Research And Application Of Image Based On FGBC-iDistance Orthogonal Range Query

Image similarity measurement is a fundamental research problem in the field of computer vision and pattern recognition,which is widely used in image retrieval,image classification and other fields.The paper is based on the project similarity check cloud service platform and the image similarity detection in the National Natural Science Foundation project application as the background to solve the problem of overfitting the VGG model on small data sets and improve the retrieval accuracy as the goal.The paper proposes a Dro-SVGG model based on Dropout and an orthogonal range query retrieval algorithm based on FGBC-iDistance.The main research work and innovation of thesis are as follows:(1)The VGG model based on Dropout.The paper combines Dropout with the SVGG model to construct the Dro-SVGG model,which solves the problem of overfitting the VGG model on small data sets,and improves the training accuracy.The paper replaces the SVGG pooling layer with two convolutional layers to improve the convergence speed of the model and reduce the training time.The theoretical and experimental analysis shows that the Dro-SVGG model improves the accuracy of VGG and improves the convergence speed on small data sets.(2)Orthogonal range query algorithm based on FGBC-iDistance.The traditional feature vector retrieval will seriously degrade performance when searching for high-dimensional vectors.In order to improve the retrieval accuracy,the paper proposes an orthogonal range query method based on FGBC-iDistance,which reduces the intersection of the orthogonal search range and space by dividing the region.The theoretical analysis and experimental results prove the accuracy of FGBC-iDistance orthogonal range query.(3)The image similarity retrieval system.The paper applies the VGG model based on Dropout and the FGBC-iDistance orthogonal range query algorithm to the image similarity detection of the Natural Science Foundation project application,and designs a complete image similarity detection system.The system is stable in function and the results are reliable and true.It can accurately detect similar images in the fund project application,providing a scientific basis for the similarity testing task of the National Natural Science Foundation project application undertaken by the Central South University Key Laboratory.