Event Location: GHC2109
Bio: Stephen is a Project Scientist at the Robotics Institute and was a PostDoc at the RI from 2008-2010. Prior to arriving at the RI he completed his PhD in Australia conducting the research jointly at the Commonwealth Science and Industry Research Organisation and the University of Queensland. He has worked on a number of different agricultural, industrial and aerial vehicle applications focusing his research on computer vision systems that can localize, track, detect, segment to collect the necessary information for the perception task at hand.

Abstract: The single-most important practice in vineyard production that impacts economic viability is the balancing of potential yield with vine canopy size. Grape growers, however, suffer from a dearth of information about the state of their vines leading to uninformed manipulation of the vines and inefficiency in their production process. Growers want to measure and control the key indicators of vine performance — the amount fruit and the size of the vine canopy. However, current practices to measure the vines are coarse, inaccurate and tedious.

We are developing a combination of visual and laser sensing mounted on vineyard machinery that automatically provides dense measurements of vine performance indicators. We use machine vision to estimate the weight of the fruit growing on the vines and laser scanners to measure the size of the vine canopy. We validate our approach in experiments covering acres of vines, on several different grape varietals, on the east and west coast of the United States. Results demonstrate timely and high-resolution measurement of vine performance that offers growers the information they need to better manage their production.