# The Effect Of Image Overlap On Accuracy Of Photogrammetry Using Uavs

## School Name

Spring Valley High School

## Presentation Topic

Math and Computer Science

Mentored

## Mentor

Mentor: Dr. Kenny McLean; CDS Corporation

3rd Place

## Abstract

Image overlap is a variable in photogrammetry. It determines the amount of overlap that two parallel images have when creating a photogrammetric model. There is both longitudinal and latitudinal overlap. The purpose of this experiment was to find a percentage of image overlap that created the most accurate model in photogrammetry. It was hypothesized that an image overlap of 80%-75% would be the most accurate in creating a photogrammetric model, because it would create the most overlap between the images. Nine different ratios of image overlap were tested to see which were the most accurate. A diamond of markers, 200 ft apart from each other was first placed in an open field. They were first input into a flight plan for the Unmanned Aerial Vehicle (UAV) through a computer. Once this was done the drone was launched and the data from the flight was collected and saved. The image pool from the flight, along with the mission area, the lens data, and the CMOS data were then loaded into a project in Pix4d. From these data sets, the software understood the projection. From there, the pictures were exported into Global Mapper where they were measured to see the distance between them and then compared to the known distance of 60.96 m (200 ft). These methods were repeated for all nine overlaps. Four of the tests failed to create an image due to insignificant image overlap. All of the other tests successfully created an image and could be analyzed to see which were the most accurate. The first flight had the smallest average variance between the established and the measured distance, thus making it the most accurate of the flights. A one-way ANOVA test was run on the results to establish if there was a significant difference in between the variances. It was determined that there was not because p = 0.144 > α = 0.05. Thus, the conclusion was made that the overlap did not affect the accuracy, it just determined whether it would have enough information to create an image.

Owens 207

## Start Date

4-16-2016 11:15 AM

COinS

Apr 16th, 11:15 AM

The Effect Of Image Overlap On Accuracy Of Photogrammetry Using Uavs

Owens 207

Image overlap is a variable in photogrammetry. It determines the amount of overlap that two parallel images have when creating a photogrammetric model. There is both longitudinal and latitudinal overlap. The purpose of this experiment was to find a percentage of image overlap that created the most accurate model in photogrammetry. It was hypothesized that an image overlap of 80%-75% would be the most accurate in creating a photogrammetric model, because it would create the most overlap between the images. Nine different ratios of image overlap were tested to see which were the most accurate. A diamond of markers, 200 ft apart from each other was first placed in an open field. They were first input into a flight plan for the Unmanned Aerial Vehicle (UAV) through a computer. Once this was done the drone was launched and the data from the flight was collected and saved. The image pool from the flight, along with the mission area, the lens data, and the CMOS data were then loaded into a project in Pix4d. From these data sets, the software understood the projection. From there, the pictures were exported into Global Mapper where they were measured to see the distance between them and then compared to the known distance of 60.96 m (200 ft). These methods were repeated for all nine overlaps. Four of the tests failed to create an image due to insignificant image overlap. All of the other tests successfully created an image and could be analyzed to see which were the most accurate. The first flight had the smallest average variance between the established and the measured distance, thus making it the most accurate of the flights. A one-way ANOVA test was run on the results to establish if there was a significant difference in between the variances. It was determined that there was not because p = 0.144 > α = 0.05. Thus, the conclusion was made that the overlap did not affect the accuracy, it just determined whether it would have enough information to create an image.