Visual perception device and method for ship navigation environment

Abstract

The invention relates to a visual perception device and method for a ship navigation environment. The method comprises the following steps: acquiring a multi-frame fused polarization image, a radar data graph of the ship navigation environment and an RGB (Red, Green and Blue) image; performing information fusion on the fused polarization image, the radar data graph and the RGB image, inputting the fused polarization image, the radar data graph and the RGB image into an improved condition generative adversarial network for data enhancement, and generating fused enhanced data; and inputting the fusion enhancement data into a multi-scale convolutional neural network for scene segmentation identification, generating a scene segmentation map, and identifying different objects in the ship navigation environment. According to the method, scene image features are acquired in a multi-scale and multi-mode manner, visual features of different scenes on water are captured, data fusion is performed based on radar and vision to improve the stability and accuracy of a visual detection result, a data enhancement method and a multi-scale convolutional neural network are adopted, the effectiveness of network identification and classification is ensured, and autonomous navigation of the intelligent ship is realized.

Description

technical field [0001] The invention relates to the technical field of autonomous navigation, in particular to a visual perception device and method for a ship's navigation environment. Background technique [0002] Traditional image semantic segmentation methods mainly include pixel-level threshold method, segmentation method based on pixel clustering and segmentation method based on graph theory partition. It mainly relies on the low-dimensional visual features of the image for segmentation. Based on visual features such as color, texture, and edge, some feature extraction algorithms are used to extract visual information such as edge features and texture of objects in the image, and then according to these low-level visual features. The regions and objects in the image are segmented. For example, commonly used image features include directional gradient histogram features, SIFT features, SURF features, local binary features (LBP), Gabor features, etc. [0003] The method...

AI Technical Summary

Problems solved by technology

[0004] In view of this, it is necessary to provide a visual perception device and method for a ship's navigation environment to solve the problem that it is difficult to accurately identify complex water surface navigation environments in the prior art

Images