Aerial image small target detection method based on feature fusion and up-sampling

A small target detection and aerial image technology, applied in the field of aerial image target detection, can solve problems such as difficulty in convergence and poor results with small batch sizes, and achieve the effects of increasing computational overhead, improving feature upsampling methods, and improving representation capabilities

Active Publication Date: 2020-07-28
SOUTH CHINA UNIV OF TECH
View PDF4 Cites 7 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Existing literature 2 proposes a feature normalization method, which improves the problem that the original batch normalization is not effective when the batch is too small during network training, and it is difficult to converge to the optimal solution.

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
  • Aerial image small target detection method based on feature fusion and up-sampling
  • Aerial image small target detection method based on feature fusion and up-sampling
  • Aerial image small target detection method based on feature fusion and up-sampling

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0058] A small object detection method in aerial images based on feature fusion and upsampling, such as figure 1 shown, including the following steps:

[0059] S1, using the backbone network to extract the feature set of the input image;

[0060] The backbone network is a residual convolution network, the residual convolution network includes five stages, each stage is composed of several similar residual modules connected in series, and the resolution of each residual module output feature map is the same; each There is a 2-fold downsampling between adjacent stages, and the length and width of the feature map after downsampling are reduced by two times; the final extracted feature set is a set composed of the last feature map of the second to fifth stages of the backbone network.

[0061] S2. Constructing a channel standardization module to standardize the channel dimensions of the features extracted in step S1;

[0062] The channel normalization module is realized by a con...

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 an aerial image small target detection method based on feature fusion and up-sampling. The method comprises the following steps: extracting a feature set of an input image by using a backbone network; constructing a channel standardization module, and standardizing the channel dimensions of the features; constructing an up-sampling layer based on learning, and performing resolution up-sampling on the features to obtain a feature set with uniform resolution; carrying out group normalization of grouping the features according to channels; splicing the feature sets to generate fusion features; performing down-sampling on the fusion features for multiple times, and constructing a feature pyramid for detection; classifying and locating targets using a head detection network. The invention relates to a feature fusion and feature up-sampling method for a target detection training and testing stage, which can significantly improve the detection precision of small targets in aerial images and only increase a small amount of calculation overhead.

Description

technical field [0001] The invention relates to the field of object detection in aerial photography images, in particular to a method for detecting small objects in aerial photography images based on feature fusion and upsampling. Background technique [0002] Compared with surveillance cameras with fixed positions and fields of view, cameras on UAVs have natural advantages, such as easy deployment, strong maneuverability, and wide field of view. These advantages are expected to serve many applications, such as security monitoring, search and rescue, and people flow monitoring. Object detection in aerial imagery is a key component in many UAV applications and is critical to the development of fully autonomous systems, thus becoming an urgent need in the industry. [0003] Although convolutional neural networks have achieved remarkable results in the field of general object detection, their performance in drone aerial photography scenes is not satisfactory. The main reason ...

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
CPCG06V2201/07G06F18/24G06F18/254G06F18/253Y02T10/40
Inventor 林沪刘琼
Owner SOUTH CHINA UNIV OF 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
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap