Sperm detection and analysis using feature description algorithms

Abstract

Computer aided sperm analysis (CASA) systems have been used in recent years to examine the mobility and morphology of human and animal sperm. While these systems detect sperm, they fail to detect more than one sperm image coinciding or overlapping between the motile spermatozoa. in addition, sensitive results can not be obtained against the light factor of the background in sperm detection. in order to improve the above mentioned problems, using the random forest algorithm, sperm detection was performed on the images obtained from the HOG, LBP and color histogram feature extraction methods. When the experimental results were examined, it was observed that 92% success rate was achieved in the images.