Technology of wireless sensor communication has revolutionized our way of living motivated in development and improvement in the field of wireless sensor network. Among many applications target tracking application requires sensor nodes to be localized with minor error. To maintain the target tracking in timely fashion with accurate measurement of target movement is difficult as it summed up with many noises during the detection. Possibilities of node failure often come across in the tracking which results in loss of tracking. Thus to overcome difficulties of tracking new improved tracking framework called as Face (Polygons) Track is proposed. In Face Track sensors are deployed in polygonal fashion. Where polygon share common edge with non overlapping pattern these polygons are the faces and edges being reconstructed to generate each face further so as to avoid loss of tracking. In Face track with Kalman filter, KF optimally selects node as target moves ahead crossing the edges of the face. The Face track network is mapped to develop the courier management system. In this courier management, optimal node selection function selects the number of nodes between source and destination on basis of distance calculated between these node so as to refer nearest node.
Methods like tree and cluster based target tracking require huge interaction among nodes. It is so often that each sensor communicate with next node and then to sink. During this multi hop communication delay is induced in target detection. Unlike clustering, the nodes are organized in polygon without overlapping. Target is tracked using face prediction instead of location prediction in face [7]. Polygonal tracking framework for nodes is organized which detects the movement of target using brink called critical region instead of predicting future movement of the target. Edge detection algorithms bring reliable interconnection among nodes. Optimal node selection algorithm is to select node and to guide target information to sink [8]. III.SYSTEM MODELConsider moving target is to be observed by wireless sensor network. A moving target always transmits signal this transmission from the target are detected by nodes in its sensing range. Using standard estimation theory node sense the presence of target by recording the signal strength and noise energy of the signal as follows: es(t)=Si(t) + Es(t) (1) es(t) is the average signal energy strength in term of received signal Si(t) and noise energy Es(t) at time t. Position of the target is defined by the velocity and the intersection of the x and y point in x-y plane in time t is as follows: Pk = [ Xk Vx(k) Yk Vy(k)]t (2) A.Assumption •Moving target to be observed and detected by WSN is single target. •All sensor nodes are synchronized. •Target detected by nodes in different time span is assumed to be the same target. •Faulty nodes are avoided to keep on tracking B.Face Track with Kalman Filter Localization of polygons, edge detection and optimal node selection algorithm are main factors in the design of the Face track system. Polygons are used to describe the moving path for target. Nodes of the polygon is given as PNwhere N is number of nodes, PN = (v1,v2,...vp). Suppose target is detected by some nodes and suppose that nodes are from polygon P2then this polygon is said to be active polygon. Suppose node v2is active node in P2polygon (Fig. 1). Active node is aware of its position, its adjacent node and adjacent polygons. Edge detection constructs the next polygon by connecting edges of the active polygon. Measurement data obtained from the sensor nodes processed for Kalman filter to arrive at final state of the system. Kalman filter is so powerful that its equations supports to calculate the past, present and future value of the state. Kalman filter is library of set of mathematical equations that provide efficient computational of two steps: first is predicting and second is correction. In prediction step, using time update equations of one step ahead of observation is calculated. In correction step, using measurement update equations correction for estimate of present state is calculated. In time update equation step the error covariance and the present state estimates is used to calculate prior estimate. In measurement update equations step prior estimate is used to calculate posterior estimates. The process is repeated as to calculate new prior state again and again with time and measurement update equations. Fig. 1, polygon network with nodes is typical scenario of Face track framework. Time update equation of KF xˆk ̄ = A. xk-1 (3) Pk ̄ = A. Pk-1 AT + Q (4) A is n×n matrix which gives the state at time from k-1 to current time k. Q is the process noise covariance which might be changed or kept constant. Time update equation (3) calculates the posterior state for projecting state ahead and equation (4) calculates the priori estimate error covariance for the next time. Measurement update equation of KF Kk = Pk ̄. HT (H Pk ̄. HT + R)-1 (5) xˆk = xˆk ̄ + Kk (zk - H xˆk ̄ ) (6) Pk = (I - KkH) Pk ̄ (7) K is the n×m matrix calculates gain which reduces the posteriori error covariance. A H m×n matrix relates to the actual measurement zkand predicted measurement Hxˆk ̄. The posteriori state estimate in equation (6) is updated with zk. Equation (7) updates the error covariance Pk. C.Edge Detection In edge detection algorithm, when target moves from one polygon to other polygon edge is detected between couple of nodes so as to confirm that target has moved from one polygon to other polygon. Detection of the edge allow target to track in timely fashion. While leaving the current polygon target may come across three phases that are: first is to target detected the edge but it will not cross it, second is that, it will cross the edge and third is that, it is about to cross the edge. Considering this three phase target tracking process proceeds further for next polygon
For target tracking application challenge is how to perceive target efficiently in WSN. The idea is to detect the movement of target using Faces. Kalman filters improve the tracking as it predicts the future position for next position and improve the location for the target. Simulation results shows that proposed framework of Face track with KF estimates target's position area, better tracking ability with high accuracy of 82% than Face track. Average error is reduced to great extent considering 50 sensors. Face track framework is developed for courier management application. For this application considering nearest polygon from Face track towards destination from the source courier places are calculated. Calculating shortest place using polygon mapping improves location accuracy with minimum distance for destination from source position.
