Data Fusion for Identity Estimation and Tracking of Centroidusing Imaging Sensor Data

Abstract

Two aspects involved in automatic target recognition namely, (i) Location and identity estimation (LIE) of a target by fusing infrared (IR) and acoustic sensor data, and (ii) centroid tracking for target state estimation using IR sensor data are discussed in this paper. The LIE has been achieved using a combination of Bayesian fusion and one of the three search algorithms namely, metropolis hastings (MH), simulated annealing (SA) and gradual greedy (GG). It was observed that the performance of the GG search algorithms was better in terms of success rate which has been evaluated through Monte Carlo simulations. For tracking of the centroid, an algorithm, where the centroid of the gray level image is tracked using probabilistic data association filter, has been implemented. Simulated data results indicate good tracking performance of this algorithm. For robust tracking of centroid, the track from the imaging sensor was fused with the track from ground-based radar using state vector fusion. It was observed that fusion generates robust tracks even when there is data loss in one of the sensors.

Fuzzy Logic Augmented Observer and Stability Analysis for Image - Centroid Tracking

“…1) The gray level image is converted into binary image (using lower/upper bounds of the target); the gray image I m (i,j)is converted into binary image with intensity β(i,j) [3]:…”

“…The centroid tracking algorithm (CTA) involves: a) conversion of the data (from the original image) into a binary image by applying upper/lower bounds for the target layers, b) the binary target image is converted into clusters using nearest neighbor (NN) criterion for DA, and then c) the centroid of the clusters is computed and this information is used for tracking the target image. There has been some work in the area of centroid tacking [1][2][3][4][5][6]: a) in ref. [1] by analyzing the molten pool infrared images, the keyhole was extracted by using the fixed threshold method; by using the keyhole images and computing the keyhole centroid, the deviations between the keyhole (centroid) and the welding seam was analyzed; b) the method of ref.…”

“…[2] is highly statistical; c) in ref. [3] a Kalman filter (KF) based centroid tracking algorithm was proposed; d) in ref. [4] the weld pool image centroid algorithm for seam tracking in an arc welding process was proposed; e) in ref.…”

Fuzzy Logic based Square Root Information Filter for Image Centroid Tracking

“…Thus, the a priori information forms a data equation similar to the measurement equation, and it is considered as an additional measurement. The square root information pair and the existing measurements are put in the following form and orthogonal transformation is applied to obtain the Least-square (LS) solution (6) where T is the Householder transformation matrix, e(k) is a sequence of residuals. Since, e(k), the residuals are available from the previous cycle (or the initial conditions); separate the e(k) into e x (k), and e y (k) and obtain the following derivatives: (…”

“…The factors like low signal-to-noise (SNR) ratio in the acquired image/s, low contrast, presence of background clutter and/or false alarms and partial occlusion of the target image necessitate the use of efficient, accurate, and numerically stable filtering algorithms for image-centroid tracking and image fusion [1]. Certain studies on square root type factorization filtering algorithms for state estimation and target tracking have been carried out [6], [7] [11], but in a limited way. In this paper fuzzy logic augmented SRIF is presented for image-centroid tracking and fusion.…”