Automatic crop row detection from UAV images

Midtiby, H. (Foredragsholder)

Aktivitet: Foredrag og mundtlige bidragForedrag og præsentationer i privat eller offentlig virksomhed


Images from Unmanned Aerial Vehicles can provide information about
the weed distribution in fields. A direct way is to quantify the amount
of vegetation present in different areas of the field. The limitation of this
approach is that it includes both crops and weeds in the reported num-
bers. To get a better understanding of the weed distribution, some way of
discriminating between crop and weed plants are required. The approach
used is based on the placement of certain crop plants into parallel crop
rows with a relative large distance between adjacent crop rows. Plants
outside the crop rows are considered weeds. We have used a Sugar beet
field as a case for evaluating the proposed crop detection method. The
suggested image processing consists of: 1) locating vegetation regions in
the image by thresholding the excess green image derived from the orig-
inal image, 2) calculate the Hough transform of the segmented image 3)
determine the dominating crop row direction by analysing output from
the Hough transform and 4) use the found crop row direction to locate
crop rows.
Periode24. nov. 2014
BegivenhedstitelAGROMEK and NJF Joint Seminar on Future Arable Farming and Agricultural Engineering: Future arable farming and agricultural engineering.
PlaceringHerning, Danmark