|
|
|
 |
Project 0708/14Project Title: Using Unmanned Aerial Vehicles for surveys of marine mammals in Australia: test of concept Chief Investigators: Dr Amanda Hodgson and Dr Michael Noad The outcomes of this project are summarised below in accordance with each objective. Objective 1: Review of UAVs Within the timeframe and budget of this project there has so far been no single UAV available to purchase or hire that could fulfil our requirements for conducting a trial survey of either dugongs or humpback whales. Of the companies available to hire, Insitu Pacific, and Cyber Technology have the most suitable UAV systems. They are both currently focused on designing UAVs for the military, so are relatively costly. However, they are interested in developing the civilian applications for their UAVs. A potential niche market for them is monitoring marine fauna as part of the regulatory and/or environmental impact assessment requirements for the oil and gas industry. Insitu Pacific have approached Woodside Energy and confirmed their interest in UAV technology, particularly for eliminating human risk in aerial surveys. Cyber Technology offered this project some low cost trials. However, these trials were stalled by the permitting requirements of the Civil Aviation Safety Authority (CASA), and the long wait time for CASA to issue their UAV Operators Certificate. Silvertone Electronics has the most promising UAV airframe for purchase, for which researchers would need to source their own payload, autopilot, data link and ground station. There are a number of benefits to hiring a UAV operator: (1) the relatively costly outlay for purchasing, insuring and maintaining a technology is avoided, (2) all risk of system failure or loss is borne by the operator, and (3) hiring multiple operators means you can trial different UAVs. The option of purchasing a UAV would require a large commitment to the development of UAVs by a single research institute, due to the permitting requirements, maintenance costs and the need to retain personnel with the skills to operate the systems. (a) Using small UAVs We used the Warrigul UAV operated by V-TOL Aerospace to conduct scoping flights over both land and water. This UAV was small (1.5 m wingspan) but robust as it was made out of polypropylene materials and could withstand relatively high impacts with minimal damage. Warrigul had limited endurance and control range however, so flights were restricted to within 10 km maximum distance from the base station. When comparing one UAV scoping flight with one manned flight, the UAV maintained the desired altitude and trackline (average 0.04 m and 5 m deviance respectively) better than the manned aircraft (average 4.5 m and 158 m deviance respectively) under the same low-wind conditions. During our over-water scoping flight, the wind speed reached 15 knots and the UAV deviated more heavily from the trackline under these conditions. The Warrigul could transmit images in real-time back to the base station and its flight path could be diverted at any time. However the video images obtained had limited resolution. We were able to depict two dolphins (which were sighted by land-based spotters first) and a manta ray using the real-time footage. The Warrigul gave the advantage of providing records of the field of view and angle of the camera, together with the exact altitude, pitch, roll, heading and GPS track. These records could be used to determine the exact proportion of the survey area sampled more precisely than can be obtained from manned flights, and consequently provided more accurate population estimates. (b) Using manned planes mounted with UAV systems The Australian Research Centre for Aerospace Automation (ARCAA) assisted us in conducting manned flights using a Partenavia mounted with their UAV data acquisition system. Images were captured at 1 frame per second and at a resolution of 1024 × 768 pixels, with the camera angle being changed during flight according to where the animals were located. One flight was conducted over a large dugong herd in shallow water in Moreton Bay. At all altitudes tested (1000, 750 and 550 ft) the dugongs were visible in the images captured. However we felt we could only reliably count the dugongs visible because they were in a large herd and we had prior knowledge that they were dugongs. If surveying animals in deeper water where they might be more obscured by the water, we felt this camera system would not be reliable. We also conducted scoping flights over humpback whales in Moreton Bay and the results were similar to dugongs. In images captured at 1000 ft we could depict whales but couldn’t have identified them to species. At 1500 ft, whales could not be reliably depicted. The combination of the typical UAV imaging system we used and the altitudes we trialled did not provide images of high enough resolution to reliably detect dugongs or whales. Rather than continuing with this system and conducting further trials at lower altitudes, we converted to with higher resolution imaging systems. Objective 3: Comparison between humans and imaging systems We used a manned aircraft to directly compare the sighting rates of dugongs from three observation platforms: (1) four human observers, (2) two high definition video cameras, and (3) a digital still camera capturing 4 megapixel images. A small line-transect survey was conducted at Shark Bay, Western Australia, where there is a high density of dugongs which offers a good opportunity to compare these platforms. The overall sighting rate per platform was analysed within a log-linear model framework. This analysis showed that the still platform’s sighting rate was significantly better than the human observers by 251% at the altitude of 900 ft. However, at 500 ft the performance of the still camera was reduced by 42% to be equivalent to the human observers. The video system performed relatively worse than human observers across both altitudes with a sighting rate of 60% that of human observers. More data would be needed to investigate this result further. Two possible explanations for the different relative performance of stills and observers at the different heights are: (1) the poor sea-state conditions experienced at the low altitude flight may have been better compensated for by the human observers who could spend more time viewing each sighting compared to the single snapshot obtained from the stills, or (2) the observers’ sighting rate may have been poorer at 900 ft than at 500 ft because they had a greater search area to observe in a limited time frame. The poor performance of the video platform was because of the low resolution these images compared to the stills, but may be improved if flying lower and pointing the cameras vertically downwards rather than obliquely. Video should not be discounted as it produces a higher frame rate than the stills providing benefits such as: (1) increasing the probability of capturing animals surfacing, (2) providing some information about the animal movement (e.g. multiple surfacing of dolphins or the white-water produced when dugongs exhale), and (3) increasing the probability of capturing animals outside of the zone of glare within the images. In conclusion, it is apparent, just by the UAV developments that have occurred in Australia during the course of this project, that the capabilities of UAVs will continue to improve. There are companies, such as CyberTech and Insitu Pacific who have UAVs capable of the range and endurance needed to conduct full marine mammal survey trials. The next step forward for the development of this technique for monitoring marine mammal populations would be to hire one of these companies and trial a range of UAV payload systems to determine the most efficient imaging system for each type of marine mammal surveyed in Australia. As there are currently fewer limitations in Australia than in the US for flying UAVs in civilian airspace, we have strong potential to research and develop UAVs for aerial surveys in Australia. |
 
|