An Algorithm for Separating Guidance Information from Row Crop Images

Astatistical-based algorithm was developed for selecting a threshold to segment row crop images into classes of information corresponding to the crop canopy and soil background. Systematic image sampling was used to determine an estimate of the population of pixels in an image. An automatic-iris lens and an optical filter were used to enhance image contrast to minimize computational image processing requirements. The threshold was selected using a Bayes classifier. The two image intensity classes were assumed to be independent Gaussian random variables. The distribution parameters used in classification were estimated using the method of moments. Data is presented on the development and performance of the thresholding algorithm on 16 representative row crop images. The algorithm was an effective classification technique for segmenting row crop images into crop canopy and soil background classes. In images with an extreme percentage of crop canopy or soil background, a priori knowledge of the class mixture was needed for convergence to the correct image thresholding parameters.