About the Project

Project Description

Forest ecosystem health is at the center of many large-scale environmental problems that we face today. Efforts to prevent wildfires and handle their fallout, for example, or to ensure that forests maintain their important role as carbon sinks while climate change progresses, will benefit from an increased understanding of these crucial environments. Mobile lidar scanning, a technique that relies on lidar, a remote sensing technology, is used to gather large amounts of three-dimensional data from forests. These data can in theory be used to create visual models or answer quantitative questions about forest makeup. In practice, it is difficult for the non-specialist to make much sense of the vast and complicated data that result. Tools exist to help with stages of the pipeline, but many require technical knowledge or were developed for airborne lidar scanning, a separate technology.

To bridge this gap, we plan to develop a start-to-end solution that is accessible by the non-specialist, and delivers results in the browser. The solution will accept mobile lidar data as input, and generate various useful statistics and visualizations which can be filtered and manipulated within a web application. This will improve upon the currently used methods, which are somewhat strung together and lack a cohesive pipeline.

---

High Level Requirements

Requirements have been gathered through multiple meetings with our client with discussions on what already exists, and what should exist. The Requirements Specification Document is linked on the Deliverables page. These requirements are split into three sections:
Functional

• Shiny UI
• Data Upload
• Data Cleaning and Normalizing
• Individual Tree Segmentation
• Point Classification
• Key Characteristics in a Point Cloud
• Change Detection in a Point Cloud
• Registration of Two Overlapping Point Clouds

Non-Functional

• Quick and Easy to Learn UI
• 80% Margin of Error for Calculations

Environmental

• R Programming Language
• GeoSLAM

---

Solution Vision

The envisioned solution of this project will serve as a resource for researchers and ecologists to analyze their own MLS data. As part of the final product, the team plans to create an easy-to-use application with the following features:

• Data Upload
• Users will be able to upload only .laz and .las files.
• Users will be able to upload as many files as they would like, under 10GB.
• Data Cleaning and Denoising
• Once data is uploaded,it will be cleaned and normalized using a statistical outlier removal algorithm (SOR).
• Individual Tree Segmentation
• After the uploaded data has been automatically cleaned, the user can run a tree segmentation algorithm to separate trees based on their bole position.
• Point Classification
• Individual points in the point cloud will be classified based on their position on a tree such as bole, branch, or crown.
• Derived Data Products
• Users will then be able to select specific trees and obtain relevant measurements such as tree height, tree diameter, and lowest branch height.
• Users will also be able to see how two temporally distinct but overlapping point clouds match up
• Data Summaries/Visualization
• The GUI will also display statistics in both a 3D interactive point cloud, as well as on a table.

The final product will be a single portable tool that any ecologist can access and use. With a straightforward GUI, the team plans for users to be more focused on their research than on learning a new piece of software. The project will focus on creating a MLS specific application to support the growing number of researchers utilizing MLS. The features explained above will be wrapped into a Shiny R application that will be portable and accessible from the web.

Below is a flowchart that details the structure of the app:

---

Technologies

The technologies used in the software development for our project are:

GitHub: The ability to work collaboratively on code allows all team members to develop portions of the code that can be integrated without issues.

R: With R's ability to effectively manage data by reshaping, cleaning and analyzing it is a key technology to be used in the management of the point cloud data generated from mobile lidar systems.

C++: Using multithreaded operations data can be efficiently loaded and used, which will improve the performance of our project. C++ also offer's tools that can provide 3d visualizations of the mobile lidar point cloud data.

---

Schedule