Improving three-dimensional (3D) range gated reconstruction through time-of-flight (TOF) imaging analysis

Abstract

This paper performs an experimental investigation on the TOF imaging profile which strongly influences the quality of reconstruction to accomplish accurate range sensing. From our analysis, the reflected intensity profile recorded appears to deviate from Gaussian model which is commonly assumed and can be perceived as a mixture of noises and actual reflected signal. Noise-weighted Average range calculation is therefore proposed to alleviate noise influence based on the signal detection threshold and system noises. From our experimental result, this alternative range solution demonstrates better accuracy as compared to the conventional weighted average method and proven as a para-axial correction to improve range reconstruction in 3D gated imaging system.

Theoretical and experimental investigation into the influence factors for range gated reconstruction

“…Gaussian form is commonly assumed for temporal function of the transmitted laser pulse P t (t) where P o represents the transmitted power and  p denotes the standard deviation of laser pulse [9,13]:…”

“…These parameters could change the characteristics and cause variations in the reflected laser pulse which directly impact the accuracy of range determination. The importance of laser intensity profile [11,12], distance interference [10], sensor [13,14], and scattering effects [16] were discussed in various literatures.…”

Time-of-Flight Imaging at 10 ps Resolution with an ICCD Camera

“…The main advantages of ICCD cameras are their wide commercial availability (compared, e.g., to SPAD cameras), low cost (compared to Streak cameras) and array format with high pixel counts (compared to both SPAD cameras and streak cameras). ICCDs are used for time-of-flight measurements, where their gating and low light detection capabilities are used for ghost imaging [11,12], underwater and 3D imaging [13,14], vehicle mounted night vision systems [15], bio-medical applications [16,17,18] and for single photon time of flight measurements [19,20] and generic LIDAR measurements [21].…”

Learning Super-resolved Depth from Active Gated Imaging

“…Andersson presented in [9] two other depth reconstruction methods based on a sampled GDP: leastsquares parameter fitting and data feature positioning. In recent years, Chua et al developed a noise-weighted average range calculation in order to handle noise influence better [10]. However, sampling the GDP requires recording many images of the same scene.…”