Simulation Research on Artificial Intelligence Matching of Image Database Based on Neural Network

The most important extension of the traditional relational database is the processing of image data, especially the query based on image content. In recent years, with the rapid development of machine learning, especially DL (Deep Learning) has achieved great success in text, voice, image and other aspects. In this paper, the AI (artificial intelligence) matching simulation of image database based on neural network is studied. Establish an AI matching algorithm for image database based on CNN (Convolutional Neural Network). Firstly, input the image to be matched, extract its feature points through preprocessing, get feature description according to trained CNN, get the feature point correlation of two images through correlation algorithm, and get the final matching result through the algorithm of removing mismatching points. The results show that the algorithm in this paper can get the most data of matching point pairs when the light intensity changes, with 35 pairs of matching points actually, of which 31 pairs are correctly matched, with the highest accuracy rate of 88.57%. Through experimental analysis, the validity and reliability of the intelligent target matching method in image database are proved.