Intelligent Aperture Identification Combining Compressed Data Acquisition with Sparse Filtering-based Deep Learning Towards Natural Gas Pipeline Leak

摘要

Aiming at the problems of natural gas pipeline leak monitoring, it proposed an intelligent pipeline leak aperture identification method combining compressed sensing (CS) and deep learning theory, which can achieved compressed sampling, adaptive feature extraction and recognition. The random Gaussian matrix was applied to realize the compressed data acquisition, and the sparse filtering based on deep learning was applied to achieve the automatic selection of the features. Finally, the high precision recognition of apertures was implemented by softmax regression. Experimentalresults showed that this method achieved the compression of the monitoring data, and the identification performance for data of compressed sensing domain was better than . traditional methods.