Research On Sperm Activity Detection Technology Based On Computer Vision

Because the current routine semen analysis is based on observation and judgment under the microscope by experienced medical personnel,the accuracy of the analysis is low,and at the same time it lacks rigorous and objective data support.Computeraided medical decision-making has always been a hot topic in the field of image processing.The use of computer image processing and machine vision to detect and evaluate sperm motility and its kinematic parameters efficiently and accurately has high clinical and scientific value.The core of vision-based sperm activity detection technology is how to accurately extract sperm targets and track them with the help of image processing technology and target detection and tracking technology.A report of sperm activity in the sample is finally given.This paper studies and discusses object detection and tracking technology of sperm targets in images,and designs a sperm activity detection system for experimental verification.The main work is as follows:(1)Research on the detection of sperm multi-target and the extraction algorithms of moving sperm target.Aiming at the problem that the existing frame difference method is easy to produce holes and ghosting when calculating the difference,this article improves the traditional frame difference method.Using the extracted current frame background image as the intermediate frame to make a difference from the current frame and the previous frame respectively.It can eliminate the situation that the sperm target image is missed or mistakenly detected due to the background and foreground pixels being similar after graying.At the same time,in view of the great correlation between the three channels of R,G,and B in RGB space,which is not conducive to the detection and segmentation of moving sperm objects.In this paper,a background model is established in HSV color space,and then a low-rank matrix decomposition method is used to extract a more accurate background image.(2)Research on multi-sperm-objects tracking algorithm.Based on the common method of multi-target tracking processing in sequence images,Kalman filter is added to the common KCF tracking algorithm to predict and estimate the position of moving target.At the same time,adaptive scale modification method is added to solve the problem of tracking loss caused by the invariable tracking window of traditional KCF algorithm.During this period,the sperm target may be blocked by impurities such as white blood cells,which may cause tracking failure.Therefore,this paper adds the target blocking judgment condition to the algorithm.Finally,the matching and correlation method based on IOU is used to match the detection target with the corresponding tracking target,complete the data correlation,and then obtain the sperm target motion trajectory.Experimental results show that the algorithm can achieve accurate tracking when the target is occluded or deformed.(3)To construct a sperm activity detection system.By using Open CV dynamic link library,we can accurately detect sperm target in semen and distinguish impurity target,recognize and distinguish the number of living sperm and dead sperm,analyze sperm trajectories and browse dynamically,and finally get the corresponding dynamic parameters and classify them.The operation of the system is relatively simple and the sample experiment shows that the system can track the sperm stably.The algorithm has small calculation and strong robustness.The detection rate of moving targets is over 94% and the tracking success rate is high.