Research and improvement of multi-sensor data fusion

Data fusion is the key technology in WSNs. One of the biggest problems in data fusion is the appearance of special data, which is called “noise” and may lead to fusion result deviations. To resolve this question, in this paper, we propose an improved method based on the pignistic probability function. First of all, according to the conversion of the pignistic probability function, we calculate the distance, and then weigh the average fusion evidence. The calculation results show that regardless of the size of the conflict of the evidence, based on the improved method of difBetP, they can quickly and accurately determine the identity of the target under testing. So even if the information given by one or a few sensors are different from the actual existence, it will not affect the fusion result.