenshot below: This is a LiDAR (terrestrial) scan that we made from an ATV for a harbour inspection demonstration. It illustrates the issues posed by 3D data quite well as it includes a number of features that should be preserved such as people, a fence and a flag pole together with lot of data noise.

Data set with noise

Data set with noise (for example the points in the red circle)

EC-3D (EIVA Cleaning 3D)

When the EIVA software development team set out to implement a set of new tools, the following characteristics were on the need to have list. The aim was to create tools that:
  • are efficient at cleaning 3D and of course also 2D data
  • include the ability to remove noise while preserving real world artefacts, which other tools might mistakenly consider to be noise
  • are sensor independent, that is, no knowledge of the sensor is required for the tool to work
  • are very intuitive to use, that is, have a simple user interface
  • ensure reproducible results, that is, a set of simple parameters should generate same results when repeated
  • include the possibility of automation, that is, be able to run without user interface based on parameters, for example in the new NaviSuite Workflow Manager
  • have an exceptional high performance

The mission is now accomplished, and we are happy to introduce our new data cleaning option EC-3D (EIVA Cleaning 3D), which collects the new tools, all designed for cleaning sonar as well as subsea and above-water laser data.

Unlike other cleaning/data processing tools, such as S-CAN or CUBE/CHRT, which are applied on DTM cells, EC-3D examines each point in the 3D sonar or laser point cloud to determine if it is noise or real data. EC-3D supports several cleaning methods:
  • Distance to nearest neighbour 
  • Number of neighbours in a sphere 
  • Mean distance to X number of neighbours
  • Grouping of adjacent points based on proximity

The various approaches are useful for different applications and can be applied on top of each other.

So how efficient is the EC-3D cleaning tool?

The first question to raise when introducing a new cleaning tool is of course how well it cleans and preserves 3D / water column features. And the answer is: Very efficiently – as shown in the example below. 
EC-3D efficiently removing noise

EC-3D efficiently removes all the noise in the LiDAR example while preserving elements such as people, flagpole, fence etc. 

Looking at other examples, EC-3D also proves its worth when cleaning sonar data from pipe line inspections – preserving the pipe line in the process. 

Removing artifacts from pipe survey

Artefacts resulting from the sonar ‘hitting’ the ROV are efficiently removed with EC-3D, while preserving the pipeline in the process.

We have been running EC-3D on a number of data sets that we have received from our customers and these have proved difficult to clean with other methods. We are happy with the results, which all add to the conclusion that EC-3D handles challenging cleaning tasks very well. 

What about performance?

Until now, S-CAN was the fastest cleaning implementation available in the market, but EC-3D is many times faster – in fact, typically 10 times faster, and in most cases handles the cleaning ‘near-instantly’, as we call it.

Cleaning a point cloud of 10 million points takes 2.5 seconds, and 3.6 million points (typical sonar block size) only takes 1 second, so there is no risk of EC-3D becoming a bottleneck in point cloud processing. 

Moreover, EC-3D will benefit from good graphics cards, as it uses GPU acceleration to achieve its high performance. We use OpenCL, so if you do not have an OpenCL-enabled graphics card, it will just use the CPU.

What if I am currently using the S-CAN cleaning tool?

S-CAN has been the primary means of automated cleaning in NaviSuite since the cleaning tool was introduced in 2010. There will be no future development on S-CAN, and, consequently, NaviSuite SMS (software maintenance and support) customers will not be charged for S-CAN in 2018 onwards. 

Existing S-CAN licences will continue to work, and future versions of NaviModel will continue to support existing versions of S-CAN.

All SMS customers with S-CAN licences will receive EC-3D for free, and can sign up to EC-3D SMS as of 2018.

The price of EC-3D is approximately 50% lower than S-CAN licences, and volume discounts and subscriptions are available, matching the NaviModel licence to which EC-3D is connected.

EC-3D will be shipped with NaviModel 4.2, which is to be released in the coming months.

Related articles

Show all news