Multimodal image stitching algorithm for weed control applications in organic farming

Abstract

In computer vision applications based on machine learning algorithms, a ground truth of datasets is necessary for efficient training of classification or data mining algorithms. Due to the increasing computer power, machine vision systems are developed for robotic, medical and industrial applications. This paper puts the main emphasis on the SURF algorithm for agricultural image applications in organic farming. The algorithm is used for a multimodal image stitching algorithm for biological weed control in the agricultural sector. Based on the algorithm a ground truth agricultural crop map is produced to allow a proper detection of weed. The calculation effort of the algorithm is examined and the experimental results are presented.