Smart crop row detection in a soybean crop – results experiments in 2024

The objective of WP1 in the final year 2024 of the project is to investigate if it is possible to detect soybean crop rows, for different angles between the driving direction and the crop rows using the WEED-IT sensor from partner Rometron. In the previous report published for this project (Janne Kool, 2023) the distinctiveness of the measurements with this sensor is introduced. The distinctiveness measures how well the sensor can distinguish between soil and crop row when driving skew over crop rows in the field. This distinctiveness is dependent on the width of the field of view of the sensors. To compare the distinctiveness and to assess the possibilities to determine the driving angle from the data, a soybean trial field was planted in Brazil. A quad vehicle equipped with the WEED-IT sensor placed half the height as in earlier seasons was used for the measurements, and these took place at 0- , 15- , 30- and 45- degree angles towards the crop rows. In the data analysis a Fourier approach was used, analogue to the approach in 2023. There is no clear peak in the Fourier transform around the expected frequencies, so it is not possible to find the unique crop row positions. The percentage of soil measured, as approximation of the distinctiveness, is compared with the measurements in 2023 around the same growing stage. This percentage is indeed higher for the measurements done with a mounted lower sensor.
On the dataset collected in the 2022-2023 season in Brazil an algorithm has been developed to predict the locations of the crop rows two meters ahead using a shorter distance. This is based on Fourier analysis and the concatenating of channels behind each other. It is evaluated how well the predictions are based on the driving distance and several down sampling rates.