Three-dimensional reconstruction of cucumbers using a 2D computer vision system

Real-time classification and grading horticultural crops are important issues in postharvest technology. Curvature, girth uniformity, length/girth ratio and volume (weight) are the basis of many classification standards for cucumbers. By three-dimensional (3D) reconstruction of the crop, these features could be easily achieved. This article describe a new image processing algorithm for 3D reconstruction of cucumbers, using lofting technique and mathematical calculations. The proposed algorithm firstly detect boundary of cucumber by image processing techniques and then fit a B-spline curve on it. The proximal and distal fruit ends as key-points are detected via applying mathematical operations on the curve. The virtual 3D reconstruction of cucumber would be obtained via revolving the points on 2D boundary curve around the fruit center line. Advantage of the proposed method is in quickly and accurately determination of the central line of cucumber and its geometrical features, as the volume, length and curvature of test samples were calculated with relative errors of less than 3.5%, 1%, and 4.5%, respectively. We concluded that the algorithm could be implemented in real-time classification processes efficiently.