Eventos Anais de eventos
COBEM 2019
25th International Congress of Mechanical Engineering
A FAST MULTILEVEL THRESHOLDING APPROACH FOR IMAGE SEGMENTATION USING GENETIC ALGORITHM AND AN APLICATION IN CROP ROWS DETECTION
Submission Author:
Pedro Ventura de Oliveira , MG
Co-Authors:
Pedro Ventura de Oliveira, Luciano Lima
Presenter: Pedro Ventura de Oliveira
doi://10.26678/ABCM.COBEM2019.COB2019-1528
Abstract
In this paper, an image segmentation method based on multilevel thresholding allowing a more efficient determination of the threshold points is proposed. Moreover, a practical application of this method for crop rows identification based on Progressive Probabilistic Hough Transform is presents, being an important initial step for the rows determination. The proposed thresholding combines a wavelet transform technique, for initially reduce the histogram length of an image converted to grayscale, with a genetic algorithm, responsible for finding the threshold points without an exhaustive search through all the possible values. The points, projected onto the original space of the histogram, are then set as the segmentation values. In addition, when used for crop rows detection, the method provide a refined threshold value, between the ones found in a multilevel segmentation, for the image binarization, which is an initial step in a protocol based on a Hough Transform method. Experiments using different images and thresholding approaches for comparative tests show the efficiency in speed and accuracy in the results for the proposed method. For the application in precision agriculture, the method show a better definition for the initial lines used to define the crop rows.
Keywords
image segmentation, genetic algorithm, thresholding, Precision Agriculture, crop rows detection
