3D gated Imaging Using a Unet Model

Abstract

Range gated 3D imaging is widely used in space imaging, distance measurement, automatic driving and other application scenarios. A range gated 3D imaging system based on neural network is proposed. The system inputs three range gated images through convolutional neural network, and then outputs a depth map. A single convolutional neural network is proposed to process range gated images, and a multi-scale loss function is used. The test results on the real data set of outdoor driving show that the model trained on the GPU(graphics processing unit) in this paper realizes range gated 3D imaging.