Active NLOS imaging based on deep learning


Non-line-of-sight imaging technology that can image objects outside the field of view has important application prospects in many fields, such as security and automatic driving. Traditional non-line-of-sight imaging studies usually use an imaging model to model the data acquisition process and reconstruct the hidden scene by solving the imaging model. At present, traditional imaging models is not accurate enough so that they can only reconstruct a simple hidden scene. In addition, traditional reconstruction algorithms have a long reconstruction time and are difficult to apply in real time applications.

In order to overcome the shortcomings of traditional imaging and improve the reconstruction speed, this paper proposes a three-dimensional reconstruction algorithm for non-line-of-sight imaging based on deep learning. The reconstruction algorithm firstly obtains 30,000 simulation data as a dataset by simulating existed imaging model. Then, we build, train and test a reconstructing algorithm based on deep learning using the conditional GAN with similar results to traditional reconstruction algorithms. Through contrast experiments, it is concluded that the deep learning-based reconstruction algorithm is more robust to parameter correction errors and can effectively reduce the high dependence on parameter calibration. In addition, based on deep learning, the reconstruction algorithm only needs 20ms to reconstruct a single hidden scene, while the traditional algorithm requires at least 0.2s. Therefore, the reconstruction algorithm based on deep learning can greatly improve the reconstruction speed of non-line-of-sight imaging. Finally, this paper analyzes the shortcomings and improvement schemes of the proposed non-line-of-sight imaging algorithm based on deep learning.

Bachelor Thesis
Click the Cite button above to demo the feature to enable visitors to import publication metadata into their reference management software.
Create your slides in Markdown - click the Slides button to check out the example.

Supplementary notes can be added here, including code, math, and images.