The inventory taking process in the archives is complex and error-prone, and the traditional manual method is inefficient. To this end, this study proposes a real-time inventory count model for archives that integrates QR code recognition and fingerprint encryption algorithms. This model adopts the Adaptive Median Filtering Algorithm (AMFA) to suppress salt-andpepper noise, and combines the improved Otsu threshold segmentation and Hough transform to correct image distortion. The fingerprint feature encryption algorithm is introduced. Dynamic keys are generated through the translation, rotation, and scaling transformations of biometric feature points, forming double-layer encryption with the improved RC4 algorithm to achieve secure storage and access control of data. Experiments show that in the multi-source archive image test, the average MSE value of the model is 28.56 and the PSNR value is 32.45, which improves the noise suppression ability by 15.3% compared with the traditional median filtering algorithm. Facing complex scenarios, the decoding rates reached 87.33% and 92.11%, respectively, and the positioning accuracy rate reached 97.38%. Under the FGSM attack, the decoding rate of the model remained at 100%, and the EER value was only 0.92%. The above results prove that the proposed model has better encryption performance. This model effectively enhances the overall efficiency and intelligence level of archive management, providing technical support for promoting the automation of archive management.