Improving pasture growth assessment using machine vision

Improving pasture growth assessment using machine vision

Frequent pasture dry matter and quality assessments are vital for determining grazing schedules and managing the feed base, and ultimately profitability. Dry matter is typically assessed from manual observations or tools such as a rising plate meter that are labour-intensive and can be prone to subjective variability. Spatial dry matter assessment can be automated using satellite imagery which may be limited by cloud cover, and on-the-go height or multispectral sensors which require manual operation to traverse the field. There is potential for infield machine vision cameras to provide pasture quality and quantity features at a daily time scale, while also significantly reducing labour and increasing accuracy and repeatability in feed base assessments. This system could also detect grazing events which could be used to automate record keeping and improve pasture management. Field and image datasets have been collected over two irrigated pasture seasons at a field within Tasmanian Institute of Agriculture’s Dairy Research Farm at Elliott where perennial ryegrass is dominant (90-95%). Daily images were collected from cameras, analysed and compared with estimates of dry matter quantity from weekly-fortnightly C-Dax measurements. A machine vision system was developed that detected pasture dry matter amount with r = 0.715 and RMSE = 381.5 kg DM/ha. The camera system used trends of assessed dry matter quantity to accurately detect grazing date. The system could monitor daily pasture growth rate and assist in making grazing and management decisions in pasture production systems. Further work includes evaluating other machine vision properties and different pasture compositions.