Precision Plowing: An Approach for Clods Size Estimation via ECOC Classifier

Automatic agricultural guidance requires an accurate soil quality assessment during tillage operations. This is particularly true for assisted plowing, where the tractor speed and the plow-specific settings (pitch, sinking, and aperture) should be adjusted to guarantee an optimal soil crumbling. In this paper, we present a novel method for estimating soil clod size after the plowing activity, using a mono-camera mounted on the cabin of a tractor. Our technique relies on the Bird's Eye View reconstruction of a soil patch and its classification using an Error-Correcting Output Codes (ECOC) classifier, trained on features extracted from grayscale images. The main novelty of our approach lies in the real-time implementation on a moving vehicle. Furthermore the use of the ECOC classifies yields precise and efficient assessment of the clod size. Experimental results on real in-field collected data demonstrate that our estimator is sufficiently robust and accurate to provide a solid basis for automatic adjustment of plow settings during agricultural operations with a moving vehicle.