Abstract: Aiming at  the diversity of multisensor data, we proposed an improved data fusion algorithm. Firstly, the wavelet technology was used to eliminate the Gaussian white noise of collected data and compress the data. Secondly, the processed data was stratified, and the coefficients were filtered by Kalman. Meanwhile,  the data was reconstructed by the Mallat fast reconstruction algorithm. Finally, the signaltonoise ratio (SNR) was calculated by maximum and minimum degree of close to the sensor data, and data fusion was carried out by SNR. Experimental results based on multisensor actual data show that the data fusion algorithm is superior to simple weighted data fusion, wavelet data fusion and Kalman filtering fusion algorithm in stability.

Key words: multisensor data, data fusion, signal-to-noise ratio