To extract the magnetic signal generated by a stress concentration of buried steel pipeline geomagnetic environment exactly, and to overcome the drawback of the existing magnetic testing, which lacks effective detection of high sensitivity multiple probe array and signal processing technology, a kind of method using magnetic gradient tensor with resonance sparse decomposition and bias monostable stochastic resonance (BMSR) to evaluate the pipeline damage identification is put forward. Firstly, the probe arrangement is in the form of a cross tensor array. Secondly, according to the characteristics of the pipeline defect and on-site interference signals, the resonance sparse decomposition of the signal with different quality factors is used to eliminate some interference signals. Finally, the stochastic resonance system with different time domain recovery is added with quantum genetic algorithm for parameter optimization. The identification method is used for the actual station pipeline tensor detection signal, compared with the results using the traditional low-pass filtering, stochastic resonance system combined with different resonance sparse decomposition, the magnetic gradient tensor resonance sparse decomposition and bias monostable processing algorithm is verified to be effective on extracting pipeline damage characterization of magnetic field and pipeline stress concentration.