Mildew is one of the important factors affecting the quality of pipe tobacco. A detection method was developed for identification of moldy pipe tobacco based on electronic nose. Five SnO₂ semiconductor gas sensors were selected to construct a sensor array of the electronic nose. Back propagation neural network (BPNN) was employed as the pattern recognition method. Two feature parameters were extracted from response curves of each sensor, and principal component analysis (PCA) and BPNN were used to process feature data of the whole sensor array. The results of PCA showed the obvious separability of moldy and normal pipe tobacco, but there was some overlap between different levels of moldy tobacco. BPNN were applied for further identification of different moldy levels. The accuracy of recognition rate for moldy pipe tobacco reached 90.00%. The experiments show that the method developed based on electronic nose is capable to distinguish moldy and normal pipe tobacco objectively and effectively which provides a feasible way in control of tobacco quality.