Quantifying photogrammetric accuracy for measuring humpback whales using Unmanned Aerial Systems

Abstract

Photogrammetry is the practice of obtaining accurate and valid measurements from 2D images. This practice can be useful in applications where it is dangerous or difficult to reach the target. In recent years, this practice is becoming more common in the marine science field to measure large and potentially dangerous marine mammals. Even more recently, Unmanned Aerial Systems (UAS) technology is being utilized to further minimize the dangers to humans, as well as to decrease the disturbance to animals To establish the accuracy of measurements taken from aerial imagery with UAS technology, this study calculates the distortion values from 3 different cameras, on three different UAS platforms. Lens correction values were calculated for images taken with the three cameras, a GoPro 4 Black, an Olympus E-pm2, and a Sony a5100. These lens correction values were then applied to images taken on the ground of a wooden board approximately 99.9cm long. The static ground images were taken every 10 meters up to 50 meters, to calculate the impact that distance and distortion has on the accuracy of photogrammetric measurements. Finally, each camera was attached to a different UAS platform, GoPro 4 Black with a 3D Robotics Iris+, Olympus E-pm2 with a Microcomputer HexaXL, and the Sony a5100 with a LemHex44. Images were taken at varying altitudes and were then able to be compared to the static ground images to quantify the impact that UAS has on the accuracy. The 3D Robotics Iris+ altitude measurements needed for photogrammetric calculations were derived solely from the onboard barometric sensor, while the MikroKopter and the LemHex44, altitude data were collected by an onboard barometric sensor as well as a Lightware SF11 pulse laser altimeter, thus allowing a comparison of the improved measurements obtained by using a more accurate reading of altitude. These methods were then applied to images of humpback whales (Megaptera novaeangliae) collected in the Antarctic Peninsula in January and February of 2017 with the Sony a5100. A total of 48 individuals were measured for total length, and due to the UAS testing it is known that these measurements are within 1.664 cm of the true length of the whales. Additionally, width measurements of mother calf pairs were compared allowing for an important first step in establishing important time periods of growth and size differences in genders.