Date of Award

Spring 2023

Project Type


Program or Major

Mechanical Engineering

Degree Name

Master of Science

First Advisor

May-Win Thein

Second Advisor

Bingxian Mu


Light Detection And Ranging (LiDAR) sensors are an effective way to unobtrusively measure snow depth using Uncrewed Aerial Vehicles (UAVs). LiDAR snow depth estimation performs poorly when scanning obstructed areas due to a reduction in LiDAR ground returns. When using Uncrewed Aerial Systems (UAS) to measure snow depth in forests, trees obstruct LiDAR laser pulses from penetrating to the forest floor, thus, reducing the number of returns. Decreasing UAV flight speed increases the total number of returns which, in turn, can improve snow depth estimates but reduces the sampling extent. In this work, it is shown that Machine Vision (MV) can be used to analyze underlying terrain and determine appropriate corresponding UAV speeds that result in a desired average number of LiDAR returns. Otsu's Automatic Thresholding method is used to differentiate between coniferous and deciduous trees which represent the two types of obstructions in the test site used for this work. The average expected number of ground returns is determined using LiDAR data and composite images of the Thompson Farm test site at the University of New Hampshire. Using the MV method, maximum speed setpoints are calculated for an appropriate pre-determined flight path trajectory. This flight path maximizes mission efficiency (i.e., decreases flight time and power consumption), while still collecting the minimum average number of LiDAR ground returns. Taking into account UAV dynamics, the setpoints are adjusted to ensure the UAV never exceeds these speeds. Simulation results show that the flight path only changes the number of LiDAR returns over obstructed areas where the returns are increased to desired levels. The minimum number of returns for the simulated flight is 300 ground returns per square meter. This number is comparable to flying at a constant speed of 2 m/s, which would require 14 minutes and 40 seconds to complete the remote sensing mission. The simulated flight only requires 6 minutes and 30 seconds, collecting the same minimum number of LiDAR returns in 44.3\% of the time as would normally be required when flying at a constant 2 m/s speed.