Row detection is a necessary first step for the analysis of high resolution remote sensing images of vine fields. This paper presents an image processing algorithm for automatic row detection which implements an active contour model. The model consists of a network of lines whose purpose is to adjust to the vine rows. After a rough initialization, the convergence algorithm proceeds step by step by minimizing an energy criterion. The energy is composed of two parts: an internal energy reflecting the internal geometric constraints undergone by the network, and an external energy which represents the effect of the image on the network. The algorithm proves to perform properly: the rows are precisely detected. This approach is the basis for future developments which consist in drawing a vigor map from measures made along the rows.