Detection of Crop Lines and Weeds in Corn Fields Based on Images Obtained from a Drone

Abstract

Precision agriculture (PA) automates agriculture by collecting and analyzing agricultural data for decision-making and obtaining efficient farming productions. Weeds are one of the main factors affecting the yield and quality of farming products. The PA is one of the technological solutions to detect weeds in corn crops through the analysis of digital images to carry out different actions that allow reducing the risk of production in a traditional or automated way. In the present work, an alternative is proposed for the detection of weeds in corn crops in the first 4 weeks of growth using images acquired by a DJI Mavic 2 Pro drone (UAV- Unmanned Aerial Vehicle) with a resolution of 5472 × 3648. The images used are publicly available online. Matlab libraries (Image Processing Toolbox) for the implementation of the algorithm was used in 4 phases: vegetation detection, crop line detection, crop exclusion, weed detection, this allows separating the weed from the crop lines detected in the images captured at 5, 10, and 15 m in height. The results obtained show that the crop lines (85%) and weeds (34.61%) of the total vegetation can be better identified in the fourth week 15 m high. With the proposed algorithm, the processing times evaluated for the finding of weeds, on average, are 3.41 s per image that reaches an area between 20 and 114 m2 at 5 and 15 m in height, respectively.