Side-scan sonar image classification method based on multi-task learning
A multi-task learning and side-scan sonar technology, which is applied to instruments, biological neural network models, character and pattern recognition, etc., can solve the problems of small sample size, poor quality, and poor effect, and achieve the effect of enriching the feature space
Active Publication Date: 2020-03-27
HARBIN ENG UNIV
View PDF9 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0004]In summary, side scan sonar is widely used in ocean detection, but due to the small sample size and poor quality of side scan sonar image targets, the current target recognition method does not perform well on classification of side-scan 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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment 1
[0057] A side scan sonar image classification method based on multi-task learning, specifically comprising the following steps:
[0058] Step 1: Construct a side-scan sonar image data set, and classify the existing side-scan sonar data sets. Here, the three categories of aircraft, ships, and targets other than aircraft and ships are used as examples to illustrate;
[0059] Step 2: Obtain the aircraft and ship images in the optical scene similar to the aircraft and ship images in the side-scan sonar image data set from the existing data set;
[0060] Step 3: Build a multi-task learning model based on convolutional neural network, which is modified on the basis of VGG11 convolutional neural network model. The model has two inputs and shares a convolutional pooling layer. Each input has its own classification layer (fully connected layer) and calculates its own loss function separately. The total loss function of the multi-task network model is the sum of the weighted loss func...
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
Login to View More Abstract
The invention belongs to the technical field of side-scan sonar image recognition, and particularly relates to a side-scan sonar image classification method based on multi-task learning. According tothe invention, a multi-task learning idea and a convolutional neural network method are combined, the convolutional neural network is used for automatic feature extraction, and compared with a traditional manually-arranged feature extractor, important features which cannot be felt by human eyes can be extracted, and the influence of factors such as side-scan sonar image noise, image edge missing and image feature deformation on feature extraction can be reduced. According to the method, the idea of multi-task learning is utilized, and an optical image classification task is introduced, so thatthe feature space of the side-scan sonar image can be enriched, and the problem of over-fitting caused by incomplete feature extraction when samples are too few is avoided. According to the method, the problems that side-scan sonar image samples are few, and the classification effect is poor when feature extraction is difficult can be solved, and the method has certain engineering and research value.
Description
technical field [0001] The invention belongs to the technical field of side-scan sonar image recognition, and in particular relates to a side-scan sonar image classification method based on multi-task learning. Background technique [0002] Compared with the optical imaging system, the side-scan sonar system can overcome the limitation of harsh conditions in the water and obtain images of the seabed from a long distance. Sonar images have incomparable uses in the fields of seabed surveying and mapping, seabed geological survey, deep-water combat, and seabed rescue. However, due to the vastness of the ocean, it is very difficult to collect enough side-scan sonar images with search targets. On the other hand, since light, temperature, sand, etc. may affect the quality of the side-scan sonar image, these factors make the target features in the side-scan sonar image far from the real target features, so only limited side-scan sonar images are used. Scanning sonar images are di...
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
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/62G06N3/04
CPCG06N3/045G06F18/214G06F18/241
Inventor 叶秀芬杨鹏刘文智李海波李传龙李响仰海波
Owner HARBIN ENG UNIV