In Trimble® eCognition® software version 10.3 (released December 2022), users gained access to a Python integration that enables advanced ruleset development and opens new opportunities for feature extraction.
Many users have already adopted this new feature. For example, Dr. Dirk Tiede from the University of Salzburg, our Trimble Innovation Partner, has implemented a new segmentation algorithm in Python Hexagon segmentation. A few years ago, Dr. Tiede shared the hexagon segmentation (tessellation)—implemented using eCognition algorithms—with the eCognition user community.
The Python version of the hexagon segmentation is significantly faster than the implementation based on eCognition algorithms. With the seamless Python integration available in eCognition 10.3, users can easily integrate the script into their rulesets.
The output of the script is a thematic layer with geometrically correct hexagons and an object level derived from a vector-based segmentation of the thematic layer.
The key advantage of this segmentation strategy is that, unlike the chessboard segmentation, each image object shares a real border with every neighbor rather than a single point with some neighbors. At any given point, the inside of a hexagon is closer to the center of that hexagon than any given point in an equal-area square would be. This can be particularly useful when analyzing each neighbor of an image object or creating aggregated information layers.
Ruleset (including Python script) is available for download here.
Hexagon segmentation in eCognition
Status of work: public domain
Reference: Dr. Dirk Tiede