Modelling grape-vines from a two-dimensional image has a number of interesting problems which have to be overcome. Due to their growth patterns and the surrounding environment vines tend to grow in ways which result in a large number of occlusions from a single perspective, making identification difficult. In addition the presence of wires, posts and other foreign objects increase complexity. We propose a method which uses the contours extracted from an image to identify canes. An emphasis has been made on identifying only sections of canes which are not occluded. Our method extracts the contours, creates a model of all edges and pairs up edges which contain similar properties such as their width and direction. High quality edge pairs are then selected to produce partial cane models. These models have been quantitatively assessed by comparing their structure to near perfect hand drawn versions. From this analysis we have found a mean overall accuracy of 71.3% (σ = 5.9%) for all images used in our study. These results are promising as the models produced have a high correspondence with the ground truth, although are missing some information. Finally we propose several techniques which could be investigated in the future in order to further improve the accuracy and reliability of our method.