Research On Image Retrieval Based On Improved Particle Swarm Optimization

With the advancement of computer hardware and software technology,the development of graphics has also entered a new stage,in the face of massive image data,traditional image retrieval methods are subject to accuracy constraints,revealing many drawbacks.Therefore,the relevance feedback(RF)algorithm proposed in text retrieval is integrated into content-based image retrieval(CBIR),to a certain extent,the search accuracy is improved.However,RF has added human interactions and will perform multiple judgments and operations based on people's wishes,resulting in a tedious search process,which is not conducive to improving the efficiency of retrieval.At the same time,the existing RF algorithms still have much room for improvement in feedback accuracy.In order to make breakthroughs in terms of search efficiency and accuracy,relevant scholars,while improving existing RF algorithms,have combined optimization algorithms with CBIR to do research and achieved some results,such as PSO-RF algorithm,GA-RF algorithm,etc,these algorithms are very helpful for improving the retrieval effect.This paper improves CBIR from the following three aspects:(1)Using the scale-free network model as the domain structure of particles,add the items far away from the worst particles in the basic particle swarm formula,and use tabu list to store the local best and the worst particles,get the fusion class scale free tabu particle swarm optimization algorithm(ST-PSO).(2)Aiming at the limitations of the existing relevance feedback algorithms,we introduce the resolution based on the user feedback information to dynamically adjust the feature weights and query vectors,and get a new relevance feedback algorithm.(3)In the process of RF,ST-PSO algorithm is applied to optimize the feedback process of improved image retrieval,which speeds up the retrieval speed and improves the accuracy of retrieval.