A calibration-free indoor positioning method based on depth feature mining is proposed to solve the problem of RSS difference and positioning accuracy offset caused by heterogeneous devices in indoor positioning based on WiFi location fingerprint. In the offline stage， the original fingerprint database is processed by the combination of the strongest access point （AP） classification and procrustes analysis， and the standardized sub-fingerprint database is acquired. Stacked denoising autoencoder （SDAE） is used to learn the standardized sub-fingerprint database to obtain the depth feature fingerprint and construct the depth feature sub-fingerprint database. In the online stage， the same fingerprint processing method as the offline stage is used to mine the depth features of RSS data， and the WKNN method is used to match the depth feature sub-fingerprint database to obtain the estimated position. Four different types of mobile phones are used in typical experimental building scenarios. Compared with the traditional standard fingerprint calibration methods， the positioning accuracy of the proposed method is improved by 5.9% and 12.5%， respectively. Experimental results show that the proposed algorithm improves the positioning accuracy and robustness.