Bull Sperm Tracking and Machine Learning-Based Motility Classification

Sperm motility measurement using computer assisted sperm analysis (CASA) has been widely accepted as a substitute for manual but still faces several challenges. In the tracking phase, errors caused by detection failure often occur when measuring fresh bull semen. Tracking two reasons: (1) move very fast, which makes them appear blurry, and (2) partial occlusion, frequently occurs. This study proposes mean angle of motion Tracking-Grid to predict position that failed be detected. The also found useful in fast-moving sperm. proposed methods reduce identity switch (ID-switch) achieve multi-object overall accuracy (MOTAL) 73.2. MOTAL result exhibits 5% less ID-switch is 15.6 points higher than state-of-the-art simple online real-time with deep association metric (Deep SORT). speed achieved 41.18 frames per second (fps), 1.8 times faster Deep SORT. classification, most researchers use one or CASA parameters static threshold value. Such method effective motile-progressive classification reliable identifying non-motile-progressive such vibrating floating machine learning-based classifier support vector three parameters: curvilinear velocity (VCL), straight-line (VSL), linearity (LIN), we call progressive (BSPMC svm3casa ). Experimental results show BSPMC ’s 92.08%, 2.51–9.67 other methods.