Detecting litter objects using an aerial drone with convolutional neural networks

Abstract

Marine litter is leaving a highly negative impact on the cleanliness of oceans since plastics do not biodegrade and remain intact for centuries, making it essential to monitor for any litter while providing relevant knowledge to develop a long-term policy to eliminate litter. Consequently, the decay of the aesthetic significance of beaches results in a reduction of profits from the tourism sector and procures higher costs in the clean-up of coastal regions and their surroundings. For these reasons, UAVs (Unmanned Aerial Vehicle) can be effectively used to identify and observe beach litter since they make it possible to readily monitor the entire beach while CNNs (Convolutional Neural Networks) can classify the type of litter that is present. This dissertation evaluates approaches that can be used for litter object detection by using highly efficient models. Object detection refers to estimating the locations of objects in each image while labelling them with rectangular bounding boxes. The process of the solution began by gathering a custom data set of different types of litter, which were then used for training the model. The data set incorporates four variances of litter: aluminium cans, glass bottles, PET bottles, and HDPE bottles. Once the appropriately trained results were achieved for each object detection model, the results could be compared by using widely known standards to determine which model is the most accurate. The Tello EDU drone was then used to capture video footage on which the detections can be made. The trained model was finally inputted to a primary system that controls the drone and in return, accepted the video feed captured by the drone. The results of this dissertation achieved satisfactory results as both models implemented were efficient, however, the Tiny-YOLOv3 model proved to be more useful since it performs better on videos due to its fast nature and capability to require less hardware by occupying less memory space. Moreover, the project can be further improved in the future by incorporating more litter types.