Unveiling Crop Growth Dynamics Through Time Series Multispectral Observation

The diverse climates and farming practices across Asia pose a significant challenge for automatically classifying and mapping a wide variety of crops. Existing methods often struggle to account for this heterogeneity, leading to inaccurate crop maps. This study proposes a novel approach that leverages dynamic time warping (TWDTW) analysis of time series multispectral data to address this limitation. We focus on the Kanigiri region of the Prakasam district in India, utilizing Sentinel-2 satellite data from 2019 to 2022. The data includes monthly average NDVI (Normalized Difference Vegetation Index) for each year, along with a reference file containing location, crop type, and crop duration information. The proposed method utilizes both the temporal (time-related) and spatial (location-related) information within the NDVI data to identify and classify different crop types. The TWDTW technique plays a crucial role in finding the optimal alignment between NDVI data from different years. This alignment allows for a detailed comparison of crop growth patterns and their evolution over time, enabling the identification of specific growth stages. The resulting maps provide valuable insights into agricultural land use within the Kanigiri region. This information can be used to support decision-making for sustainable farming practices and promote food security. In simpler terms, the proposed approach utilizes satellite data and advanced analysis techniques to automatically identify and map different crop types over several years. This empowers farmers and policymakers to gain a deeper understanding and improve the management of agricultural land, ultimately contributing to a more sustainable agricultural system.