Environmentally Adaptive Segmentation Algorithm for Outdoor Crops Using Gaussian Mixture Model

Abstract

Crop segmentation from outdoor images is still an open problem. In this paper, we proposed a novel crop segmentation method using Gaussian Mixture Model (GMM), which is robust and not sensitive to the challenging outdoor light conditions and complex environmental elements. The method mainly consists of two stages, supervised learning stage and segmentation stage. The GMM is utilized in the former stage to establish crop color model in the HSI color space and a decision function is provided in the latter stage to realize the final crop segmentation. Comparing experimental results show that our method outperforms the other commonly used methods in yielding the highest performance of 94.91% with the lowest standard deviation of 3.14%.