Viable vestige segregation by support vectors: An implementation for background, shadow and unnecessary fragment amputation of Guava

We propose a simple yet effective pre-classification method for background removal and bruise separation for automatic fruit sorting and grading systems. Different fruits require different pre-classification segmentation approach driven by a unique logic in regard of specific appearance of the concerning fruit. The proposed method is experimented upon guava samples having different tones of bruises, shadow, and highlights. Support Vector Machine (SVM) is used for training regions of respective classes within the samples and Otsu method is used for further efficient processing. We have successfully achieved 94.5% efficiency through our pursued method.