Identification of Haploid Maize Seeds using Gray Level Co-occurrence Matrix and Machine Learning Techniques

• Published in: 2018 International Conference on Artificial Intelligence and Data Processing (IDAP) pp. 1 - 5
• Main Authors: ALTUNTAS, Yahya, KOCAMAZ, Adnan Fatih, COMERT, Zafer, CENGIZ, Rahime, ESMERAY, Mesut
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.09.2018
• Subjects: ANN; Artificial neural networks; decision tree; Decision trees; Embryo; Feature extraction; GLCM; haploid identification; Image color analysis; kNN; maize; Mathematical model; texture features; Training
• DOI: 10.1109/IDAP.2018.8620740

Doubled haploid (DH) technique is used effectively in maize breeding. This technique is superior to conventional maize breeding in terms of both time and homozygosity. One of the important processes in DH technique is the selection of haploid seeds. The most common method for selecting haploids is the R1-nj (Navajo) color marker. This color marker appears in the seed endosperm and embryo. Only endosperm color seeds are selected and continued to the germination stage. This selection is usually done manually. The automation of haploid seed selection will increase success and reduce the labor and time. In this study, we used 87 haploid and 326 diploid maize seeds as dataset. Texture features of maize seeds embryos were used. These features were obtained from gray level co-occurrence matrix. The feature vectors are classified using decision trees, k-nearest neighbors and artificial neural networks. The classification performance of machine learning tecniques was tested by using 10-fold cross-validation method. As a result of the test, the best performance was measured in decision tree with the classification success rate as 84.48%.