| Summary: | Canopy temperature (T<sub>c</sub>) is used to characterize plant water physiology, and thermal infrared (TIR) remote sensing is a convenient technology for measuring T<sub>c</sub> in forest ecosystems. However, the images produced through this method contain background pixels of forest gaps, thereby reducing the accuracy of T<sub>c</sub> observations. Extracting T<sub>c</sub> data from TIR images is of great significance for understanding changes in ecosystem water status. In this study, a temperature threshold method was developed to rapidly, accurately, and automatically extract forest canopy pixels for T<sub>c</sub> data obtention. Specifically, this method takes the temperature corresponding to the point with a slope of 0.5 in the curve composed of the normalized average temperature and the normalized cumulative number of pixels as the segmentation threshold to separate the forest gap pixels from the forest canopy pixels in the TIR images and extract the separated forest canopy pixels based on the pixel coordinates for T<sub>c</sub> data obtention. Taking the T<sub>c</sub> values, measured using a thermocouple, as the standard, T<sub>c</sub> extraction using the new temperature threshold method and traditional methods (the Otsu algorithm and direct extraction) was compared in cork oak plantations. The results showed that the temperature threshold method offered the highest extraction accuracy, followed by the direct extraction method and the Otsu algorithm. The temperature threshold method was determined to be the most suitable for extracting T<sub>c</sub> data from the TIR images of cork oak plantations.
|
|---|