Key points are not available for this paper at this time.
In this letter, the relationship between corn crop progress stages and fractal dimension derived from remote sensing data has been revealed. Due to the effects of soil background and farming practices on emerged and harvested stages of the corn crop, the roughness of their corresponding normalized difference vegetation index (NDVI) images significantly changes. Therefore, image fractals, which normally indicate such roughness, are proposed to detect corn progress stages. There are four steps in the proposed procedure: 1) NDVI maximum composite process is conducted to eliminate the influence of cloud cover or missing data; 2) Cropland Data Layer (CDL) is used to remove noncorn pixels; 3) fractal dimension is calculated to estimate the spatial roughness of NDVI image; and 4) curve fitting is used to detect peaks that infer the certain progress stages. It is worth mentioning that a dimensionality-reduction-based differential box-counting algorithm is developed to estimate the fractal dimension of NDVI image in an irregular region of interest. The algorithm is applicable for masked NDVI image. Experiments based on Moderate Resolution Imaging Spectroradiometer NDVI time series and CDL of the State of Iowa are conducted. Comparison of the results with corn crop statistics from the National Agricultural Statistics Service indicates that the proposed method is able to detect corn emerged and harvested stages with success.
Shen et al. (Tue,) studied this question.