I³S is an acronym of Interactive Individual Identification System. I³S Contour is an extension of the I³S system using the contour of e.g. the tail, dorsal fins or lining along the side. This name explains most of I³S Contour’s functionality. First, we will focus on the interactive part. I³S Contour requires user interaction and is meant to support and not to replace the researcher. Initially, you have to point out the relevant part of the contour of the unknown individual animal. I³S Contour has a tracing algorithm which tries to find the contour between the points selected by the user. In the next step, I³S Contour assists you in the tedious task of matching animals for identification purposes. It automatically matches an annotated image with all annotated images in the identification database and shows a ranked list of images. However, the user will always be responsible for making the final match between the unknown image and an image from the identification database. Please note that we distinguish between your database with all images taken and the identification database. The identification database contains only the best images of individuals and preferably 3 to 5 of each individual.

The Contour algorithm

To help recognize individual characteristics like the flukes of Whales, I³S first traces and extracts the fluke contours of the unknown tail. The image should be taken ideally perpendicular to the line of sight and no more than 30 degrees off that line. For contour tracing the goal is to find the optimal path between a start node and a set of goal nodes. This means finding the globally optimal path from a start pixel to a goal pixel, in particular, pixels represent nodes and edges are created between each pixel and its 8 neighbors. Optimality is defined as the minimum cumulative cost path from a start pixel to a goal pixel where the cumulative cost of a path is the sum of the local edge (or link) costs on the path.
The next step is to match the contour with all identified contours in the identification database. The contour matching software is based on the ideas from the EUROPHLUKES project developed by the Center for Mathematics and Computer Science in Amsterdam (CWI).
To be able to compare two contours it is essential to transform them into a single common space. This is done in a number of consecutive steps:

1. Translation and rotation. The entire contour is translated such that the start of the contour has coordinates (0,0). Simultaneously, the contour is rotated to end at the x-axis.
2. Scaling. The contour is scaled to have the last point on the contour end at (500,0).
3. Histogram. To allow for fast comparison the area below each line piece in the contour is discretized into 500 bins.