PhD student designs a machine that recognizes potato plants and kills them off. About 95 per cent of potato plants die off.
Ard Nieuwenhuizen, who now works at Plant Research International, designed a machine to do the same thing fully automatically when at the Farm Technology Group. He spent four years working out the details of all the parts required for the automatic potato weeding machine. He received his doctorate for this research on 13 October.
According to Nieuwenhuizen, the most difficult bit was recognizing the young potato plants scattered over a field of young sugar beet plants. Instead of the experienced eye of the farmer, he had two cameras: a colour camera and an infrared camera. 'The problem is that there isn't much difference in colour between sugar beet plants and potato plants. And that colour can vary. The variety being cultivated, the amount of nitrogen in the soil and the level of disease all have an effect, making that subtle difference in colour vary from site to site. This meant we needed an adaptive recognition system.'
Nieuwenhuizen was able to to recognize all the potato plants using the two cameras, but this 100 per cent score also meant 10 per cent of the sugar beet plants were wrongly identified as potato plants. He applied algorithms - rules using statistical probability calculations - to achieve optimum recognition, with 95 per cent of the potatoes and about 2 per cent of the sugar beets being sprayed.
Nieuwenhuizen had to develop a new technique for the spraying. The familiar mist spray method was not specific enough as it would kill off not just the potatoes but the adjacent sugar beets as well. 'Glyphosate kills everything', he says. The herbicide in the new spraying technique is in thick drops that do not drift on to the adjacent plants.
Once the cameras, software and spray had been combined in a machine, it was put to the test by being pulled by a tractor in a field of beet. 'The calculations have to be carried out within a certain time frame if the system is to work properly while moving across the field', says Nieuwenhuizen. The machine worked well at a speed of three kilometres an hour. 'If you go any faster, it will miss potato plants'. The limiting factor turned out to be not Nieuwenhuizen's computer program but the speed at which the valves on the sprayer open and close. He thinks this problem could be solved with a different design.
Nieuwenhuizen has produced a PhD thesis and a prototype for the machine. He expects the machine to come on the market in about five years' time. His research was financed by the technology foundation STW and supervised by five companies. They have the first right to the knowledge he has built up and will proceed further with the development together.