Page Banner

United States Department of Agriculture

Agricultural Research Service

Title: Modeling and Calibration of a Multi-Spectral Imaging Sensor for in-Field Crop Nitrogen Assessment

Authors
item Kim, Yunseop
item Reid, John - JOHN DEERE

Submitted to: Applied Engineering in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: September 11, 2006
Publication Date: November 1, 2006
Repository URL: http://handle.nal.usda.gov/10113/3126
Citation: Kim, Y., Reid, J.F. 2006. Modeling and calibration of a multi-spectral imaging sensor for in-field crop nitrogen assessment. Applied Engineering in Agriculture. 22(6):935-941.

Interpretive Summary: This paper introduces a vision-based multi-spectral imaging sensor for in-field real-time assessment of plant nitrogen stress in corn crops. The nitrogen status of plants is estimated by measuring reflectance of plant canopies. This paper describes how to formulate a mathematical model of reflectance and calibrate the multi-spectral imaging sensor. A sensor response model was formulated based on solar energy transformation and consists of several parameters such as ambient illumination, average gray-level value of plant image, and exposure and gain of the imaging sensor. Calibration procedure determines the relationship between the sensor’s reflectance response and that of a known target reflectance by setting the imaging sensor, ambient illumination sensor, and an absolute reflectance panel normal to the solar position. Sensor calibration proved the validity of the reflectance model such that dynamic adjustment of the camera parameters according to gain change maintained the linearity of log response of the imaging sensor. The calibration constants were obtained for four imaging sensor units. Comparable performance across the units was achieved by sensor calibration. The difference in some units were found and mainly caused by erroneous AI calibration values. Modified AI calibrations were determined, which improved the MSIS responses by producing nearly identical responses over all units except one unit with image formation problems.

Technical Abstract: Assessment of nitrogen content from crop leaves has been of interest worldwide to help growers adjust N fertilizer rates to meet the demands of the crop. A multi-spectral imaging system was developed for in-field real-time assessment of plant nitrogen stress in corn (Zea mays L.) crops indicated by plant reflectance and measured using a vision-based multi-spectral imaging sensor (MSIS). The objectives of this paper were to formulate a mathematical model of reflectance and calibrate the MSIS system. A MSIS response model was formulated based on solar energy transformation. The MSIS was calibrated with a known reflectance target to determine the relationship between the MSIS reflectance response and that of a known target reflectance by setting the MSIS, a sensor for ambient illumination (AI), and an absolute reflectance panel normal to the solar position. Sensor calibration proved the validity of the reflectance model such that dynamic adjustment of the camera parameters according to gain change maintained the linearity of log response of the MSIS. Calibration constants were carefully determined and validated with reflectance responses of multiple MSIS units. The calibration constants were obtained for four MSIS units paired with AI sensors. Comparable performance was achieved across the units. The difference in some units were found and mainly caused by erroneous AI calibration values. Modified AI calibrations were determined, which improved the MSIS responses by producing nearly identical responses over all units except one unit with image formation problems.

Last Modified: 9/21/2014
Footer Content Back to Top of Page