hortiturkey
Editor-in-Chief

Classifying Flower Bulbs Using Vision Technology and Deep Learning

1 Minute(s) Reading
Tuesday, May 4, 2021
Floriculture
flower bulb
Summary
With the use of 3D cameras, errors in the classification of bulbs seem to disappear.
0
Share

After the bulbs are harvested, they are peeled, sorted and checked for defects. Peeling and separating onions can damage the onions, which can affect the quality of the product. In addition, during manual quality controls, some bulbs are overlooked and fall into the wrong class.

Manual separation
Manual separation

To avoid these errors, the University of Wageningen continues to work on an image technology to determine the volume of bulbs and detect defective bulbs.

Before the collected bulbs are packaged, several processes are carried out. In the first step, the onions are first cleaned of the clods and then peeled. It is then finished on a sieve where the large onions are separated from the small ones. While these processes are taking place, there may be some errors, such as small oval-shaped onions entering the wrong class. As a result of these errors, a tulip grower may get uniform non-uniform bulbs and therefore the bulbs may bloom unevenly.

Related Products

The University of Wageningen investigated, with a selection of several hundred bulbs, whether the volume and size of the bulbs could be accurately determined with a 3D camera and compared the results with manual volume measurement by immersion. In this comparison, it was seen that the volume measurements made with the 3D camera were much better than the manual measurement.

3D camera measuring bulbs
3D camera measuring bulbs

With the 3D camera, it is checked whether the bulb is peeled properly, whether it is damaged and whether the skin quality is good, and the bulbs with these defects are sold to the producers as sub-class products.

With these measurements, it has been proven that it is possible to select the volume, size and defective bulbs using a 3D camera. Wageningen University will explore in 2021 whether this measurement can be made on larger volumes.

Sources