Multi-source Sensor Data Fusion And Its Applications In The Target Detection

Multi-source sensor data fusion is widely used in the field of target detection, it can greately improves the reliability of target detection and the accuracy of the system. Also, it has great impact on the interdisciplinary fields, such as target tracking, target recognition and so on. At present, many researchers from all the world actively study multi-source sensor data fusion, the theory and applications of multi-source data fusion, especially the applications of battlefield environment, disaster area, etc. This paper mainly studies the target detection based on multi-source sensor data fusion. The main work is as follows:Firstly, the models of multi-source sensor data fusion and the target detection based on multi-source sensor data fusion are studied. The advantages and disadvantages of the structure of centralized target detection and distributed target detection are analyzed. This paper mainly analyzes and studies the parallel topology structure of distributed target detection. The classical fusion rules: Chair-Varshney fusion rule and Counting Rule fusion rule are studied and their influenced factors are also analyzed.Secondly, the fusion rule based on failure of sensors is studied. The traditional target detection fusion is deeply analyzed. Most of literatures consider enery consumption, channel noise, sensor coverage, security issues and so on. Few literatures consider the failure of sensors. This paper does some jobs on the failure of sensors:(1) considering the problem of failure of sensors,(2) introducing the Bathtub-Shaped failure rate of the theory of reliability,(3) establishing the probability of failure of sensors,(4) modifying the classical fusion rule of Chair-Varshney,(5) proposing the Extended Log-likelihood Ration Test(ELRT),(6) conducting numerical experiments for ELRT and Chair- Varshney. From the numerical results, we can see that the performances of the system declines with the increasing the number of failed sensors. But the ELRT can outperform the traditional fusion rules when failed sensors are present.Lastly, the big data and macine learing are studied. This paper applies the method of machine learning to the distributed target detection based on multi-source sensor data fusion. The Logistic Regression Fusion Rule(LRFR) and the Logistic Regression Fusion Algorithm(LRFA) are studied, after deeply discussing the essences of Counting Rule fusion rule and K/N fusion rule. Some numerical experiments are conducted in this paper including the performances of LRFR, the system of using LRFR affected by the signal power’s changes, the comparison of the LRFR and the K/N fusion rule, the comparison of the LRFR and the the Counting Rule fusion rule. From these numerical results, we can see that the LRFR has a set of optimal algorithms that can obtain the values of its parameters easily. But, two of the drawbacks of the LRFR are that it depends on the number of samples and the quality of samples.