Study On Some Issues Of Target Tracking And Multi-sensor Information Fusion

Target tracking and multi-sensor information fusion, is becoming one of the most advanced and active research topics for the requirement of model military affairs and civil development. Control theory, signal processing, artificial intelligent and statistics etc. are the main theory basement of this technology. Tracking and fusion have been studied extensively in past decades, and a mass of theories and methods have been achieved. As yet, there are many problems in this area, and especially with tracking and fusion technology used in various industries and the theory basement developing, the object becoming more and more complicated, a lot of new challenges lies in the research on tracking and fusion.This thesis mainly focuses on issues of the multi-sensors registration, high speed and high precision target tracking, infrared small weak targets detection, infrared/radar maneuvering target tracking via sequence filtering, multi-sensor multi-character information fusion. The main contents of the dissertation are outlined as following:For the issue of the multi-sensors registration, a new multi-sensor registration algorithm via fuzzy c-mean cluster is presented after analyzing some previous multi-sensor data registrations. The calculation on the registration error of the low cost multi-sensor is implemented.For the issue of the high speed and high precision target tracking, a new Fuzzy-Kalman Filter is presented: a Fuzzy bend surface that presents relationship from measurement error mean and covariance to system noise and measurement noise is put forward to overleap the reasoning of Fuzzy rule, and get the relation between the filter sampling time and target error fuzz, so the tracking and estimation of target is improved.For the issue of the infrared small weak targets detection, after analyzing the characteristics and shortage of some old algorithm about the infrared small weak targets detection, aim at the single frame image, we first time put forward a kind of new detection algorithm— small weak targets detection via HOSA.For the issue of the multi-sensor maneuvering target tracking, we propose that the multi-sensor information should be space-time fused firstly, then the fused data could be sent to the particle filter. The traditional particle filter isalso be improved by us.For the issue of the multi-sensor multi-character information fusion, based on high order spectral reduction for rulers, a new structure changed Fuzzy Neural Network (scFNN) is presented. The resource of multi-sensor/multi-character is optimized by optimizing the structure of FNN.