In the wind speed (0~1 m/s) space flow field measurement, the sensor accuracy requirement is high, the realtime online instrument data accuracy is not enough, hysteresis of data acquisition; Consider using multiple sensors measurement improve accuracy, but also has the problem of data fusion. Velocity flow field measurement, the author of this paper, based on kmeansdata collection and pretreatment of RBF neural network soft measurement model, firstly, intermediate variable (current value), using the kmeans clustering, use RBF network training to get a single sensor data; Based on correlationsensor data fusion algorithm of kalman filtering, eliminate invalid data points, and get accurate fusion wind speed prediction. Measurement experiment and data results show that this method processing data results of hysteresis is small, processing speed, high precision of data.