Unlock instant, AI-driven research and patent intelligence for your innovation.

A Sentinel-1 radar image classification method based on convolution neural network is presented

A convolutional neural network, radar image technology, applied in the field of radar image classification, can solve problems such as holes and isolated points, affecting the accuracy and effect of image processing, and achieve good economic and social benefits.

Active Publication Date: 2019-02-19
EAST CHINA JIAOTONG UNIVERSITY
View PDF4 Cites 7 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] At present, many studies at home and abroad focus on the classification of icebergs in the TerraSAR-X band, but there are few studies on the classification of radar images in the Sentinel-1C band, and the traditional radar images often have holes and isolated points, which easily affect the accuracy and quality of image processing. Therefore, a Sentinel-1 radar image classification method based on convolutional neural network is proposed to solve the above problems

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
  • A Sentinel-1 radar image classification method based on convolution neural network is presented
  • A Sentinel-1 radar image classification method based on convolution neural network is presented
  • A Sentinel-1 radar image classification method based on convolution neural network is presented

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0034] Since the input of the CNN network is a two-dimensional image of a fixed size, it is necessary to cut the satellite image collected by the radar into a 75*75 ROI image by cutting; after the satellite image is cut, 1604 HH and HV are generated Band ROI image dataset.

[0035] CFAR algorithm: When performing the CFAR algorithm, it is necessary to determine the size and position of three windows, namely: the box CFAR window, the cell under test (CUT) and guard window. The box CFAR window represents the range of statistical calculations. Since the sample image It is not big, so the box CFAR window is set to be the same size as the image after ROI, ie 75*75. The cell under test (CUT) is set at the center of the image, a center point of 75*75, numbered from 0 according to the pixel coordinates should be (75-1) / 2, and the coordinates of the center point should be (37, 37) . The guard window is set to 21*21 pixels according to the approximate size of the target.

[0036] The...

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 discloses a convolutional neural network-based Sentinel. a radar image classify method comprises that follow steps: step A, after the image is inputted, preprocessing the image, wherein, the image preprocessing comprise ROI cutting, normalization processing, CFAR algorithm processing and RGB graphic analysis; B, training the preprocessed image, selecting a group of functions f1, f2... Fn, training the data, and selecting the best function f; C, testing that best function f selected in the step B so as to classify the collected images. The invention provides an image preprocessing method suitable for a data set, which utilizes a CFAR method to extract features of the image, removes outliers and fills holes. A 4-layer convolution neural network model with multi-channel input mode is constructed. When CNN is used to train the data set, over-fitting can be avoided by regularization and data enhancement, and good classification effect of 91% can be obtained.

Description

technical field [0001] The invention relates to a radar image classification method, in particular to a convolutional neural network-based Sentinel-1 radar image classification method, which belongs to the technical field of radar image classification applications. Background technique [0002] In the application of an adaptive fast CFAR algorithm based on automatic truncation in high-resolution SAR image target detection, an adaptive CFAR detection algorithm is proposed, and the corresponding clustering statistical model is constructed to realize the SAR radar image target quick extraction. In the automatic detection of icebergs and ice classification navigation based on TelasAR-X images, TerraSAR-X is used to monitor sea ice and icebergs on HH polarized images. After extracting text features, it is used as the input of the neural network, and then used The iterative CFAR algorithm detects the image, finds icebergs and sea ice areas, and enhances the interpretation effect ...

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
IPC IPC(8): G06K9/62G06K9/46G06K9/32G06N3/04
CPCG06V10/25G06V10/40G06N3/045G06F18/24
Inventor 宋岚
Owner EAST CHINA JIAOTONG UNIVERSITY