Explore

UAS Imagery Classification Analysis

Introduction

In this lab, we explored object-based classification using high-resolution RGB imagery collected from Unmanned Aircraft Systems (UAS). Unlike previous exercises that used multispectral imagery, this lab focused on classifying standard RGB data within ArcGIS Pro. The goal was to identify and categorize land cover types, simplify them into broader permeability classes, and analyze pavement conditions by detecting cracks. This hands-on process provided valuable experience in image segmentation, classification, and spatial analysis

Why This Lab is Important to UAS

Object-based classification is a critical tool in the UAS industry because it allows for highly detailed analysis of surface conditions. UAS imagery provides extremely high spatial resolution, making it ideal for applications such as infrastructure monitoring, environmental management, and airport maintenance.

For example, airports must constantly assess runway conditions to ensure safety. Being able to detect pavement cracks and distinguish between permeable and impermeable surfaces helps improve drainage planning and maintenance scheduling. UAS-based analysis enables faster, more cost-effective inspections compared to traditional ground surveys. This technology supports safer operations, reduces costs, and enhances decision-making across aviation and other industries.

Steps and Processes

1. Data Preparation

Copied the project data into a temp folder

Created a new project in ArcGIS Pro

Loaded datasets including Countyparkclipped and Crackdetectionclipped

Resampled imagery to appropriate resolutions (10 cm and 5 cm)

2. Object-Based Classification (Land Cover)

Next, we performed classification on the county park imagery:

Used Segment Mean Shift to divide the image into meaningful objects

Created at least 8 land cover classes

Applied training samples to guide classification

Cleaned results using:

Majority Filter

Boundary Clean tool

3. Area Calculation

After classification:

Added a new field to calculate area

Used a field calculation to determine square meters for each class

Produced a map showing land cover distribution and area coverage

4. Reclassification

The detailed land cover classes were simplified into:

Permeable surfaces

Impermeable surfaces

Area calculations were repeated, and a new map was created to visualize these categories.

5. Crack Detection Classification

A second classification focused on pavement analysis:

Used the crack detection imagery

Created three classes:

Pavement

Cracks

Vegetation

Applied the same cleaning tools

Calculated area coverage for each class

Generated a final map displaying crack distribution and coverage

⁠

Summary and Lessons Learned

This lab demonstrated how object-based classification can extract meaningful information from high-resolution UAS imagery. By segmenting images into objects rather than individual pixels, we achieved more accurate and realistic classifications.

One key takeaway was the importance of parameter selection during segmentation; adjusting spectral and spatial detail significantly impacts results. Additionally, cleaning tools like the Majority Filter and Boundary Clean are essential for improving classification accuracy.

Overall, this lab highlighted the practical value of UAS imagery in real-world applications such as infrastructure monitoring and environmental analysis. The ability to quantify land cover and detect surface damage provides critical insights that support safer and more efficient operations, especially in aviation environments.

Want to print your doc?
This is not the way.

Try clicking the ··· in the right corner or using a keyboard shortcut (

CtrlP

) instead.