VBM granulometry for real applications in the fertilizer industry

• Nelson de F. Traversi FURG
• Douglas A. Goulart FURG
• Julio Cezar O. Mendonça FURG
• Silvia S. C. Botelho FURG
• Luciane B. Soares FURG
• Emanuel S. D. Estrada FURG
• Paulo L. J. Drews FURG
• Vinícius M. Oliveira FURG

Abstract

The particle size analysis is a laboratory process that takes a considerable time to be carried out. This control is mainly made using mechanical sieving to regulate the production. Searching for a faster method, this paper explore the alternatives for the mechanical sieving with digital images, using a top view of the conveyor belt. This paper explore the viability of conventional segmentation methods, like K-means, Watershed, Canny and OTSU to segment clear fully covered images of grains, and the impact of the possible real life noises like dust, motion and lighting conditions. All noises applied in varying degrees of influence to determine the impact each noise have. Due to the difficulties of obtaining properly annotated data from the factory floor a simulator developed and used to generate images with noises simulating the factory floor.

Keywords: Industries, Image segmentation, Service robots, Motion segmentation, Digital images, Education, Lighting