Manually harvesting broccoli for the fresh produce market is highly labour-intensive and responsible for some 35% of the total production costs of this crop. Alongside considerable cost savings, robotic harvesting would also decrease the load on horticultural workers in the peak season. Researchers at Wageningen Plant Research have developed image recognition software which has brought robotic broccoli harvesting a step closer.
A fully automatic harvesting robot would need to be able to identify broccoli in the field and determine its size and ripeness.
Researchers at Wageningen Plant Research have developed image recognition software that can distinguish a broccoli plant from the background based on texture and colour. This is an important first step in the development of a fully automated harvesting robot.
The automatic recognition software has been tested in various field trials and evaluated using images of broccoli of various sizes and of both marketable and non-marketable ripeness. With an accuracy of 94%, the technology looks very promising.
The results were presented at AGRICONTROL 2016, the 5th IFAC conference on Sensing, Control and Automation Technologies for Agriculture, in Seattle in the US.