Automatic litter detection using AI in environmental surveillance aircraft

Abstract

Plastic pollution is an always-growing problem in earth’s oceans. In this paper, we propose an aerial method to detect marine plastic litter, which can be utilized on oil pollution control aircraft already in use in many parts of the globe. With this approach resources are saved, and emission are reduced, as no additional aircraft has to take off. To prevent the growing accumulate of plastic litter in our oceans, two major approaches are necessary. First, one has to detect and collect the plastic that has already reached the ocean. Second, sources of plastic litter have to be found to prevent more plastic from reaching the oceans. Both approaches can be targeted using sensors on airborne platforms. To achieve this, we propose a method for litter detection from aircraft using artificial intelligence on data gathered with sensors that are already in use. For oil pollution control multiple aircraft are already flying in different regions all over the world. Sensors used on these aircraft are partially adapted and utilized in a new way. The detection of plastic is performed using a high frequency, low resolution visual line sensor. If plastic is detected, a high-resolution camera system is targeted on the detected plastic using a gimbal. These high-resolution images are used for verification and classification purposes. In addition to the development of the method for plastic detection, results from intermediate field tests are presented.