CRowNet: Deep Network for Crop Row Detection in UAV Images

Nowadays, the development of robots and smart tractors for the automation of sowing, harvesting, weeding etc. is transforming agriculture. Farmers are moving from an agriculture where everything is applied uniformly to a much more targeted farming. This new kind of farming is commonly referred to as precision agriculture. However for autonomous guidance of these agricultural machines and even sometimes for weed detection an accurate detection of crop rows is required. In this paper we propose a new method called CRowNet which uses a convolutional neural network (CNN) and the Hough transform to detect crop rows in images taken by an unmanned aerial vehicle (UAV). The method consists of a model formed with SegNet (S-SegNet) and a CNN based Hough transform (HoughCNet). The performance of the proposed method was quantitatively compared to traditional approaches and it showed the best and most robust result. A good crop row detection rate of 93.58% was obtained with an IoU score per crop row above 70%. Moreover the model trained on a given crop field is able to detect rows in images of different types of crops.