A plant-tracking system in a crop field is presented, it uses sequences of outdoor images taken from the field under natural illumination. Images are segmented with a gray level threshold and the result is the input to the tracking system. Plants are identified through a clustering of the two-dimensional regions obtained from the segmentation. A shape description that handles the splitting and merging of regions is computed for each plant. A Hough transform-like technique is applied to select matches that are consistent with the rigidity constraint. Camera motion parameters are found through a least squares minimization. Results with real images are presented.