Using image processing techniques to maximize the value of autonomous MPA monitoring for managers

Monitoring vessel activity near marine protected areas (MPAs) is an important part of effective management. The Marine Monitor (M2) autonomously documents vessel activity in nearshore areas using X-band marine radar and a high-definition camera which helps MPA managers maximize available resources. But there are drawbacks to automated data collection, including capturing images during poor visibility, a high volume of images, and time required to review them. We developed a heuristic model that used traditional image processing techniques to classify images collected by M2 as vessel presence or absence with 72.7% and 72.9% accuracy, respectively. The majority of images with vessel presence associated with a unique vessel record were also classified correctly. Therefore, the computationally inexpensive model is an effective method for selecting images to be used for further classification of vessel type using more complex models. By autonomously identifying these images, managers can more effectively concentrate on data of interest.