Research And Implementation Of Fine-grained Detection And Management System For IT Equipment Based On CNN

With the higher demand for network transmission and data calculation,the scale and number of IT(Information Technology)room equipment assets have also increased.In the traditional server room management mode,managers need to make an inventory for a large number of devices manually,and to input information on paper.Manual information collection and management bring huge amounts of human consumption,and cannot guarantee the accuracy and speed of equipment information input.With the rapid development of the communication and Internet industries,it’s urgent to informationize and intelligentize the management of IT server room asset.Aiming at the problems of low efficiency,high cost,long time consumption,and low reliability of traditional equipment asset management methods in IT equipment rooms,this thesis designs and implements a IT equipment management system based on convolutional neural network(CNN).The subsystems in the system are highly decoupled.They interact through message middleware and the RESTful interface.It has good operating performance and robustness,and can well meet the demand of managers for the informationization and digitization of equipment asset management.At the same time,CNN-based object detection technology is introduced.And filters are used to extract deep features of the collected images.Finally,the bounding box and predictive labels of the target device are calculated to implement the automatic detection of IT room equipment.Based on the YOLO algorithm model,an improved network model,which is suitable for the equipment room equipment detection scenario,is innovatively proposed.Experiments are designed to verify its feasibility in this application scenario.At the same time,this thesis proposes an innovative design of the dataset automatic building module.This module uses user data to expand the training set,and the algorighm model will be trained on the dataset iteratively.It reduces the model’s overfitting,optimizes the performance and the generalization ability of the object detection algorithm model.The automation of equipment information input during computer room management,which is implemented by object detection technology method,improves the efficiency of equipment room equipment management to a certain extent,and improves the real-time and reliability of equipment information management.First,this thesis introduces the relevant background and technology.Then combined with the use case diagrams,the user’s demand in the application scenario are analyzed.Combined with UML class diagram and flowchart,the detailed design of the subsystem is brought forward.Then each function is tested to confirm the completeness of the system functions.After a series of experiments and tests,this system can meet the demand of IT equipment room equipment information management in a better way.The equipment detection module can clearly locate and identify multiple equipment objects in the image,which plays a good role in improving the efficiency of IT equipment room equipment information management.Finally,the work of this thesis is summarized and the future work is prospected.