Cotton yield estimation based on similarity analysis of time-series NDVI

Multi-temporal remote sensing images contain more crop yield information than mono temporal images, and it is very significant to mine information from multi-temporal remote sensing data for improving the precision of crop yield estimation. In the paper, taking the cotton field of First Regimental, Agriculture First Division, Xinjiang Production and Construction Corps as the studying area, a method of cotton yield estimation was proposed by integrating the concept of cotton growing area with similarity analysis of time-series NDVI data. Firstly, the NDVI was determined as the dominant factor of cotton yield estimation through correlation analysis between vegetation index and cotton yield from all sampled plots. Secondly, the whole studying area was divided into several cotton growing areas according to cotton variety and soil condition. And then the linear-fitting analyses were used to acquire the coefficient of yield model for each growing area. Finally, multiple linear regression coefficients for each cotton pixel were determinated by similarity analysis between NDVI vectors from unknown-yield cotton pixels and all known ones as the investigated yields. Thus, cotton yield estimation throughout the whole studying area was realized by time-series NDVI data. The analyses show that the coefficient?of determination (R2) between the estimated and investigated yield can reach to 0.77, which indicates that the method is reasonable and adaptable.