Infrared/Laser Multi-Sensor Fusion And Tracking

In recent years, multi-sensor data fusion technology has gotten an extreme attention driven by its versatility. Compared with single sensor, multi-sensor information fusion has obvious advantages. It effectively improves the data accuracy and enhances the system reliability by integrating the measurements from multiple sensors. At the same time, multi-sensor detection system can improve the ability of observation process through the relative positions or movements among the multiple sensors.As a typical application of multi-sensor data fusion system, infrared detection system and laser radar detection system are cooperated to detect targets. The complementary properties of infrared/laser detection system can be utilized better, which can make the system more intelligent in target detection and recognition. Besides that, it efficiently improves the system's noise-resistant, survivability and working ability.This paper mainly studies the infrared/laser radar data fusion tracking algorithm. First of all, the background of this paper is introduced. The basic principle of multi-sensor fusion and tracking, multi-sensor fusion model, and the common algorithms for information fusion and target tracking filtering algorithm are expounded in detail. Through analyzing the present situations of multi-sensor information fusion and the infrared/laser radar data fusion, the shortages of current research on heterogeneous sensors asynchronous information fusion are pointed out.There are two challenges for infrared detection system and laser radar system on tracking target. The sampling frequency of laser radar is far slower than that of infrared detection system. Besides, in tracking applications, target dynamics are usually modeled in Cartesian coordinate, while the measurements are directly available in the polar coordinate. The traditional solution is to reduce the asynchronous data fusion problem to a synchronous one by data registration. Then the synchronous data is fused by transforming real measurements into pseudo measurements, and through modeling under rectangular coordinate system to get a new measurement equation. So that the converted measurement Kalman filter algorithm can be used to track target. This processing method is simple, but can only solve the situation of time alignment.Aimed at the disadvantages of traditional algorithm, an effective algorithm of data fusion estimation algorithm is presented by combining the theory of multi-scale and converted measurement Kalman filter. The algorithm is suitable for the infrared/laser radar detection system whether they are time alignment or not. Multi-scale model is established according to the different sampling frequency of sensors. And each for a measurement information, a filtering estimation can be achieved. In the case of time alignment, the unscented Kalman filter is used to estimate the target state at fine scale with the angle information of infrared detection system, which makes full use of the frequency characteristic of infrared detection system to effectively improve the estimation precision of target state. At coarse scale, as a result of the laser radar sampling period for infrared detector integer times, the angle information with high precision can be directly fused with the distance information of laser radar, so that the spatial location of target can be determined. And the combined information is thought as the observation of filter at coarse scale. Then the improved converted measurement Kalman filter algorithm can be adopted to estimate the target state. When the sampling time is not aligned, the processing of fine scale is same to the above algorithm, while there is different processing at coarse scale. Because of the laser radar sampling period is no longer the integer times of infrared detector, the observation need to be chosen according to the specific situation at the coarse scale. And due to the infrared/laser radar asynchronous sampling data, the system state transition matrix is constantly changing, no longer a constant matrix.Infrared/laser system is set in tracking a target in 3-D space to verify the algorithm, and the simulation results shows that the algorithm improves the precision of state estimation of infrared/laser effectively.