Research On Multi-sensor Data Fusion Algorithm And Applications

In currently, as multi-sensor information fusion system emerging applied to a variety of complex application background and extensive used, information fusion algorithm research has become a hot research of scientists from various countries.Firstly this paper analyzes the development of multi-sensor data fusion and several multi-sensor information fusion algorithm that used widely, and summarizes the advantages and disadvantages and the use thereof of each of these methods exist, Especially the D-S evidence theory fusion algorithm and BP neural network fusion algorithm are researched and analyzed in detail. According to a large number of materials and books to understand their improved methods. In-depth research is did based on the improved methods of these algorithms and the effect of improved methods, such as D-S evidence theory encountered difficulties in resolving conflicts,and the weak adaptability of BP Fusion network,because of the output is easy to converge to a local minimum, so that we can' t get the global optimum weights. BP neural network has a poor memory about weight of each layer, even has no memory about the weights and the thresholds, so we need to re-train the BP neural network when learning samples increases. Based on the characteristics of these two algorithms, we propose a new improved algorithm, that is based on the accuracy of the results of the integration of two algorithms, the two algorithms are assigned two different weights, and the result obtained by multiplying the weight improves the correct rate of final result of the fusion.In the case of automobile anti-theft alarm system often makes mistake, I present a new vehicle anti-theft alarm system which using multiple sensors around the car to check the environment, then combine the information which collected by sensors. Using the improved algorithms, this paper designed a model based on the ratio of monitoring values and threshold.In order to verify the effectiveness of the new algorithm, the proposed algorithm is applied to the test and calculated using MATLAB software, and the results obtained shows that information fusion used the proposed algorithm makes the car's alarm system has a higher accuracy, and the more sensors, the lower uncertainty, this satisfies the requirements of forecast.