Research On Image Retrieval Technology For Public Security Investigation And Application

The application of video image investigation technology for case solving,in the 21 st century,has become an increasingly important routine investigation method of the public security departments.In order to implement the requirements of "rapid reaction,rapid detection and rapid solution" in criminal investigation,retrieving the target suspects rapidly and accurately is one of the key issues that must be solved,which is of great significance to solve the case quickly.With the development of big data,cloud computing and artificial intelligence technology,especially the breakthrough in image retrieval technology based on deep learning algorithm,it can be possible to achieve quick and accurate retrieval of target suspects by using deep learning as a supplement in massive data of surveillance video.Based on the above analysis,research on image retrieval technology for applications of public security investigation is carried out,and the details can be seen as follows:In the first section of the paper,the basic theories of image retrieval are reviewed firstly.Then,the structure of convolutional neural network and the way of parameter propagation are systematically introduced.Finally,the existing image retrieval algorithms for detection applications are presented.And in the second section of the paper,a new method of image retrieval applied in public security investigation is proposed.This method mainly includes two phrases that we called "coarse-level search" and "deep-level search".The "coarse-level search" adopts the existing image retrieval method based on color features,realizes the rough retrieves of target image with database,and the unmatched images can be quickly removed from the image database.During the "deep-level search" phase,we use the results got from the "coarse-level search" phase to retrieve further and get the most similar image to the target image.The third section of the paper is experiment and simulation verification.From the aspects of retrieval time-consuming and accuracy,the image retrieval method proposed in this paper is verified and analyzed in experiment on the public datasets of Market 1501 and Duke-MTMC.The results show that image retrieval method in this paper proposed is superior to the existing methods,and it meet the requirements that improving accuracy and less retrieval time-consuming of public security investigation.Finally,after completing the improvement of theoretical method and the simulation experiment above,the paper uses the development tools and languages such as Flask,SQLite,Java Script,Python and so on,to design and implement an image retrieval prototype platform for the application of public security investigation.By testing the prototype platform in different application environment,the reliability and applicability of the proposed image retrieval algorithm are verified.