Development of a Ground-Based Hyperspectral Remote Sensing System for Crop Health Sensing
Crop Production Systems Research Unit
2012 Annual Report
1a.Objectives (from AD-416):
To develop a new ground-based hyperspectral remote sensing system that will provide crop health sensing and establish a ground truthing platform to facilitate development of a rapid analysis system with high spectral and spatial resolution for aerial variable-rate chemical application.
1b.Approach (from AD-416):
A new ground-based hyperspectral remote sensing system is proposed to permit fast evaluation of crop growth and provide ground truth measurement to validate the analysis of aerial image analysis. The proposed system is a visible near infrared hyperspectral imaging system (VNIR-HIS) with an effective spectral range from 400 nm to 900 nm. The hyperspectral camera is a 14-bit PCO1600 CCD (charge-coupled device) high resolution camera. The PCO1600 camera has resolution of 1600 x 1200 pixels and is thermoelectrically cooled to -50ºC compared with ambient temperature. Image data transfer from the camera to computer is through IEEE 1394 “firewire” link. In front of the camera is an ImSpector V10E spectrograph with a 30 micron entrance slit. The Prism-Grating-Prism spectrograph disperses incoming light to generate spectral information of the target. Push-broom scanning technique is used in the imaging system to produce continue scan of the target. Furthermore, the variable binning capability of the camera allows image acquisition at user specified spatial and spectral resolutions. The system will be developed based on an existing prototype developed by a scientist at Mississippi State University. The prototype was originally developed for detection of aflatoxin in corn, and will be converted to be used in field measurement. To implement push-broom scanning in field, the VNIR-HIS will incorporate a patented line scanning technique that requires no relative movement between the target and the sensor. It will scan an input image within the focal plane of a front lens and disperses an input image line (via the spectrograph) vertically as a function of the spectral wavelengths. The front lens can be selected to fit for field applications. The system utilizes a linear motor to move the front lens for the focal plane line scanning. This kind of hyperspectral focal plane scanning eliminates the requirement of a mobile platform in a pushbroom scanning system.
The VNIR-HIS has great potential for field level crop damage assessment. It can provide high spectral and spatial resolution image at the same time. Field level crop damage assessed at canopy level generally used a multispectral camera or a spectrometer. Comparing with multispectral cameras, the fine spectral resolution from the VNIR-HIS reveals more subtle changes in canopy reflectance, especially when crops are under stress conditions. The VNIR-HIS also outperforms a spectrometer which only collect spectral information from a certain area that contains mixed spectral signatures. The fine spatial resolution enables better image segmentation of canopy structures. Finally, the use of focal plane scanning in data acquisition eliminates the requirement of a mobile platform. The imaging system can be mounted on field equipment such as a tractor or sprayer. Data acquisition could be implemented in a stop-and-go mode. In this case, the field equipment will first travel to the survey site. It will then stop for data acquisition and subsequently move to the next survey site.
Two greenhouse experiments have been conducted to evaluate a hyperspectral imaging system in detection the onset of the soybean injury caused by glyphosate application. Results indicated that the hyperspectral imaging technique has a good potential to identify crop herbicide injury in early growth stages of soybean plants. Further, a modified derivative index method has been developed for early crop injury detection.