Abstract
Crop mixtures are often beneficial in crop rotations to enhance resource utilization and yield stability. While targeted management, dependent on the local species composition, has the potential to increase the crop value, it comes at a higher expense in terms of field surveys. As fine-grained species distribution mapping of within-field variation is typically unfeasible, the potential of targeted management remains an open research area. In this work, we propose a new method for determining the biomass species composition from high resolution color images using a DeepLabv3+ based convolutional neural network. Data collection has been performed at four separate experimental plot trial sites over three growing seasons. The method is thoroughly evaluated by predicting the biomass composition of different grass clover mixtures using only an image of the canopy. With a relative biomass clover content prediction of R2 = 0.91, we present new state-of-the-art results across the largely varying sites. Combining the algorithm with an all terrain vehicle (ATV)-mounted image acquisition system, we demonstrate a feasible method for robust coverage and species distribution mapping of 225 ha of mixed crops at a median capacity of 17 ha per hour at 173 images per hectare.
Highlights
Crop mixtures of grass and clovers have many advantages such as reduced use of industrial fertilizer, lower production costs, increased protein self-sufficiency, increased yield stability, and improved feed quality [1]
The result section is ordered into sections regarding semantic segmentation of images, biomass composition prediction using images, and large scale mapping using images
A 2nd stage DeepLabv3+ model can be substituted by a previous FCN-8s 2nd stage model to isolate the contributions of each work
Summary
Crop mixtures of grass and clovers have many advantages such as reduced use of industrial fertilizer, lower production costs, increased protein self-sufficiency, increased yield stability, and improved feed quality [1]. The technique has been shown to work with deep learning based methods [2,3], there is a need to verify and enhance the stability of model predictions across different growth conditions, camera systems, and mixtures of species and varieties. While the use of morphological operations to distinguish grass from clovers based on the leaf widths proved valuable, it was highly sensitive to factors such leaf sizes, camera setup, lighting, and image resolutions. This led to a need for ongoing parameter-tuning, specific for each acquisition setup and growth stage
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have