Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Abstract. Precision farming is regarded as a key pathway for improving productivity and sustainability in rice. Advancements in Artificial Intelligence(AI) and robotics has allowed for the exploration remote sensing methods that gather and analyses large amounts of data to guide precision management. In this study, we explored the use of an infield ground robot and artificial intelligence to monitor early season growth in rice. The study was conducted in an experimental plot of transplanted rice located in Minamisoma, Fukushima, Japan. The robot has dimensions of 140 cm x 120cm width x145 cm and weighs 180 kg. It is equipped with RTK GPS, a variety of sensors and cameras (Sony FDR-X3000) for image capture. Data was collected manually and with robot at two week intervals after transplanting during June – July 2019. Based on manual data collected, the number of tillers per plant was categorized into 5 levels ranging from 1 to 40 tillers. AI to detect the tiller level of plants in images captured were developed for each survey date using YOLOv3 and the Darknet Framework. The AIs had detection rates (% of plants detected in image) of 60%, 35% and 17% while detection accuracies (% of correct detection) were 92%, 83%, 93% respectively. Spatial growth maps were developed by counting and plotting the level of detections in QGIS to guide location based management of the fields. The results demonstrate the potential of replacing conventional survey and prediction methods, for monitoring crop growth by using field robot and artificial intelligence.