With the continuous maturity and improvement of machine learning, an intelligent maritime monitoring system is constructed for the identification and tracking of dynamic fuzzy targets in maritime video images. After studying the characteristics of the Marine target environment, it is found that the imaging effect is poor due to the violent movement of the offshore imaging platform, and the field of view of the ship is relatively small, so it is difficult to extract the characteristic information of the corresponding target. Depending on these reasons above, under the CAFFE framework of deep learning, the original and improved yolov2-a network model, SVM support vector machine, HOG feature, multi-scale transformation and multi-thread technology are adopted to identify multiple targets. The fast and effective intelligent automatic detection and recognition algorithm for fuzzy image is completed, which supports a variety of standards and definition. The final research results can detect the weak floating, small boats, speedboats, cruise ships and other targets, the identification accuracy reached more than 95%, for hd video image, real-time up to 30 frames/second.
Study on the Recognition of Visible Image at Sea Based on YOLOv2-Network Model
Lect. Notes Electrical Eng.
2021-10-30
12 pages
Article/Chapter (Book)
Electronic Resource
English
Study on the Recognition of Visible Image at Sea Based on YOLOv2-Network Model
British Library Conference Proceedings | 2022
|An Improved Convolution Neural Network for Object Detection Using YOLOv2
British Library Conference Proceedings | 2018
|