Artificial intelligence (AI) processing for enhancing an intelligent sensor - I

Abstract

This report has summarized the work done during the final year project. This project focuses on the reconstruction of photoacoustic images using deep learning algorithms. First, we learned the physical principles of photoacoustic imaging and the generation process of photoacoustic signals. In addition, we introduced the knowledge of deep learning, especially convolutional neural networks. After that, compared with different neural networks, such as U Net, Res U-Net, FD U-Net, we chose to make some improvements on the basis of the classic FD U-Net, and constructed the network used in this project. Through preprocessing and data augmentation of the original dataset, we obtained training set, validation set and test set. Then, we trained the network, analyzed the performance of the network, and finally observe the reconstruction results of the network on the downsampled images.