Seed identification of ten rangeland species based on machine learning using combination of RBF and Feed Forward neural networks

ABSTRACT: Seeds surface characteristics are valuable in assisting the evaluation of generic and taxonomic relationships. Identifying the seeds of rangeland species is important for restoration and improvement of rangelands. Machine vision could be an effective method for rangeland seed identification. This paper presents an automatic system for rangeland seed identification by MATLAB software using neural networks. Results of this study led to five different classifications for seeds based on their morphologic and color characteristics. Different combinations of morphological and color features led to a range of training and test accuracy figures during seed identification. In combinations of just color features, the highest average values of training and test accuracy belonged to five and six color features with 100%. In different combinations of one or two morphological and color features, the highest average values of training and test accuracy were seen in combinations of one or two morphological and five and six color features with 100% accuracy.