Ship Target Identification Method Based on Multi-level Feature Deep Fusion in SAR Image

Abstract

The invention discloses a SAR image ship target identification method based on deep fusion of multi-level features, which includes the following steps: (1) using Haar-like feature templates to process SAR image samples and reduce the dimension to obtain low-level Haar-like features; (2), use convolutional neural network to process SAR image samples to obtain high-level deep features; (3), use multi-level deep learning network to fuse low-level Haar-like features and high-level deep features to obtain multi-level Feature weight coefficients, and then obtain the optimal SVM classifier through learning and training; (4), use multi-level feature weight coefficients and SVM classifiers to identify the input SAR image sample slices to be identified. The invention can effectively improve the detection performance of the ship target in the SAR image, and has high engineering application value.

Description

technical field [0001] The invention relates to the technical field of SAR image target recognition, in particular to a SAR image ship target identification method based on deep fusion of multi-level features. Background technique [0002] Synthetic Aperture Radar (SAR) is a high-resolution imaging radar with all-weather and all-weather observation capabilities. The research and technology development of ship target identification and surveillance using SAR images has been highly valued in the field of marine remote sensing, and it is a research hotspot in the marine application of SAR images at this stage. [0003] So far, a large number of feature extraction algorithms based on edge textures have been proposed and applied in the field of target identification, including Local Binary Patterns (LBP), Histogram of Oriented Gradients (HOG), Competitive Local Binary Patterns (CompletedLocal Binary Patterns, CLBP), Multi-ScaleCompleted Local Binary Patterns (MS-CLBP), Gray Leve...

AI Technical Summary

Problems solved by technology

However, when the traditional deep learning network performs target feature extraction and identification, it will cause the loss of low-level features such as the edge texture of the ship target. The accuracy of ship target identification in SAR images with low resolution and small samples is not high, which limits its direct Applied to SAR image ship identification

Especially in complex environments such as multiple targets, the lack of samples and edge features reduces the target discrimination rate of traditional deep learning networks.

Images