SSD-mobile Net Application In Official Archives Seal Detection

Using deep learning theory,the machine automatically completes the feature extraction of the image in the image library,and realizes the function of archival image seal detection to meet the actual needs in the digital archive utilization system.This method can reduce a lot of repetitive work,reduce the workload of managers,and improve the level of digital file management.Aiming at the problem of file seal detection,the target detection model based on SSD-Mobile Net technology is selected and applied to the official document file utilization system of B / S calculation mode.The main contributions of the research work include the following three aspects.Firstly,according to the situation in the real scene,the 2000 seals required for the simulation experiment were stamped,and the electronic archives were generated by the scanner in the actual use process to scan them into individual archive stamp images.Then for some cases where the seal is not clear due to material reasons,the image is filled with holes after closing operation to improve the quality of the file.Then all the collected images are uniformly named to make them conform to the VOC data set format,in order to provide convenience for model training.Then,replace VGG16 in the SSD-VGG model with Mobile Net to establish the SSDMobile Net model.Because Mobile Net is a lightweight network,the established model has the advantages of fast speed and few parameters.Simulation experiments show that the optimized SSD-Mobile Net model’s seal detection method is effective,achieving 93.2% m AP.Among them,the accuracy rate of the official seal required by colleges and universities reached 99.9%.Combining the two indicators of accuracy and speed,applying this model to the system can meet the actual requirements of the colleges and universities.Finally,in order to integrate with the existing file utilization system,a B/S architecture model consistent with it was selected to build a digital file seal detection system.The system is based on the Flask framework,developed in Python,and uses the SSD-Mobile Net algorithm to implement seal detection.The experimental test proves that the system can accurately detect the target seal,which has strong real-time performance and high accuracy,which lays the foundation for practical application.