Validation of the Soil Line Transformation Technique

Abstract

Farmers can use aerial images to assess crop conditions. Crop parameters, such as leaf area index (LAI) and normalized difference vegetation index (NDVI), are calculated from aerial images and used to determine crop status and infer other characteristics. However, direct comparison of images taken at different times in the growing season is not possible, as the images have to be calibrated to true reflectance or normalized to a standard. Various methods of image normalization have been proposed. Most of these methods rely on reflectance from standardized panels included in the image, which are often not available under farming conditions. The soil line is a relation between the red and near-infrared bands of a multispectral image and can be used to characterize bare soil pixels. Therefore, the soil line transformation (SLT) technique was developed to normalize images based on the soil line's characteristics. Utilizing the soil line concept, images can be normalized to a reference soil line, so that crop parameters from multi-temporal images can be compared directly. The objective was to verify that the SLT could effectively normalize images and that the NDVI extracted from SLT transformed images would be directly comparable to the NDVI values from true reflectance images. NDVI was calculated for a series of ten images that were all available as raw, true reflectance, and SLT transformed forms. The SLT program was used on a sub-image known to contain cotton and bare soil pixels. There was no significant improvement of the NDVI values when calculated from the SLT transformed images compared to the values from the raw images. Improvements to the SLT technique that may provide better performance are suggested. Normalization of images would be useful for comparison of multi-temporal data, but the existing SLT technique requires further development.