Title : From gaming to high throughput phenotyping: Use of an X-box camera to model soybean 3D structure and morphological traits
Abstract:
The world population increase coupled with the depletion of natural resources are a dangerous combination that is threatening humanity and other life around the world. At the current rate, crop production must be doubled by 2050 in order to meet population demand. High throughput phenotyping in agriculture has drawn increasing interest in recent years. This research introduces an unprecedented sensor, Kinect sensor, for 3D assessment of crops on a high-clearance tractor platform. The development of crop canopy 3D models, and calculation of canopy traits such as plant height, number of pods per node, etc can be achieved through light detection and ranging scanners (LIDAR). However, its complexity and high cost makes this technology difficult to implement for research and breeding proposes. The advantage of Kinect sensor besides its price, $200 per unit, is the ability to capture depth images from as many viewing angles as possible. Depth images can then be converted to point clouds, which will be further processed and combined to form a single and complete 3D model of the plant. Those 3D models can help to extract phenotypes of plants such as plant height, canopy area, number of pods per plant, and node, will provide valuable information for agronomists, breeders, geneticists, and physiologist to study their correlation with growth, resistance to biotic and abiotic stresses, and ultimately yield. Except Kinect sensors, some other sensors are also deployed on the platform, including ultrasonic sensors, thermal sensors, and spectrometers to monitor plant height, crop canopy temperature, and canopy reflectance, respectively.
