Research And Implementation On The Anti-occlusion Algorithm For Sperm Multi-target Tracking

Computer assisted sperm analysis (CASA) system is an innovative product combined ofdigital image processing technology and medical science, overcoming the subjectivity,inefficiency and errors of the traditional manual analysis. With the advantages of automation,high efficiency and accuracy, it has been widely used in medical field. The core of CASA isthe sperm multi-target tracking algorithm, which has achieved analysis functions of imagesegmentation, target recognition, motion prediction, trajectory tracking and parametercalculation automatically. However, when sperm targets are occluded by white blood cells orother sperm targets, the original algorithm will produce the problems such as failure detectionand tracking error, influencing the robustness and accuracy of the system seriously. The topicbackground of this article is the multi-target tracking algorithm of CASA, needed to solve theproblems under the condition of occlusion.Combined with sperms motion model’s characteristics of large quantity, small size, lowdefinition and weak moving regularity, this paper presents an improved anti-occlusionalgorithm. Based on targets’ size and morphological characteristics, the improved algorithmdetermines the type of occlusion and occurring period. Sperms occluded by white cells will betracked with virtual positioning on Kalman filter, else occluded by each other will be trackedwith retentive information. It formulates corresponding strategy of template updating andcompensatory correction. The algorithm also can timely and accurately detect the end ofocclusion, automatically return to normal process of target tracking without occlusion.This paper introduces the processing procedure of anti-occlusion algorithm of spermmulti-target tracking in detail and carries on the simulation experiment. Experimental resultsshow that, in CASA the improved anti-occlusion algorithm can detect sperm targets, detectthe type of occlusion and start time, track the target motion trajectory of sperm effectively,and match back the target at the end of occlusion, increasing the accuracy and robustness ofthe sperm multi-target tracking algorithm This study provides a solid theoretical foundationand practical data for the application in the project.