| This research investigated the technology for developing a multi-spectral imaging sensor to detect corn nitrogen deficiency based on crop canopy reflectance information from Green, Red, and Near-Infrared bands of radiation spectra. A background segmentation algorithm was developed to filter the soil background from the canopy images to eliminate the effects of background noise on leaf reflectance analysis. A dynamic image calibration model was developed to compensate for ambient light factors to obtain consistent leaf reflectance with similar nitrogen deficiency levels. The developed algorithm and model improved the capability of the sensor in detecting corn nitrogen deficiency. Investigation results indicated that the correlations (R) between reflectance of the corn plant and corn leaf chlorophyll (soil plant analysis's development; SPAD) values in Green, Red, and NIR channels were 0.86, 0.79 and 0.15, respectively. The correlations between vegetation indices and the SPAD values, in GNDVI (Green and NIR channel), NDVI (Red and NIR channel) and ratio (NIR/Green) were 0.85, 0.85 and 0.85, respectively.;In addition, it was found, via the effect of leaf shadow area investigation, that the ratio of NIR/G showed not only strong correlations based on all leaf conditions (whole areas, bright areas or shadow areas), but also low computation intensity. All the above results indicated that the developed multi-spectral imaging sensor can be used to estimate corn SPAD values "on-the-go" with reasonable accuracy during field operations.;Integrating the multi-spectral imaging sensor with a variable-rate nitrogen application system, can provide the fundamental means for performing site-specific variable-rate nitrogen side-dressing applications. Using the estimated SPAD values to support sensor-based variable rate nitrogen application at specific crop growth stages, has the potential of not only increasing the yield, but also of reducing the total N consumption. This result, however, was obtained based on only a one-year field test. Additional field tests are recommended to validate the effectiveness of sensor-based N application. |
