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

Deep learning-based adaptive weight convolutional neural network underwater sonar image classification method

A technology of convolutional neural network and self-adaptive weight, which is applied in the direction of biological neural network model, neural architecture, instrument, etc., can solve the complex seabed situation, increase the difficulty of underwater sonar image classification, and cannot take underwater acoustics into account. problem of accepting useful information of images, etc.

Active Publication Date: 2018-08-21
HARBIN ENG UNIV
View PDF7 Cites 29 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The traditional underwater sonar image classification method uses different feature extraction methods to complete the classification, but a specific feature extraction method cannot take into account all the useful information of underwater sonar images, resulting in the bottleneck of underwater sonar image classification
In addition, due to the imaging of underwater sonar images, it is difficult to separate the targets of underwater sonar images from the shadow parts and the reverberation area of ​​the seabed. In addition, the seabed is complicated and noisy, which makes the classification of seabed targets difficult. become difficult
At present, the work on the classification of underwater sonar images is still in its infancy. The seabed is complicated, and the images captured by sonar on the seabed are different every time. Even one class of images is from various angles, which further increases The Difficulty of Classifying Underwater Sonar Images

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
  • Deep learning-based adaptive weight convolutional neural network underwater sonar image classification method
  • Deep learning-based adaptive weight convolutional neural network underwater sonar image classification method
  • Deep learning-based adaptive weight convolutional neural network underwater sonar image classification method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0081] The present invention is described in more detail below in conjunction with accompanying drawing example:

[0082] combine figure 1 , the concrete steps of the present invention are as follows:

[0083] (1) Generate DBN two-dimensional parameter matrix

[0084] The underwater sonar images belong to the small-sample undisclosed data set. The experimental data set of the present invention comes from laboratory collection and collection over the years. The data set is divided into six categories, including underwater sand patterns, sunken ships, sunken planes, stones, tires and school of fish. Due to the obvious advantages of deep learning in big data, the present invention considers various situations of underwater sonar images, such as image angle inclination, noise, etc., and expands the data set. These include H, hv, S, V channel conversion, R, G, B single channel conversion, image flip operation, in order to simulate the complex environment of the seabed, Gaussian ...

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 provides a deep learning-based adaptive weight convolutional neural network underwater sonar image classification method. The method comprises the steps of (1) according to characteristics of underwater sonar images in a data set, generating a DBN two-dimensional parameter matrix; (2) adaptively adjusting distribution of a CNN filter weight matrix; and (3) realizing deep learning-based adaptive weight CNN underwater sonar image classification. According to the deep learning-based adaptive weight convolutional neural network underwater sonar image classification method provided bythe invention, the problem of randomness of filter weight initialization in a CNN can be solved, so that the situation of falling into local optimum is avoided, the classification correctness can bebetter improved, and certain validity is achieved.

Description

technical field [0001] The invention relates to a method for classifying underwater targets, in particular to a method for classifying underwater sonar images. Background technique [0002] In recent years, the underwater target classification technology based on sonar images has been greatly developed, and its application range is becoming wider and wider. Due to the low contrast, blurred edges, and weak texture of underwater sonar images, the image quality is not ideal, which will seriously affect the classification of underwater sonar images, making it a difficult problem. Scholars at home and abroad have conducted in-depth research on the classification of underwater sonar images and achieved important results. Among them, the most famous and effective classification methods in the existing literature mainly include: 1. Underwater sonar image classification based on image segmentation and texture features: 2012 Khidkikar Mahesh, Balasubramanian Ramprasad.Segmentation an...

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): G06K9/62G06N3/04
CPCG06N3/048G06N3/045G06F18/214G06F18/24
Inventor 王兴梅焦佳孙博轩王国强刘安华
Owner HARBIN ENG UNIV
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