Volume Estimation of Convex 3D Objects from three 2D Projective Images

Dr Josh Chopin, Professor Stan Miklavcic and Dr Hamid Laga

In this project we propose a new recipe for estimating the volume of convex 3D objects, such as fruits and vegetables, using information obtainable from three orthogonal projections of the object.  In general, accurate estimation of the volume of objects requires high resolution images, multiple viewpoints and 3D reconstruction. These requirements create a time consuming and computationally expensive procedure, making automated phenotyping impractical. To estimate the volume, the recipe that we propose requires only three orthogonal images of the object and does not require any 3D reconstruction. The method is based on a novel lower volume bound on this volume, combined with an application of a well-known upper bound.

We applied the method to a number of simple geometric shapes whose true volume can be computed using well known formulae. We also applied the method to several fruit and vegetables, validating our results using a variant of the water displacement method. The results showed that our method provides a fast, accurate and simple means of estimating volume of 3D objects. Future directions include non-convex objects and an application to biomass.

Approximately Convex Fruit and Vegetables


Chopin, J, Laga, H & Miklavcic, SJ 2017, 'A new method for accurate, high-throughput volume estimation from three 2D projective images', International Journal of Food Properties, pp. 1-14.

Areas of study and research

+ Click to minimise