DEEP LEARNING FOR AUTOMATIC RECOGNITION OF BOATS AND SHIPS TO AVOID COLLISIONS IN MARINE TRANSPORT

Abstract

Listen

Translate article

The objective of this work is to explore the use of deep learning for the automatic recognition of ships and boats. The study uses a deep convolution neural network (DCNN) architecture to analyze a large data set of ship and boat images. The data set is prepared and labelled for supervised learning, and CNN is fine-tuned using the YOLO series group of versions 5, 6 and 8 to improve recognition accuracy. The proposed method involves training a deep convolution neural network on a large data set of ship and boat images and using the trained model to classify new images. The performance of the model is evaluated on a separate test set and compared to other state-of-the-art methods. The results of this study show that the deep learning model is effective in automatically recognizing ships and boats with an accuracy of mAP (Mean Average Precision). The model’s performance is also compared to traditional machine learning algorithms, and CNN outperforms these methods. In this work we present the architecture, design and implementation of an object detection model deployed on an IMX8M Plus hardware board, to be used on the gathered image data model to recognize and label the ships and boats at the edge. We conduct transfer learning on the state-of-the-art trained YOLO model by introducing a labelled BS(boat-ship) image data set. We use the trained model to do predictions on a test image set to evaluate the model’s performance. The result of the model can predict labels with an accuracy of 72.1 mAP of YOLOv8 and inference FPS time that detect it to do so in real-time with the board. Results show that the proposed deep learning approach outperforms existing methods, achieving high accuracy, and demonstrating the potential of deep learning for the automatic recognition of boats and ships in marine environment system.