Accurate analysis of high-resolution transmission electron microscopy (HRTEM) images is important for the characterization and design of materials. However, conventional analyses rely mostly on manual procedures, which are time-consuming and lack accuracy, especially when the image contrast is low. Here, we propose an advanced analysis method for extracting crystal features from HRTEM images based on a 2D fast Fourier transform and U-Net based deep learning model. By using HRTEM images of iron oxide nanoparticles as examples, we show that our method is capable of providing information on the crystallinity profile, distribution of crystal planes, phases and defects automatically with high accuracy. In an era of data-driven technological development, we believe that deep learning based analysis tools will facilitate great progress in fundamental research on crystalline materials.