Harmonic analysis is a fundamental tool in signal and image processing that is essential to many applications, including pattern identification, audio and speech recognition, image compression, and many more. In-depth discussion of harmonic analysis methods and their relevance to signal and image processing are provided in this study. This work proposes a novel method for classifying land features and detecting hotspots based on harmonic analysis of NOAA/AVHRR annual composite pictures. Due to their extensive coverage, high frequency, and free availability, the NOAA/AVHRR photos were chosen. Three steps make up the proposed algorithm's primary section. In order to rectify geometric distortions and perform radiometric calibration of the data, the NOAA/AVHRR photos first go through preprocessing. As a result, measurements are reliable and constant across the whole time period. The preprocessed images are used to identify cloud and water pixels, and harmonic analysis is used to examine the 10-year time series of AVHRR images. The NDVI (Normalised Difference Vegetation Index) and various AVHRR band values are temporally changed in the phase and amplitude images created by the harmonic analysis. While the amplitude information represents the size of these shifts, the phase information sheds light on their time and frequency. Images of various bands' amplitudes are processed using image-processing algorithms. The objective of this stage is to identify hotspots, or places that have undergone major changes over time, and to categorise the region of interest using these traits. The proposed method produced categorization results with a high accuracy rate of over 91%.
Key words: Image processing, NOAA/AVHRR, time series, normalization.
|
