Radiometric normalization of multi-temporal images based on image soil lines

A radiometric normalization procedure utilizing the soil line concept is developed and tested. The routine, referred to as the soil line transformation (SLT) technique, is developed to provide a system for transforming multi-temporal images to a common reference scene. The routine involves developing soil lines not only for the bare soil images, but also for images with vegetative cover, assuming that some pixels within the entire image would represent bare soil conditions or pseudo-invariant features. The procedure uses the soil lines developed for the image-to-be-transformed and the reference image to form transforming equations using pixels within a specified interval of both soil lines. In this research, the SLT technique is evaluated using multi-temporal images of crop development. Digital aerial images of two corn (Zea mays L.) fields are used to test the SLT routine based on expected change in pixel intensity due to crop development. The SLT technique is also compared against an alternative histogram matching technique based on the percentage of pixels matching the expected growth pattern away from the soil line. For two fields over two growing seasons, the SLT technique outperformed histogram matching, resulting in at least 87% of pixels demonstrating the expected growth pattern between the images.