Feature Extraction for the Identification of Two-Class Mechanical Stability Test of Natural Rubber Latex

摘要

Rubber latex concentrate is a popular raw material widely used for making many common household and industrial products. As its quality is not consistent due to either, the source, weather, storage time, etc. there is a need to be able to measure its quality. A common measure of its quality is the mechanical stability, which is defined as the time at the first onset of flocculation when the latex is subjected to physical stress. Currently, the assessment is performed manually by trained personnel, closely adhering to the specifications defined by the ISO 35 standard mechanical stability test that is widely adopted by the rubber industry. Nevertheless, there is some level of subjectivity involved as the test heavily depends on the human eyesight as well as the technician's experience. In this paper, we proposed a new feature set for a computer vision-based mechanical stability classification system that is based on the current standard test. We investigated this with several features as well as a new feature set that is based on the particle size. These were classified with a feedforward neural network. Experimental results demonstrated that the proposed system was able to provide good classification accuracies for this two-class MST problem.