رديابي اسپرم با استفاده ار الگوريتم فيلترهاي ذره اي

چكيده انگليسي :

Sperm tracking using particle filter algorithm Jaber roohi Jaber roohi@ec iut ac ir Jan 15 2010 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language FarsiM Ahmad zadeh Prof Supervisor Ahmadzadeh@cc iut ac irAbstractUsing and advancement of machine vision techniques in industrial military and medical applications wereoutstanding in last two decades It has been employed for determining some diseases by non invasive methods It is employed in order to determinate infertility in males Different Diseases in sperms leads to deforming oftheir shapes Deforming in shape of head tail neck of sperms causes different motion of sperms In sequence extracting sperms motion feature is used for classifying sperms disease Tracking of human sperm cells is achallenging task in computer vision due to the motion uncertainty In this work extracting trajectory of spermsfor determining sperms disease has been presented Particle Filter PF and Kalman Filters are two of the most reputable algorithms that used in object tracking Inorder to finding optimum response of Kalman filter constraints such as linear state model and Gaussian noiseshould be satisfied Also probability density function of state model variables should be Gaussian function However in most systems these conditions are not met and other solutions such as Monte Carlo method areused Monte Carlo has been proposed to solve problems associated with nonlinear non Gaussian probabilitydistribution and non Gaussian noise In this work a new method of multiple sperm tracking based on Particle filtering and C means clusteringalgorithms is considered which reduces sperms clutter effects in tracking This algorithm is functioning in twostep enhancement after removing the background and small areas successfully Additionally sperms aredetected by comparing their color histogram using Bhattacharya distance method with the same features of areference specimen Finally by executing a simple threshold sperms have been detected In tracking step first single sperm tracking is implemented Estimated trajectory of kalman is compared withthat of PF Then multiple sperm tracking was evaluated In multiple object tracking and in the case whichobjects are similar problems such as clutter and occlusion are occurring However in multiple sperm tracking occlusion would not be a problem Two methods were considered in this section which one is by not assumingclutter effect and another one is by assuming this effect In First case proposed algorithm works well onlywhen sperms are far apart And as sperms become close together this algorithm fails In order to prevent thisfailure a second method for reducing clutter effects has been presented that would track sperms acceptablyeven if they become close together This algorithm is based on clustering of particles of PF algorithm usingfuzzy C Means algorithm PF algorithm generates some particle for each sperm Observations from particles ofsperms that become close together would be affected by other sperms In this case observation of particles thatplaced between two or more sperms have less reliable Fuzzy C means algorithm for weighting the particles ofsperms has been used Therefore particles that are close to other sperms get smaller weight and vice versa Bydirectly applying this weight in observation equation we can reduce clutter effects Simulation results showedthat proposed method works better rather than the conventional method Key WordsComputer vision fertility Particle Filter nonlinear model multiple object tracking clutter C means

روحي، جابر

رديابي اسپرم با استفاده ار الگوريتم فيلترهاي ذره اي

پردازش ويدئو , هيستوگرام رنگي , فيلتر كالمن