Target Tracking Based On Gray System Theory

Since20th century the research about target tracking technology has been a focus inthe related field, it has very important application in many field. In recent years we have adepth study, but limited to the influence of various interference factors, thus it has a longdistance for its application in real life. As a result,present motion targets tracking techniquescannot satisfy the need in military and civilian field. Therefore, researching on motiontargets tracking techniques still has important significance.The major innovation and research results in this study are as follows:Firstly, according to human motion tracking, this study proposes human motiontracking method which is based on matching normalized auto-correlation and grayprediction. Aiming at the weakness of real-time tracking template correlation matchingalgorithm and the powerlessness in processing obstructions, this study proposes a modifiedGM(1,1) model, obstructions criterion and normalized auto-correlation match. GM(1,1)model Combines initial condition of the gray differential model which is based on thesequencex1of the nth component and the quadratic interpolation method toconstruct background value, and thus improves the prediction accuracy of GM(1,1) model.On the condition without obstructions, gray prediction model narrows the templatematching area and enhances the real-time performance of the algorithm; On the conditionobstructions, this method takes the predicted value with prediction error to instead the realvalue, improves the accuracy of human motion tracking and thus enhance the robustness ofthe algorithm.Secondly, aiming at the human face rapid real-time tracking, this study proposes a newmethod which combines Camshift algorithm and gray exponential law data sequence model.This method firstly on the basis of offline calculation skin probability distribution,combining with the method based on gray model to predict EMG Camshift algorithm oflarge area skin interference in the process of tracking, arm shades the face and face shadesitself when the head rotates. The EGM model prediction information narrows the searchrange of Camshift algorithm. At the same time, the EGM model of metabolism of historicaldata in a timely manner, effectively solve the error accumulation effect between predictiondata and the measured data. Experiments show that this method reduces the number ofiterations Camshift algorithm, improves the operation speed, enhances the real-timeperformance of the system, at the same time the method improves the tracking accuracyunder interference environment, algorithm for good robustness. Aiming to solve the problem brought by the influence of the background for thetemplate and the reduction in matching computation, this study puts forward a kind of basedon dynamic template new algorithm of target tracking and motion prediction. When usingSSDA, this method only employs the template target pixels to participate in the match,reduces the risk of background interference; And constantly revises the template, ensuresthe validity of the template data. Employing the area of actual curve on in the interval asbackground value, and using the minimum square value sum of difference betweensimulation sequences and an accumulation generation operator sequences to optimize thetraditional GM(1,1) model. With GM(1,1) prediction model to get the initial search point,reduces the search area, enhances the real time tracking. When tracking target is sheltered,instead of the real value of forecasting for target location and ensures the continuity androbustness of the tracking of the movement. The experimental results show that the methodis effective.