Affiliations: [a]
Department of Software Engineering and Artificial Intelligence, School of Computer Science, University Complutense of Madrid, Spain
| [b]
Department of Computer Architecture and Automatic, School of Computer Science, University Complutense of Madrid, Spain
| [c]
Centre for Automation and Robotics, (CSIC-UPM), Arganda del Rey, Madrid, Spain
Correspondence:
[*]
Corresponding author. Martín Montalvo, Department of Software Engineering and Artificial Intelligence, School of Computer Science, University Complutense of Madrid, Spain. E-mail: mmontalvo@ucm.es.
Abstract: In this article, a new strategy to detect green plants in a maize crop has been developed. This strategy is based on five main stages: segmentation, reduction of dimensionality based on PCA, Otsu thresholding, threshold combination and final thresholding. Images are taken in RGB color model and transformed to grayscale image using basic vegetation index and PCA method to reduce the dimensionality of data. After this step, a combination of Otsu thresholds by PCA is designed in order to generate a new threshold and binarize the previous grayscale image. The performance of the proposed strategy is validated with an image set and compared with other strategies, being the best of all the methods analyzed in a quantitative mode.
