This title appears in the Scientific Report : 2015 

phenoSeeder - A robot system for single-seed handling and measurement of seed traits
Roussel, Johanna (Corresponding author)
Hombach, Thomas / Fischbach, Andreas / Kochs, Johannes / Herber, Torge / Embgenbroich, Patrick / Brenscheidt, Marcus / Putz, Alexander / Scharr, Hanno / Jahnke, Siegfried
Pflanzenwissenschaften; IBG-2
Plant 2030 Status Seminar 2015, Potsdam (Germany), 2015-03-04 - 2015-03-06
Plant Science
The quality of seeds is essential for plant breeding with very different variability of relevant traits for different crops. In DPPN we have developed a modular phenoSeeder system consisting of an industrial pick-and-place robot and different measurement stations. It enables precise and effi-cient seed handling with low error rates and collecting seed traits such as size or weight of dif-ferent plant species. The robot is equipped with different tools used for seed handling. Depend-ing on seed size, each tool uses a specifically designed needle to which vacuum or overpressure is supplied to suck or release individual seeds, respectively. The picking system has been success-fully tested for seeds of various species such as Arabidopsis thaliana, Brassica napus or Hordeum vulgare varying in size from 180 μm to 8 mm.The seed handling cycle starts at a so called seed source station. It consists of a circular glass pane on which the seeds are manually dispersed and which is uniformly illuminated by an LED ring from below. A color camera underneath the glass pane acquires images of the seeds. The images are analyzed to extract both seed positions and properties such as projected seed area, length and width. For the different seed types individual segmentation parameters have been determined. The position of a single seed is defined by its image coordinates describing the seed’s center of mass. A 2D-homography is used to map image coordinates to world coordinates which can be approached by the robot. After picking up a seed, it can be carried to measure-ment stations like a balance or a 3D-reconstruction station. Each single seed gets its own ID and all measured traits are transferred to a database allowing seed-to-plant tracking in phenotyping pipelines. The modular software concept makes it possible to simply add new measurement stations without having to change the existing software.After traits are measured, the seed is released either into a pot with soil or at a defined position of a well plate. Currently a sowing procedure without moving to a measurement station takes 2.9 to 3.3 seconds depending on the pot’s distance to the seed source station. In case that a seed could not be released, the robot moves to a needle cleaner station where two rotating brushes clean the needle. Another part within DPPN is the construction of a quickSeeder system, based on a similar robot as that of the phenoSeeder. It will enable fast planting of seeds of which the respective traits were scored and which were placed in well plates by the phenoSeeder sys-tem beforehand.