Identifying object location and orientation are important tasks in target tracking and obstacle avoidance. This paper proposes an algorithm for identifying target location, size, and orientation without any prior knowledge based on machine vision. The algorithm uses target measurements taken from a video camera to update estimates of the target. Two forms of the Kalman filter were compared as methods for estimating the targets' states. The first, the extended Kalman filter (EKF), uses a Taylor series expansion about the current state to estimate the nonlinear measurement update. The second, the square-root unscented Kalman filter (SRUKF), a form of the unscented Kalman filter (UKF), uses perturbed sigma points to estimate the mean and covariance of the measurement update. The algorithm was tested using periodic vehicle motion and random target location, orientation, and area. Both Kalman filter methods converged on the targets' 3-D position, orientation, and area but required an irregular trajectory to accurately estimate the orientation and area. The estimators were simulated using identical parameters to compare them based on equivalent conditions. This produced similar results from the EKF and SRUKF and therefore neither showed a significant improvement over the other.