Classification of Soybean Seed Using Support Vector Machine with Image Enhancement Techniques

Soybean is rich in high-quality protein and serves as a vital dietary source for humans and livestock, offering essential amino acids that are crucial for nutrition. Hence, classification of soybean seed holds paramount importance in modern agriculture and the food industry. With the ever-growing demand for soy-based products, accurate classification ensures optimal crop management. In this paper, we present a groundbreaking approach for classifying soybean seed image datasets using a support vector machine (SVM) integrated with innovative image enhancement techniques. The research begins with a detailed exploration of data collection methodologies, which involves the meticulous categorization of soybean seeds into two distinct types and further classification into ten specific classes. Preprocessing techniques are applied to the dataset to remove the noise and enhance the quality of an image. The study meticulously delves into feature extraction processes, emphasizing the extraction of pertinent features crucial for accurate classification. Lastly, the SVM classifier is used to determine the class of the soybean image. Following a thorough experiment, the soybean dataset demonstrated impressive results, achieving a classification accuracy of 95.6% for good seeds and 96.1% accuracy for bad seeds, with a ratio of 85:15. Notably, the integration of preprocessing techniques led to remarkable improvements. Specifically, a notable 17% increase in classification accuracy for good quality seed and a significant 16% gain for bad seed was achieved when compared to the soybean dataset without preprocessing. These findings underscore the pivotal role of preprocessing methods in refining the dataset, enabling the classification model to discern subtle patterns and nuances within the data, ultimately enhancing the accuracy of seed classification processes.