Research On Recognition Method Of Specific Target Image In Complex Background

Dossiers are the truth for cases and also a handler record of the judiciary,which is necessarily to be reviewed by the investigators in the process of handling case.Due to various reasons,the investigators have a heavy workload.Aiming at the problems of seal missing in files of the dossiers,this thesis analyzes the object detection under complex background,which is a important checkpoint in dossiers review.To satisfy the seal check requirement,the F1 should be greater than 80 percent.Contents of the thesis are as follows:(1)Due to the lack of the seal dataset in judicial field,we make a scheme for seal dataset building and enhancement,which is the input for the training and testing of the seal detection model.The seals on the files are usualy unclear,such as blurry,incomplete,overlapping,which are difficult to detect by the model.In order to solve this problem,we propose a seal detection model,which is improved by YOLOv3.More specifically,we prepose a clustering algorithm to get the anchors,which are more suitable for seal object;In order to increase the performance of the model in complex background,we use dilated convolutions in the first three sampling layer and ASFF in the neck network;Finally,we use GIOU loss in the loss function of YOLOv3.The F1 value of the seal detection model increased by 8.43 percent,which reaches the seal detection F1 index.(2)By analyzing the requirements of the seal detection in judicial field,we develop seal detection system,which is base on B/S mode,besides,we use springboot as the backend architecture of the system,which makes the structure of system more clear and easier to maintain.The function of the system are as follows:dossiers creating,dossiers browsing,dossiers searching,files uploading,files classification,files viewing,seal checking,report viewing,report modification,defect location and so on.Seal detection system helps investigators discover the defect of the dossiers,besides,it can also manage the files,which improves the efficiency of dossiers review.