|M.Sc Student||Vered Dekel Ben-Yaakov|
|Subject||Attempt to Detect Water Stress in Melissa Plants Using|
hyperspectral Thermal Imaging
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Linker Raphael|
|Dr. Victor Alchanatis|
|Full Thesis text - in Hebrew|
This research examined the ability of a thermal sensor with 40 spectral channels to identify water stress in plants. The camera used was SPECIM LWIR with a microbolometric sensor (uncooled). Since it is not cooled, the camera temperature is not maintained and the camera may warm up during operation time. This heat radiation reduces the signal to noise ratio, reduces the camera sensitivity and can cause a phenomena called signal drift. All these pose challenges during image acquisition and limit the information that can be extracted from the images.
This study focused on two objectives: one was to characterize the camera based on the following parameters: Thermal drift, signal to noise ratio (SNR) and the ability to distinguish a temperature differences. The second challenge was to develop an image acquisition, calibration and data analysis methodology, which would overcome the limitation of the camera and allows water stress detection in plants in general and in the field particularly, by using its spectral characteristics.
The sensor characteristics and capabilities were tested using black bodies at the laboratory. These experiments helped determine the SNR of the camera and to detect thermal drift, if there is one.
To develop image acquisition, calibration and data analysis methodologies, Melissa plants underwent a controlled and gradual drying process in greenhouse conditions. Several sets of images including the stressed plants together with healthy ones and black body were acquired. Five days of image acquisition were held at noon, four of which were held under open sky with the camera placed above the vegetation and on the last day the camera was placed in the shade horizontally relative to plants, while the plants were under the sun.
At the processing stage, each image was calibrated for radiance values by using Empirical line method. The effective Temperature values of the dried and green vegetation was assessed by comparing the vegetation radiance curve to that of a black body at different temperatures. The emissivity signature was calculated on the basis of the effective temperature that was found.
A significant difference among the treatments in radiance and in effective temperature values was found only in images of the second and the last days of the experiment. A significant between the emissivity values was found only for the last day and only in two bands. Additionally, a large camera thermal drift was found.
Use the camera in more realistic situation, and in particular using it to monitor plant stress in field condition, would require significant improvements of the calibration and drift correction methodology so that no more than one black body would be required to appear in each frame. For these reasons, it appears that despite its attractive price, portability and relative ease of use, this camera is currently not suitable for agricultural applications such as crop stress monitoring.