Computer vision for broccoli harvest
The work Barth was awarded for lies in the addition of computer vision based selectivity to the harvesting process of broccoli. Broccoli grows at diﬀerent speeds per plant, resulting in high variance of crop size. Currently all broccoli are harvested with a robot in one pass. The result is a suboptimal harvest containing a high share of broccoli that is either too small or too big. The innovation solves this problem by adding industrial cameras to the harvesting robot which produce high resolution images of the broccoli during the harvest. Image analysis then performs a set of crop speciﬁc texture functions in order to separate the broccoli from the rest of the crop and in a next step to determine the position and size of the broccoli. Using this information, it can be calculated if the broccoli has the proper size and is ready to be harvested. If this is the case, a signal with a location of the broccoli is sent to the harvesting unit of the robot. By using the new system, food waste can be reduced whilst yields can be increased.