RESUMEN
Breeding higher yielding forage species is limited by current manual harvesting and visual scoring techniques used for measuring or estimation of biomass. Automation and remote sensing for high throughput phenotyping has been used in recent years as a viable solution to this bottleneck. Here, we focus on using RGB imaging and deep learning for white clover (Trifolium repens L.) and perennial ryegrass (Lolium perenne L.) yield estimation in a mixed sward. We present a new convolutional neural network (CNN) architecture designed for semantic segmentation of dense pasture and canopies with high occlusion to which we have named the local context network (LC-Net). On our testing data set we obtain a mean accuracy of 95.4% and a mean intersection over union of 81.3%, outperforming other methods we have found in the literature for segmenting clover from ryegrass. Comparing the clover/vegetation fraction for visual coverage and harvested dry-matter however showed little improvement from the segmentation accuracy gains. Further gains in biomass estimation accuracy may be achievable through combining RGB with complimentary information such as volumetric data from other sensors, which will form the basis of our future work.
RESUMEN
BACKGROUND: In-field measurement of yield and growth rate in pasture species is imprecise and costly, limiting scientific and commercial application. Our study proposed a LiDAR-based mobile platform for non-invasive vegetative biomass and growth rate estimation in perennial ryegrass (Lolium perenne L.). This included design and build of the platform, development of an algorithm for volumetric estimation, and field validation of the system. The LiDAR-based volumetric estimates were compared against fresh weight and dry weight data across different ages of plants, seasons, stages of regrowth, sites, and row configurations. RESULTS: The project had three phases, the last one comprising four experiments. Phase 1: a LiDAR-based, field-ready prototype mobile platform for perennial ryegrassrecognition in single row plots was developed. Phase 2: real-time volumetric data capture, modelling and analysis software were developed and integrated and the resultant algorithm was validated in the field. Phase 3. LiDAR Volume data were collected via the LiDAR platform and field-validated in four experiments. Expt.1: single-row plots of cultivars and experimental diploid breeding populations were scanned in the southern hemisphere spring for biomass estimation. Significant (P < 0.001) correlations were observed between LiDAR Volume and both fresh and dry weight data from 360 individual plots (R2 = 0.89 and 0.86 respectively). Expt 2: recurrent scanning of single row plots over long time intervals of a few weeks was conducted, and growth was estimated over an 83 day period. Expt 3: recurrent scanning of single-row plots over nine short time intervals of 2 to 5 days was conducted, and growth rate was observed over a 26 day period. Expt 4: recurrent scanning of paired-row plots over an annual cycle of repeated growth and defoliation was conducted, showing an overall mean correlation of LiDAR Volume and fresh weight of R2 = 0.79 for 1008 observations made across seven different harvests between March and December 2018. CONCLUSIONS: Here we report development and validation of LiDAR-based volumetric estimation as an efficient and effective tool for measuring fresh weight, dry weight and growth rate in single and paired-row plots of perennial ryegrass for the first time, with a consistently high level of accuracy. This development offers precise, non-destructive and cost-effective estimation of these economic traits in the field for ryegrass and potentially other pasture grasses in the future, based on the platform and algorithm developed for ryegrass.
