Unmanned aerial vehicle (UAV) for sea turtle skeleton detection in the Mexican Pacific

By means of a Parrot ANAFI UAV, we obtained 8000 images with a spatial resolution of 1.40 cm along 30 km of a nesting beach of Lepidochelys olivacea. The images were used to construct orthomosaics for the detection of turtle shells using object-based image analysis. Four attributes describing the shapes of the shells were obtained from the processes of segmenting the images: compactness, convexity, roundness and elongation. A total of 54 sea turtle skeletons were quantified in the field, 51 were detected by object-based image algorithm, two records turned out to be 20-L plastic bottles and one not detected because it was very old and covered by sand. The value of the coefficient and its 95% confidence interval was kappa = 0.787 (0.015), indicating considerable agreement. The shape attributes of roundness were consistent with values for circular shapes, and the convexity values were greater than 1, indicating that they were convex, which is consistent with the shape of marine turtle shells. The advantages of using the Parrot ANAFI to detect skeletons on the beach is that it can survey a visual image field from a nadir angle of 90, which enables all skeletons to be detected regardless of the slope of the berm, as well as its affordable price.