Physical model of snow precipitation interaction with a 3D lidar scanner

Abstract

Snow precipitation interaction with a generic 3D lidar is modeled. The randomness and the intensity of the signal as a function of the visibility and snowflake size and density distribution are reproduced. To do so, a representative snow density distribution is modeled as a function of visibility. Taking into account the laser beam and pulse characteristics, the probability to have one or many snowflakes of a given size in the lidar sampling cell is calculated. Knowing the number and the size of the snowflakes, the magnitude of the lidar signal is calculated. Finally, a filtering algorithm based on the relative intensity of the snowflakes is discussed.

3D ToF LiDAR in Mobile Robotics: A Review

“…For the soft targets, the spatial impulse response function of rain or fog was described by the backscattering coefficient, and the authors gave a rough estimation of the extinction coefficient in case of dry and wet snow. Roy et al [58] modeled the interaction of snowflakes with 3D LiDAR taking into account the physical model of laser power transmission and the statistical model of snowflakes' characteristics. Hahner et al [60] developed a LiDAR noise simulator under fog according to the physical equation of laser power transmission.…”

DSOR: A Scalable Statistical Filter for Removing Falling Snow from LiDAR Point Clouds in Severe Winter Weather

“…Rasmussen et al [13] show that small snow particles act as Lambertian targets and create significant backscattering. To better understand the interaction of snow with automotive 3D LiDAR, Roy et al [14] modeled the interaction between snowflakes and laser pulses. They show that snow detections are concentrated near the sensor and detection statistics are governed by the characteristics of the laser beam and rate of precipitation.…”

“…Noise introduced by snow particles in LiDAR scans can be filtered out in a number of ways; [1] and [14] introduce intensity based filters. These filters assume LiDAR returns from snowflakes will have a lower intensity and remove snow noise by intensity-value thresholding.…”

Statistical models for the lidar technology: false alarms, receiver operating characteristic curves, and Swerling models

“…11,12 Our team has recently issued a paper in which a statistical model is developed for predicting the effect of snow on the lidar signal, model further used to develop a filtering algorithm to remove the clutter due to snow from the 3D image. 13 This filtering algorithm has been experimentally tested since and we will discuss in this paper the fact that it can be seen as an application of the CFAR technique. Radar and lidar technologies will become the rule for comprehensive data fusion from numerous interconnected vehicles in global intelligent transportation systems 14 and the detection theory must be rigorously used in view of not overloading the data processing systems with nuisance alerts.…”