Method and system for ship detection in video images under the influence of ocean waves based on deep learning

Abstract

The present invention provides a method and system for detecting ships in video images under the influence of ocean waves based on deep learning. The video image data is marked, including the labeling of ship data and ocean wave data; the ROI of the region of interest is extracted to obtain the ships and easy objects in the image. The position information of the wave target confused with the ship; the ROI of the region of interest is arranged and combined in pairs, and the dual-branch convolutional neural network structure constructed includes an input layer, multiple hidden layers of two branches, and an output layer; Paired ROIs of the same category or ROIs of different categories are used as the network input, and the similarity measurement results of the two images are used as the target output. The characteristics of the paired images are learned through the dual-branch convolutional neural network structure, and the network parameters are adjusted until they are trained well. A deep learning network; for the video frame images in the ship monitoring system on the shore under windy conditions, the trained deep learning network is used to distinguish between waves and ship images to achieve ship detection under the influence of waves.

Description

technical field [0001] The invention belongs to the field of target detection, and specifically relates to a method for detecting ships in video images under the influence of ocean waves based on deep learning. Background technique [0002] As a part of object detection, ship detection has always been a hot topic in the field of computer vision and pattern recognition. In many applications such as marine security, maritime monitoring and management, ship detection can provide decision-making information for it. [0003] Traditional target detection is usually based on machine learning, through artificial feature design, feature extraction, and inputting the extracted features into a classifier to achieve target detection tasks. With the development of deep learning, object detection has entered a new stage. Different from traditional artificially designed features, convolutional neural network (CNN) can automatically extract more representative features. Since no manual fe...

AI Technical Summary

Problems solved by technology

However, the method used in this literature is a traditional method, and image feature selection requires manual intervention, which is highly subjective.

In addition, there is a lack of quantitative indicators for evaluating the accuracy of ship detection under the influence of sea waves in the literature

In recent years, in the research of target detection based on deep learning, the more popular method is YOLO, which has the advantage of being able to meet the needs of real-time detection, but this method is not applicable under the influence of ocean waves, and its detection accuracy is low

Images