Automatic detection of field furrows from very high resolution optical imagery

Abstract

Surface features such as furrows are here detected using high spatial resolution images such as those provided by IKONOS (pixel size 1 m × 1 m in panchromatic mode) and WorldView-1 instruments (pixel size 0.5 m × 0.5 m), through a new detection methodology based on image analysis. This methodology includes four different steps. First, segmentation of the whole image is performed in order to label each agricultural field individually. Then, for each field, mathematical morphological processes are applied (erosion followed by functional geodesic reconstruction) in order to enhance the image contrast. Then, an automatic thresholding process is applied to construct a binary image of the furrows or other features in the field. In order to reduce the noise due to the overlap between the classes of furrow and background, the a priori information that furrows are straight lines exhibiting regular patterns is introduced. The Hough transform is applied to detect straight lines (represented in polar coordinates) and furrow directions as accumulations on the angle axis in the Hough space. The proposed method was applied on a WorldView-1 image acquired over the Lokna catchment within the Plava watershed in Russia (around 180 km2) and validated against ground truth observations.