Collaborative Sensor Registration without a Priori Association

Abstract

Sensor registration is an important prerequisite for successful multi-sensor data fusion. In this paper, we consider a cooperative sensor registration scenario that the Precise Location Messages (PLM) can be received by the tracker periodically from cooperative targets through wireless data link and utilized to estimate sensor systematic biases. A 2-D sensor registration algorithm is presented to jointly estimate the site location bias and the measurement bias without a priori knowledge of track-to-track association. At first, a point set is obtained by mapping all possible pairs of sensor and PLM measurements. Next, a credit function is defined as the arithmetic mean of the likelihood function of these points. Since the registration biases can be estimated by finding the maximum credit point, a two-step searching algorithm is proposed to jointly estimate location and measurement biases. Its statistical performance is compared to the hybrid Cramér-Rao lower bound (HCRLB).

Relative sensor registration with two‐step method for state estimation

“…For example, the real-time quality control method [4], the least square (LS) method [5], the generalised LS method [6], and the exact maximum likelihood method proposed in [7], all the above methods are considered only measurement bias existed. A two-dimensional (2-D) sensor registration problem with measurement biases and location biases in a cooperative tracking scenario has been investigated in [8], the author defined a credit function as the arithmetic mean of the likelihood function to estimate the bias. An unscented Kalman filter algorithm-based on an augmented state space model, which includes target state as well as space and time misalignment parameters with two different types of 2-D sensors was proposed in [9].…”

Approach of system error registration for two‐station coast radars for sea surface monitoring

“…The corresponding target is defined in this paper as a target that can be effectively detected and steadily tracked by radars located in the common detection area of the two radars. The track information is used as the input of the EML algorithm [4].…”