Eureka AIR delivers breakthrough ideas for toughest innovation challenges, trusted by R&D personnel around the world.

Large-breadth SAR image ship target detection and identification method based on fine segmentation

A target detection and fine segmentation technology, which is applied in the field of SAR ship target detection and recognition, can solve the problems of high false alarm rate, large SAR image size, and inability to achieve positioning, and achieve the effect of improving detection speed and increasing positioning accuracy

Active Publication Date: 2021-09-17
HUAZHONG UNIV OF SCI & TECH
View PDF4 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] (1) For a large amount of large-format SAR data captured in near real-time by remote sensing satellites or UAVs equipped with SAR, the existing SAR detection network is difficult to take into account the detection speed and detection accuracy, and cannot achieve efficient and high-precision detection of massive large-format data
[0007] (2) SAR data labeling is difficult, and the labeling form of the rectangular box is the main form. The detection network trained based on this rectangular box can locate the ship target relatively rough, and cannot achieve a finer positioning like the rotating box
[0008] (3) Due to the large size of the SAR image and the small ship target, the detection network often misidentifies objects such as islands, reefs and bright spots as ship targets, resulting in a high false alarm rate
[0011] (2) Although the rotating frame detection network can achieve precise positioning of the ship target to a certain extent and outline the general outline of the ship target, the number of existing rotating frame data sets is not enough to support the network for effective training
Therefore, we need to find a new method to solve the problem of rough ship target positioning and realize the fine positioning of ships.
[0012] (3) The imaging mechanism of SAR leads to the lack of significant texture features of ship targets, and most of them are bright cross spots. For islands and ports with buildings of the same material, it is more difficult to distinguish them through the detection network.

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Large-breadth SAR image ship target detection and identification method based on fine segmentation
  • Large-breadth SAR image ship target detection and identification method based on fine segmentation
  • Large-breadth SAR image ship target detection and identification method based on fine segmentation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0084] Aiming at the problems of high false alarm rate and rough target positioning in existing large-format SAR image detection methods, this invention proposes a large-format SAR image ship target detection and classification method based on fine segmentation. In this method, the sliding window strategy and the detection network are used to detect large-scale SAR images, and the detection results of each slice are obtained. These results are mapped back to the original image, and after the fusion of voting strategies, the preliminary detection results of the detection network are obtained. The results contain certain false alarm targets, and the target positioning is not accurate. The area corresponding to the detection result is sent to the segmentation network for segmentation, and the fine positioning result of the ship target is obtained. The target in the minimum bounding box area of ​​the final segmentation result is sent to the classification network, and the target i...

Embodiment 2

[0114] For this invention, this section provides a specific implementation case, from the collection and production of data sets to the training, testing and effect display of each model, such as Figure 4 shown. The specific embodiments described here are only used to explain the present invention, not to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0115] (1) Collection and production of data sets.

[0116] The SAR data taken by the ALOS PALSAR remote sensing satellite from 2006 to 2011 were collected from the NASA Earth Observation System Data Information System. The shooting scenes were selected from the ports and near the coast of the United States, Japan, Taiwan, etc., and the data included 5600×4700, 11300 ×9400 in two sizes. After converting the downloaded data into 16-bit deep single-chan...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention belongs to the technical field of SAR ship target detection and identification, and discloses a large-breadth SAR image ship target detection and identification method based on fine segmentation. The method comprises the following steps: carrying out stretching and cutting of a large-breadth SAR original image, and obtaining a plurality of slice images; detecting the slice images to obtain a detection result, mapping the detection result back to the original image, fusing redundant detection frames by utilizing a voting strategy, and calculating a new target position; sending each target area into a segmentation network FCN to obtain a fine pixel-level segmentation result of a target; and taking out a minimum bounding box of a segmented target, sending the minimum bounding box into a classification network AlexNet, and suppressing a false alarm target to obtain a final detection and identification result. According to the method, the positioning precision and the detection speed of the target are effectively improved in the detection stage, after segmentation and classification network processing, the accuracy of a prediction frame is further improved, the false alarm rate is reduced, and detection and identification of SAR ship fine positioning are achieved.

Description

technical field [0001] The invention belongs to the technical field of SAR ship target detection and recognition, in particular to a large-format SAR image ship target detection and recognition method based on fine segmentation. Background technique [0002] Synthetic Aperture Radar (SAR) has the characteristics of strong penetrating power, high resolution, and can work around the clock. It is one of the most important means for people to observe the earth at present, and has been highly valued by various countries. In 2016, my country's Gaofen-3 satellite was successfully launched, marking my country's first self-developed high-resolution microwave remote sensing satellite. The successful launch of the satellite has effectively improved my country's high-resolution SAR dependence on imports. More and more SARs are carried on aircraft and missiles for battlefield reconnaissance, ocean surveillance and guidance. SAR data also shows an increase year by year. trend. [0003] S...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06T7/11G06T7/00G06K9/62G06N3/04G06N3/08
CPCG06T7/11G06T7/0002G06N3/08G06T2207/10044G06T2207/20132G06T2207/20221G06T2207/30212G06N3/045G06F18/25G06F18/259G06F18/24
Inventor 颜露新曹旭航邰园龚恩谭毅华石清芳黎瑞王健
Owner HUAZHONG UNIV OF SCI & TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Eureka Blog
Learn More
PatSnap group products