The Grapevine Rover (GRover) developed for this project is an adaptable proximal-sensing platform designed to provide rapid, high quality information on plant performance in the field. GRover integrates a light radar (LiDAR) and positioning sensors to produce precise (± 3 mm), high resolution, three-dimensional data on vineyard structure. Scans of vine rows are converted to ‘point-clouds’, which can then be visualised as well as analysed. In addition, the LiDAR used in this study provides data on the laser reflectance of the vine structure, allowing different tissue types, such as leaves, wood or fruit, to be distinguished. Calibration data against destructive and non-destructive (manual) measures of vine size have been developed, based on LiDAR scans at a number of sites, for wood volume, pruning weight and canopy size. The system worked well within a given vineyard with good correlations obtained between LiDAR scans and manual measures for these parameters, demonstrating a high effectiveness at distinguishing differences between individual vines. However, current limitations in the analytical tools used, limit the strength of the calibrations between sites where there can be significant differences between canopies associated with vineyard set-up and management systems and their effects on canopy architecture and size.
Plant breeding programs involve large numbers of crosses and there is an associated need to generate extensive datasets on the growth and performance of the progeny from these crosses in field environments. This project addresses this gap for grapevines. New technologies are needed to automate the process of collecting and analysing the large datasets required. Large scale measurement of many aspects of an organism’s outward characteristics, or phenotype, is termed ‘phenomics’. Phenomic tools developed to measure vine traits also have potential to be used at a commercial scale for vineyard management, for example, being able to assess disease incidence, the light environment around the bunch zone or detailed yield estimation. To address this gap, we developed the Grapevine Rover (GRover), a mobile and versatile phenotyping platform, which can be used to test a range of proximal sensing instruments that might be utilised to replace conventional, manual measurements, such as those used to estimate vine growth.
GRover was constructed in collaboration with the High Resolution Plant Phenomics Centre (HRPPC) part of the Australian Plant Phenomics Facility (APPF) in Canberra, Australia. GRover’s frame was made from lightweight aluminium. It is powered by two electric wheels mounted on the front. There is one free-pivoting wheel in the rear. It has a collapsible mast on the front of the frame that is 3 m tall when deployed for measurements. The primary instrument currently attached to GRover’s frame, is a SICK LMS-400 LiDAR (Light Radar), which is used for measuring vine size and structure. LiDAR operates in a similar way to sonar, but uses light instead of sound. A LiDAR emits rapid beams of light and detects the reflected pulses. This allows a three dimensional structure to be determined. GRover is also equipped with an LMS-400 eye-safe red laser (605 nm) that is capable of 70 degree scans, producing a two-dimensional field of view 270 times every second. The third-dimension is achieved by moving the LiDAR i.e. GRover through the vineyard. As GRover moves along the vineyard rows with the LiDAR angled towards the canopy, a wheel encoder tracks the linear distance travelled. In this way, three-dimensional point clouds of vine growth can be produced. Additionally, the LMS-400 LiDAR is capable of not only measuring the X,Y,Z coordinates of a given laser return but it also provides information about the remissivity, or reflective properties of the surface.
The LiDAR setup was used successfully to capture a variety of vine size and growth parameters at a range of different vineyard locations. GRover was able to produce scans of canopies of different management styles at any time during the growing season. Additionally, by using simple computational methods, metrics from LiDAR scan data were shown to be strongly correlated with destructive measurements of pruning weight. Work was also carried out to determine how well LiDAR scans correlated with plant growth on the basis of leaf area and trunk and cordon weight. Each trait was measured on vines grown in four different regions (Adelaide, Nuriootpa and McLaren Vale, South Australia and Irymple, Victoria).
Field trials showed that LiDAR scans involving multiple measurements over time 6 were highly predictive at measuring parameters of vine size within a vineyard, however calibration accuracy was limited when comparing different locations. Correlations between LiDAR scans, leaf area and pruning weight, differed significantly between some field sites. However, within a given field site, the LiDAR could be used to assess an applied treatment, such as: reduced irrigation, elevated temperature or a change in management as effectively as using destructive measurements, but with much greater speed and scale. Initial attempts to utilise the system for yield assessment in typical Australian commercial vineyards were not successful because bunches were obscured by leaves and did not appear to have different reflectance values to leaves.
The deployment of the high resolution LiDAR system in a vineyard for relatively large scale testing against vine structural parameters has successfully demonstrated its potential. Although there were issues with correlations across multiple locations, it is anticipated that improvements in data analysis and collection could be made that would ameliorate this issue. For example, a LiDAR with different properties could be used. The LMS-400 offers an excellent scanning rate and precision but is affected by sunlight during outdoor use. Alternative positioning systems could also be used and more sophisticated algorithms for analysis could potentially be developed.
Continued development of phenomic tools such as GRover require a highly collaborative team effort involving both technology and biology skills, particularly suited to a diverse organisation such as CSIRO. Further development of LiDAR technologies could yield cost effective tools for management purposes; for example tools that accurately assess canopy structural parameters such as light penetration.