Animal Detection Using Thermal Images and Its Required Observation Conditions

Yu Oishi, Hiroyuki Oguma, Ayako Tamura, Ryosuke Nakamura, Tsuneo Matsunaga
2018 Remote Sensing  
Information about changes in the population sizes of wild animals is extremely important for conservation and management. Wild animal populations have been estimated using statistical methods, but it is difficult to apply such methods to large areas. To address this problem, we have developed several support systems for the automated detection of wild animals in remote sensing images. In this study, we applied one of the developed algorithms, the computer-aided detection of moving wild animals
more » ... DWA) algorithm, to thermal remote sensing images. We also performed several analyses to confirm that the DWA algorithm is useful for thermal images and to clarify the optimal conditions for obtaining thermal images (during predawn hours and on overcast days). We developed a method based on the algorithm to extract moving wild animals from thermal remote sensing images. Then, accuracy was evaluated by applying the method to airborne thermal images in a wide area. We found that the producer's accuracy of the method was approximately 77.3% and the user's accuracy of the method was approximately 29.3%. This means that the proposed method can reduce the person-hours required to survey moving wild animals from large numbers of thermal remote sensing images. Furthermore, we confirmed the extracted sika deer candidates in a pair of images and found 24 moving objects that were not identified by visual inspection by an expert. Therefore, the proposed method can also reduce oversight when identifying moving wild animals. The detection accuracy is expected to increase by setting suitable observation conditions for surveying moving wild animals. Accordingly, we also discuss the required observation conditions. The discussions about the required observation conditions would be extremely useful for people monitoring animal population changes using thermal remote sensing images. species are threatened with extinction by hunting or human activities, such as land-use changes. Invasive alien species were introduced intentionally or unintentionally by humans from overseas or from other areas in the country. Overabundant species cause serious damage to agriculture and forestry and affect ecosystems by their considerable population increases and expanding distributions. These problems are not independent, but rather are connected [1]. In particular, conflict between wildlife and humans is a serious problem at the blurred boundary between species near urban areas in Japan. These imprecise boundaries are caused by abandoned farmlands and depopulation by hunters accompanying the rapidly decreasing and ageing rural population. 'Satoyama' is defined as a unique human-influenced natural environment that has been shaped and sustained over a long period by diverse human activity. Studies are needed to determine how to manage satoyama landscapes in such a way that the balance between the needs of humans and nature can be restored, particularly in regard to damage to forests and farmland caused by wild animals [2] . In particular, sika deer (Cervus nippon) have caused serious damage to the agricultural and forestry industries by tree bark-stripping and ecosystem changes. The amount of damage to agriculture by wild animals totaled about 166 million US dollars, and that by sika deer was about 57 million US dollars in 2014 (1 US dollar = 115 Japanese yen) [3] . To address these problems, the conservation and management of wild animals is necessary. The adaptive management of wild animals, a systematic approach for improving resource management by learning from management outcomes [4, 5], is essential. Adaptive management of wild animals consists of the prediction of increases and decreases, plans for hunting goals, abundance adjustment, and monitoring abundance using population indices to determine changes in the population size of wild animals. However, there is insufficient population information for sika deer, which are crepuscular animals with large habitat areas [6] . Although wild animal populations have been estimated using statistical methods, it is difficult to apply such methods to large areas because they require man power and are tremendously laborious. To resolve these issues, remote sensing is a promising method. However, even in open areas, it is difficult to identify animals in remote sensing images because the shapes of objects may differ markedly when viewed from above instead of from the side, as humans are accustomed to doing. Moreover, there is the potential for oversight because an enormous amount of data must be analyzed [7] . To address this issue, we have developed several support systems for the automated detection of wild animals in remote sensing images. These systems reduce the number of person-hours required to survey wild animals from large numbers of remote sensing images. One of these is the computer-aided detection of moving wild animals (DWA) algorithm [8] . The algorithm automatically extracts moving wild animals using the time difference between two overlapping visible images. The advantages of the algorithm are as follows. (i) Almost no detection errors occur, even in sparse forest areas: In forest areas, a tip of a tree appears in different positions when viewed from different points. Accordingly, simple differences between two images can cause detection errors. The DWA algorithm can be applied to sparse forests because relief displacement effects do not cause false detection; (ii) Applicable to large areas: The DWA algorithm does not require fixed cameras because it uses overlapping areas in photographs taken from a flying platform. Thus far, the use of the algorithm has been limited to the daytime because visible and near-infrared images have been used. However, many large mammals, such as sika deer, are crepuscular. For this reason, we used thermal images, in which animals can be identified in semi-dark conditions. Very few studies have used thermal remote sensing images to monitor wild animals [9] [10] [11] [12] . Furthermore, it is difficult to distinguish animals from trees in thermal images in some observation conditions [9, 10] because surface temperature contrast between detection targets and the background is essential to extract targets from thermal images. Therefore, existing studies of the application of thermal remote sensing images to monitor wild animals [11, 12] are limited to open, cool areas. Urban areas contain many hotspots, such as streetlights; thus, we attempted to use pairs of overlapping thermal images obtained at different times to automatically extract only moving animals.
doi:10.3390/rs10071050 fatcat:tgvht5wimbfzxofefnpigti3km