Feature extraction and multi-sensor data fusion in monitoring and pre-warning system for security of pipeline based on multi-seismic sensors

In recent years, aiming at the increasingly serious pipeline damages because of the artificial factors, we investigate the characteristic of seismic signals and develop a monitoring and pre-warning system for security of pipelines based on multi-seismic sensors. There are many sensors and processing modules to acquire the seismic signals generated by the ground targets. If these modules all transmit the original data to host machine, it will jam the information channel. So in this system, the seismic signals are processed and extracted the features in respective modules. The non-stationary signal analysis method based on empirical mode decomposition is used to process the seismic signals and acquire some intrinsic mode functions. The target feature vectors were composed of the normalized kurtosis extracted from the decomposition results. The single sensor's judgment was made by the main normalized kurtosis values in the important decomposed frequency bands. Then the features vectors are transmitted to the host machine which will finish the pattern recognition, multi-sensor data fusion and target classification analysis. It will greatly reduce the dimension of feature vectors, the data amount and complexity. This paper will mainly introduce the feature extraction and multi-sensor data fusion. This D-S evidence reasoning was to fuse the recognition results for improving the target recognition accuracy. Then the last judgment was made. The processing methods above were proved effective by the experiment data analysis.