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

Method of detecting network for underwater target detection

An underwater target and network technology, applied in the direction of instruments, character and pattern recognition, computer components, etc., can solve problems such as large precision gaps, and achieve the effect of enhancing interactivity and fluidity, ensuring speed, and high precision

Active Publication Date: 2020-04-03
DALIAN UNIV OF TECH
View PDF6 Cites 6 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Redmon et al. proposed YOLO to directly predict the class and location of each object using an end-to-end CNN, but there is still a large accuracy gap between YOLO and other two-stage methods

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
  • Method of detecting network for underwater target detection
  • Method of detecting network for underwater target detection
  • Method of detecting network for underwater target detection

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0015] In order to make the purpose, technical solution and advantages of the present invention clearer, the specific implementation manners of the present invention will be further described in detail below.

[0016] Implementation using CUDA10.0 and cuDNN7.3.1 backend on NVIDIA TITAN XP GPU, Intel Xeon CPU E5-2680 v4. The UnderwaterNet is implemented on PyTorch. The image resolution for both training and inference is 512×512. The Lookahead optimizer with Adam was used and the initial learning rate was set to 2.3e-5. The batch size is 32. We used zero-mean normalization, random flipping, random scaling (between 0.6 and 1.3) and clipping to augment the data. Use the UDD dataset as the training data for UnderwaterNet. UDD is a real marine ranch target detection dataset, which contains 2227 pictures (1827 training and 400 testing) of three types of detection targets: sea cucumber, sea urchin and scallop.

[0017] I conducted ablation experiments on the MBP and MFF modules t...

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 computer target detection, and provides a method of detecting a network for underwater target detection. The method comprises the following steps: constructing a neural network based on two lightweight modules, and constructing an UnderwareNet through an MFF module and an MBP module, wherein the former enhances interactivity and fluidity of information in one module, and the latter performs down-sampling through Gaussian blur of different scales, so that translation invariance of the network is enhanced, and different levels of blur feature mapsbeneficial to small target detection are generated. Both the two networks have the characteristics of light weight and multiple scales, thus being suitable for being deployed on an underwater robot, and achieving high precision while ensuring the speed, and a new solution is provided for an underwater target detection task.

Description

technical field [0001] The invention belongs to the technical field of computer target detection, and relates to a deep neural network method for underwater target detection. Background technique [0002] Nowadays, with the increasing demand for ocean exploration, the demand for underwater target detection tasks is becoming more and more obvious. The goal of underwater target detection is to identify and locate creatures in underwater images, for example: automatic recognition of underwater capture robots in marine ranches And locate marine life (sea cucumbers, sea urchins, scallops), etc. In recent years, Convolutional Neural Networks (CNNs) have achieved remarkable success in computer vision tasks and become the main method for object detection. Different levels of features in the CNNs network contain different information, shallow features are rich in image details, and deep features have stronger semantic information. Recently, many studies related to visual recognitio...

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): G06K9/62
CPCG06V2201/07G06F18/253G06F18/214
Inventor 王智慧李豪杰刘崇威王世杰唐涛
Owner DALIAN UNIV OF TECH