The Technology Research Of Multi-sensor Data Association And Track Fusion

With the development of the electronic technology, the information fusion technologyhas been used in many fields. Because the single sensor system has more disadvantages in theaspect of reliability, multi-sensor system plays a very important role in recent years step bystep. The essence of multi-sensor information fusion is to create the more reliable data andmore correct information than the single sensor system. So we must combine all theinformation from all the sensors by some certain methods. Meanwhile, the technology ofmulti-sensor information system can enlarge the system rate of coverage in the time andspace domain. Besides, it can improve the utilization rate of the information.This paper does the research work on the data preprocess technology、data associationand information fusion method which are the three important parts of the information fusion.And the paper puts forwards some new methods on the base of the traditional methods.Through the simulation, the new methods are proven to be effective.The data preprocess technology is the preparing stage of the information fusion and isthe key to the information fusion technology. It includes the problem of data registration andoutliers inspection. And the two problems is the key link of the data preprocess technology.The data registration can be divided into time registration and space registration. To solve theproblem of time synchronization and the coordinate unity of the multi-sensor data, this paperdoes detailed research on the issue. And the paper puts forwards the online data inspectionmethod on the base of the traditional method through analyzing the outlier inspection. Thesystem can utilize the multi-sensor data more effectively through judging and delete theoutlier.For the data association problem, this paper analyzes the nearest method、probabilitydata association algorithm and joint probability data association algorithm. Then the papertests and verifies the methods effective through the simulation to single target and multi target.For the data classification problem, the paper puts forwards using the statistical screening andthe principle component analysis to classify the data from multi-sensor. And it is proven to beeffective through the simulation.For the information fusion methods, this paper mainly discusses them from the measurement fusion and track fusion. About measurement fusion, the paper studies threebasic algorithm: enlarge dimension filter、false sequential filter and composite measure filter.At the end of this part, it discusses the partial conversion of kalman filtering and verifies thisalgorithm effective through the simulation from the target speed、location and other aspects.About track information, this paper puts forwards the adaptive digital filter algorithm on thebase of studying the simple convex combination track fusion and weighted fusion algorithm.And this algorithm studies the correlation of the data and proceeds the fusion from thespecific data from the different sensors. At the end of the chapter, this algorithm is proveneffective through the simulation.According to the uncertainty of the evidence in the information fusion, the paper studiesthe D-S evidence theory in the application of information fusion. The basic theory andalgorithm process of the D-S evidence theory are studied in the paper also. And the paper putsforwards an improved D-S evidence theory algorithm according to the high conflict evidencecondition. Through simulation, the new algorithm is proven to be effective in this conditionand can solve the contradiction between the conflict evidence.