There are some limitations in geographical discrimination of wild edible mushroom using single organic or mineral element fingerprint technique. According to the complementarity and synergy of two different fingerprint analysis techniques, the chemical profiles of different parts and sources were fused to explore the feasibility of this protocoland supply a novel reference and basis for tracing the origin of wild edible fungi 124 sporocarps of Boletus edulis collected from seven origins in Yunnan Provinces. The content of 15 mineral elements in the caps and stipes was detected, respectively. In addition, Fourier transform infrared spectroscopy (FTIR) was collected using the powder of fruit body. The original spectra were preprocessed by standard normal variable (SNV), second derivative (2D) algorithm et al. Based on the low and mid-level fusion strategy, the preprocessed spectra and mineral elements of caps and sipes were fused to established support vector machine (SVM) models, including the models of stipe, cap, FTIR, low-level data fusion (stipe+cap, stipe+cap+FTIR) and mid-level data fusion (The cap+stipe +FTIR). The most reliable method that was used to discriminate the B. edulis quickly, was chosen by comparing the model parameters. The results indicated that: (1) the content of Cd, Cr, Cu, Li, Mg, Na, P and Zn elements in caps was higher than the average content of stipe, the average content of Ba, Ca, Co, Ni, Rb, Sr and V elements in the stipe is higher than that in caps. The mineral elements Ca, Cu, Mg, P and Zn), which were essential mineral elements of human, were much higher than the average content of wheat, rice and fresh vegetables, whose content was similar to that in dried animal food. (2) the optimal pretreatment protocol of mineral element dataset was EWMA. The combination of 3D and SNV was the best in FTIR dataset. (3) c value of SVM model of stipe, cap, FITR, low- and mid- level fusion was 8 192, 4 096, 1.414 2, 11.313 7, 1 and 0.707 11, respectively, which indicated that potential over-fitting risk existed in the SVM model using the single mineral element dataset of stipe and cap. (4) the number of samples was misclassified in three models (FTIR, low- and mid- level fusion) was 7, 9, 7 and 0. The accuracy of mid-level fusion model (stipe+cap+FTIR) was the highest. The results illustrated that the mid-level fusion strategy fused the mineral element and FTIR spectra of fruit body was an effective pathway for geographical origins of wild edible mushroom.