The basic principle of multi-sensor information fusion technology is like the human brain process of synthesizing information, various sensors to multi-level and more complementary information and optimize the combination of the space, finally produce the consistency of the explanation to the observation environment. In the process to make full use of multi-source data for reasonable disposal and use, and the ultimate goal of information fusion is based on the sensor for the separation of observation information, based on the information level, the various combination of export more useful information. This is not only take advantage of the multiple sensors cooperating each other, but also the comprehensive processing data from other sources to improve the whole intelligent sensor system. Multiple sensor fusion system has four prominent features:
1, the information redundancy: some characteristics, for the environment by multiple sensors multiple different times (or a single sensor) receive more information, the information is redundant, and have different reliability, through the fusion processing, can extract more accurate and reliable information. In addition, the information redundancy can improve the stability of the system, thus can avoid because of a single sensor failure on the impact of the whole system.
2, information of complementary: different kinds of sensors can be for the system to provide the different nature of the information, the information described in the environmental characteristics of object is different, they are complementary to each other. If you define a composed of all the characteristics of the coordinate space, so the information provided by each sensor will only belong to the whole space a child space, and other sensors to form independent space.
3, the timeliness of information processing, processing of each sensor are independent of each other, the whole process can be used in a parallel processing mechanism of thermal conductivity, so that the system has faster processing speed, provide more timely processing results.
4, the low into nature of information processing, multiple sensors can spend less cost to get the equivalent of single sensor can get the amount of information. On the other hand, if not will be provided by a single sensor information is used to implement other functions, the sum of the cost of the cost of a single sensor and multiple sensors is considerable.
Information fusion in multi-sensor information integrated processing, with the complexity of nature. In the process of information fusion processing, according to the raw data processing method is different, the architecture of information fusion system are mainly three types: centralized, distributed and hybrid.
1, centralized: centralized the raw data of each sensor gain directly sent to the CPU fusion processing, can realize real-time fusion, the data processing of high precision, the algorithm is flexible, the disadvantage is that the processor demand is high, low reliability, large amount of data, it is difficult to achieve.
2, distributed: each sensor to get original data to local treatment, including the pretreatment of the raw data, classification and extract characteristic information, and through their decision criteria decision-making respectively, and then send the results to the fusion center to get the final decision. Distributed to low communication bandwidth requirements, fast computing speed, reliability and good continuity, but there is no centralized tracking precision is high.
3, hybrid: mostly is the above two different combinations, to form a hybrid structure. It retains the advantages of these two kinds of systems, but in communication and computation costs a more expensive. However, such a system has incomparable advantage, the above two kinds of system in the actual situation tend to adopt such a structure.
