Abstract
3D single-photon LiDAR imaging plays an important role in numerous applications. However, long acquisition times and significant data volumes present a challenge for LiDAR imaging. This paper proposes a task-optimized adaptive sampling framework that enables fast acquisition and processing of high-dimensional single-photon LiDAR data. Given a task of interest, the iterative sampling strategy targets the most informative regions of a scene which are defined as those minimizing parameter uncertainties. The task is performed by considering a Bayesian model that is carefully built to allow fast per-pixel computations while delivering parameter estimates with quantified uncertainties. The framework is demonstrated on multispectral 3D single-photon LiDAR imaging when considering object classification and/or target detection as tasks. It is also analysed for both sequential and parallel scanning modes for different detector array sizes. Results on simulated and real data show the benefit of the proposed optimized sampling strategy when compared to state-of-the-art sampling strategies.
Highlights
Light detection and ranging (LiDAR) used with timecorrelated single-photon detection is receiving significant interest as an emerging approach in numerous applications such as Defence, automotive [1], [2], environmental sciences [3], long-range depth imaging [4]–[9], underwater [10], [11] or through fog [12] depth imaging, and multispectral imaging [13], [14] [15]
Single-photon LiDAR acquire histograms of counts for each pixel location leading to large data volumes as we increase the spatial resolution and/or the spectral resolution
The matched filter in (15) can be computed with O(T logT ) using the fast Fourier transform (FFT) leading to an overall complexity of the integral per-pixel in (16) given by O(KLJT logT ), where K is the number of classes considered, L is the number of wavelengths, J is the computational cost of the evaluated integrand and T is the number of the temporal bins
Summary
Light detection and ranging (LiDAR) used with timecorrelated single-photon detection is receiving significant interest as an emerging approach in numerous applications such as Defence, automotive [1], [2], environmental sciences [3], long-range depth imaging [4]–[9], underwater [10], [11] or through fog [12] depth imaging, and multispectral imaging [13], [14] [15]. Random sampling was considered in sparse-to-depth reconstruction using deep learning approaches Methods such as [28]–[30] used random scanning to obtain dense depth maps using sparse random measurements acquired from a LiDAR sensor and an RGB image from a camera. This paper proposes a new framework for task-optimized adaptive sampling (AS) of the scene to jointly improve both the acquisition and processing of single-photon sensing systems without making use of any other imaging modality. The proposed model accounts for data Poisson statistics and parameter prior information, to build a posterior distribution of the parameters of interest These parameters include spatial labels to locate pixels with or without a reflective surface, the class of each pixel based on a known spectral library, and depth estimates for pixels having an object.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
