Extraction of new features for classification of porcine sperm motility

Abstract

This paper presents an artificial vision system for analysis and classification of cell motility of porcine spermatozoa. In the identification of sperm were used morphology features. In evaluating the motion, the descriptors VCL, VAP, VSL, LIN, WOB, PROG, and BCF, used frequently in the literature, were extracted, also propose the inclusion of a new descriptor (MSM) based on the statistical mode of motion and crosses between VAP and VSL. The classifier used is a multilayer perceptron ANN. The results obtained show that with this method of detection and classification, the movement paths of porcine sperm cells are calculated to an accuracy of up to 98%.