Abstract. This paper proposes an automated curvatureinvariant method for crop row identification in images, acquired from Unmanned Aerial Vehicles. The proposed approach is composed of three steps: Step (i)image segmentation is performed based on the combination of Excess of Green Index and Otsu’s method to obtain the region of interest. Step (ii)local feature extraction computes the local orientation using a radial search descriptor employed to detect the predominant orientation of perpendicular row signals. Step (iii)lines are detected by linking the responses for each window using a gradient’s peak arrangement method. The preliminary results indicates the robustness and effectiveness of the approach for crop row identification.