Automated Sail Trim Analysis

Mr. Mark de Gids M.Sc. studied Medical Computer Science at Utrecht University. His Master’s Thesis was about real-time automated sail trim analysis with pattern recognition techniques.

Medical Computer Science is just like conventional Computer Science, but with a explicit medical application. As you can imagine we combine the power of today's computers with the doctors’ demand for support on interpreting diagnostic images (for example like CT and MRI scans). At this moment a lot of research is focused on automated analysis of these images to support the doctor with his diagnosis.

Being an enthusiastic sailor as well as a gifted researcher on automated image analyses with pattern recognition, at RoboSail systems, he found the perfect project to combine these interests. As the company stimulates research into all areas where advanced computer technology and nautical applications intersect, he was given the opportunity to start this project.

Mr. de Gids states: “Sailing a sailboat is only possible when one is trimming the sails in a way the wind can employ a forward-directed force on the boat. (Sailing close to the wind isn't possible with all the sheets released) As you can see looking at pictures of sailboats sailing close to the wind, you have to trim the sail so that it acts like the wing of a plane. In this wing shaped sail the 3 dimensional curvature is very important and a lot of research is being done by sail makers to improve the so-called sail shape.”

At this moment, sail trim analysis is still a human task. A good sailor looks at his sails and knows which sheet to release or tighten to improve the curvature of the sail and thus improve the boat speed. Real-time automated shape analysis of the sails by computers is still not widely used and is therefore a very interesting research task.

In this project we have combined the knowledge of the sailors on sail trim with the shape and pattern matching algorithms invented by computer scientists. To do this we had to research which parameters are of a sailor's interest when he trims the sails and how he "measures" these parameters just by looking at the sails.

Next, we had to design a pattern to project on the sails (for example with a laser in the mast or simply with black tape on the sails). This shape can be a set of points, but more likely it is a set of lines. You can image that you cannot measure the vertical curvature of the sail when you project only horizontal lines on the sail and vice versa for the horizontal curvature with using only vertical lines. More likely we project a raster on the sail which alternates within special regions of interests. Of course this pattern design (and probably the alternation) depends on the parameters requested by the sailors.

Next we acquired images from the sail with a camera on the deck and in the mast. We then fed these images to the computer program which does the pattern recognition and parameterization of the images.

This pattern recognition is the most interesting part of the project because a lot of pattern recognition algorithms are available but none of them is designed for sail trim. Changing these algorithms to make them suitable for measuring the parameters we like to extract from the images was the most challenging task.

The end result then, finally, is that the sailor is advised to adjust his sails so that the curvature is more optimal and therefore has an optimal chance of winning a race. The system was implemented on an Olympic Finn, and will be converted to larger class vessel in the near future.