Development of a field-based high-throughput mobile phenotyping platform

We developed a mobile platform to carry various sensors for high-throughput field phenotyping.The hardware and software design was modular, allowing easy sensor addition and removal and flexible system expansion.We examined possible position errors caused by time delays in data acquisition.We examined the effects of ambient light and temperature on sensor readings.We proposed a method of correcting IRT measurement error with an additional ambient air temperature sensor. In this study, a mobile, field-based, high-throughput phenotyping platform was developed for rapid measurement of plant characteristics. The platform consisted of three sets of sensors mounted on a high-clearance vehicle. Each set contained two infrared thermometers (IRT), one ultrasonic sensor, one Crop Circle multi-spectral crop canopy sensor, and one GreenSeeker crop sensing system. Each sensor set measured canopy temperature, crop height, and canopy spectral reflectance of a plant plot. Thus, three plots were measured simultaneously in a single pass. In addition to the sensors, the platform was equipped with a laser distance sensor to measure the height of the sensor beam and an RTK-GPS system that provided precise, accurate position data for georeferencing sensor measurements. Software for collecting, georeferencing, and logging sensor data was developed using National Instruments LabVIEW on a laptop computer. The hardware and software design was modular, allowing easy addition and removal of sensors and flexible system expansion. The fast sampling rates for sensors allowed the phenotyper to operate in field at a ground speed of 3.2km/h. Two verification tests were conducted to evaluate the phenotyping system. In the first test, data timestamps were analyzed to determine if the system could collect data at the required rates and if the time delays would cause significant position errors. Test results showed that data from all sensors were received within the desirable time frame and the largest position error was 17.9cm when the phenotyper was moving at a speed of 3.2km/h. The position errors can be corrected during data post processing. The second test determined whether changes in ambient light and ambient temperature had statistically significant effects on the accuracy of the sensor measurements. For the IRT sensors, a correction method using ground truth temperature measurement made during two periods within a day was recommended to correct the errors in surface temperature measured by the IRTs.