Análise de técnicas de agrupamentos para classificação de sementes de soja

Abstract

Soy is the main product of Brazilian agriculture and the fourth most cultivated bean in the world. Since the cultivation of soy tends to increase and due to this large market, the guarantee of the product quality is an indispensable factor for enterprises to stay competitive. To acquire information and evaluate the quality of soy planting, industries perform vigor tests. The tetrazolium test, for example, provides information about moisture damage, bedbugs or mechanical damage. However, the verification of the damage reason and its severity are done by an analyst, one by one. Since this is a massive and exhausting work, it is susceptible to mistakes. Proposals involving different supervised learning approaches, including active learning strategies have already been used and brought significant results. Therefore, this paper analyzes the performance of non-supervised techniques for classifying soybeans. An extensive experimental evaluation was realized, considering (9) different clustering algorithms (partitional, hierarchical and density based) applied to (5) image datasets of soybeans seeds submitted to the tetrazolium test, including different damages and/or their levels. To describe those images, (18) extractors of traditional features were considered. (4) metrics (accuracy. FOWLKES, DAVIES, and CALINSKI) and two dimensionality reduction techniques (PCA and TSNE) were considered for validation. Results show that this paper presents important contributions, since it makes it possible to identify descriptors and clustering algorithms that shall be used as pre-processing in other learning processes, accelerating and improving the classification process.