Benefits Of Koala Detection With Aerial Surveillance

The use of a drone also known as Remotely Piloted Aircraft Systems (RPAS) is being used more frequently to collect image data of wildlife populations. Ariel surveys are manually interpreted however this can often be bias, tedious and time-consuming which causes low probabilities of detection and higher rates of false negatives (Evangeline Corcoran et al. 2019). This study was set up to assess the accuracy of a new automatic detection method for koalas in RPAS-derived thermal imaging using detection algorithms in a real field setting with a variety of vegetation. Using an automated and manual detection system of thermal imaging on the koalas in Petrie Queensland was then compared and matched with other surveys done using locations of other large groups of koalas fitted with radio collars to verify this new method of detection.

The finding of this study confirms that using automated detection was more accurate than manual examination when detecting the koalas. It is also said that there is the potential to reduce detection errors, overcome problems of site inaccessibility, and gain more detailed information about population size and trends of surveyed species (Evangeline Corcoran et al. 2019). The thermal imagery allows the computer vision to easily detect and count their thermal signatures between mammals and their background environment. Using the automated method had a higher likelihood of detection and this was (85%) whereas the manual method was (52%), with a range of 68–100% detection compared to 40–55% detection for manual (Table 2). It is stated that the manual assessment of RPAS imaging is more likely to give a false negative error as opposed to the semi-automated and automated detection methods, which cause a great difference between predicted and actual population size (Evangeline Corcoran et al. 2019)

Koala populations are widely spread and found on private property, making it time consuming and tough to survey by direct observation. The likelihood of being able to detect all individuals through direct observation both on the ground and in colour lowers significantly as koalas are often hidden. As stated, only 60-75% of koalas have been detected within a survey area by experienced observers. Experienced koala spotters in south-east Queensland were able to observe 70% of koalas that were present. With the decline of experts and increased vegetation density, the likelihood of detection has shown to be decreasing significantly with non-expert observers only able to observe 23% of koalas present in an area. As seen throughout all the results found the automated detection method was more accurate and had a higher detection rate as opposed to the manual method.

Overall, it is shown that the automated method is more beneficial than the manual method for detecting koalas. This method is revealed to be more efficient, non-invasive and also less time consuming as opposed to the on-ground observational surveys. As shown throughout the results the automated method had a higher overall probability of detection in all of the testing’s. This states that the automated method is more accurate and has a higher precision rate than that of the manual method.