Pixel values, and the resulting group identity, may change among images to create a sequence of objects. Pixels are arranged into groups such that pixel proximity, orientation and similarity create a group identity. The atomic unit of data in computer vision is an image pixel that represents colour in the visible spectrum. Computer vision is a form of image-based computer science that uses pixel values to infer image content (LeCun, Bengio, & Hinton, 2015). Computer vision can increase the breadth, duration and repeatability of image-based ecological studies through automated image analysis (Dell et al., 2014 Kühl & Burghardt, 2013 Pennekamp & Schtickzelle, 2013). While image capture has greatly increased sampling, our ability to analyse images remains a bottleneck in turning these data into information on animal presence, abundance and behaviour. To reduce cost, labour and logistics of observation, ecologists are increasingly turning to greater automation to locate, count and identify organisms in natural environments (Pimm et al., 2015). Direct observation of these events can be disruptive to wildlife, and potentially dangerous to observers. Animal presence and behaviour may vary over broad spatial and temporal scales, and depend on important but infrequently observed events, such as breeding, predation or mortality. Many animals are rare, secretive and inhabit remote areas. Observing biodiversity can be expensive, logistically difficult and time-consuming.
0 kommentar(er)
