Improved YOLOv3 minimum remote sensing image target detection method and device and storage medium

A remote sensing image and target detection technology, applied in the field of target detection, can solve the problems of high false alarm rate, slow detection speed, and low detection rate of extremely small remote sensing image targets, and achieve the effect of improving detection speed and detection performance

Active Publication Date: 2020-07-28
UNIV OF SHANGHAI FOR SCI & TECH
View PDF7 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] Aiming at the deficiencies of the prior art, the present invention discloses an improved YOLOv3 minimal remote sensing image target detection method, device and storage medium,

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
  • Improved YOLOv3 minimum remote sensing image target detection method and device and storage medium
  • Improved YOLOv3 minimum remote sensing image target detection method and device and storage medium
  • Improved YOLOv3 minimum remote sensing image target detection method and device and storage medium

Examples

Experimental program
Comparison scheme
Effect test

Example Embodiment

[0027] Example 1

[0028] This embodiment discloses an improved YOLOv3 method for detecting extremely small remote sensing image targets. The method includes the following steps:

[0029] S1 obtains the target data, and merges the convolutional layer feature output of the YOLOv3 network to form a pyramid feature layer;

[0030] S2 combines the feature output of the YOLOv3 shallow convolutional layer to form a pyramid feature layer;

[0031] S3 merges the bidirectional combination of the pyramid feature layer;

[0032] S4 changes the downsampling layer of the YOLOv3 network to a 3×3 convolutional layer;

[0033] S5 uses 1×1 convolution to reduce the dimension of the network model and output data.

[0034] In the S1, the convolutional layer feature output of the last layer of the YOLOv3 network is fused with the convolutional layer feature output of the adjacent upper layer to form a top-down pyramid feature layer.

[0035] In the S2, the convolutional layer feature output of the shallow lay...

Example Embodiment

[0038] Example 2

[0039] This embodiment discloses an improved YOLOv3 detection device for extremely small remote sensing image targets, including an FPN module, a memory, a processor, and a computer program stored in the memory and running on the processor. The computer program When executed by the processor, the improved YOLOv3 method for detecting targets in extremely small remote sensing images described in the foregoing embodiment is realized.

[0040] Add additional bottom-up, horizontal connection paths on the FPN module.

[0041] This embodiment also discloses a storage medium on which a computer program is stored, and when the computer program is executed by a processor, it implements the improved YOLOv3 method for detecting a very small remote sensing image target described in the above embodiment.

Example Embodiment

[0042] Example 3

[0043] In this embodiment, image classification can predict the category of a single target in the image, and target detection has its own challenges. It is necessary to predict multiple target categories and corresponding positions in a single image. In order to solve this problem, a pyramid feature method that represents the target feature layer at multiple scales is proposed. The feature pyramid network is one of the representative model architectures for generating the pyramid feature representation for target detection. It uses a top-down, horizontal connection and combination of two adjacent feature layers in the backbone network model to build a feature pyramid to generate high-resolution, strong semantic features. YOLOv3 draws on the idea of ​​the above feature pyramid model, such as Figure 4 Shown. It is based on the backbone network model of Darknet53, with a total of 75 convolutional layers, 5 residual blocks, and 5 times of downsampling. It uses ...

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 relates to the technical field of target detection, in particular to an improved YOLOv3 minimum remote sensing image target detection method and device and a storage medium. According tothe invention, an additional bottom-up and transverse connection path is added to the FPN module to improve the performance of the low-resolution feature; a top-down and bottom-up feature pyramid network is constructed, a bidirectionally combined pyramid feature layer is fused and applied to target detection of a remote sensing image, the dimension of a network model is reduced by adopting 1 * 1convolution, and the detection speed of the network is improved. And finally, quantitative and qualitative comparative analysis is carried out on VEDAI and NWPU VHR remote sensing vehicle data sets and the most advanced YOLOv3 network. The result shows that the detection performance of the improved network is obviously improved compared with the original network, the detection speed of the networkis hardly changed, and the problems of low target detection rate, high false alarm rate and low detection speed of the minimum remote sensing image at the present stage are solved.

Description

technical field [0001] The present invention relates to the technical field of target detection, in particular to an improved YOLOv3 method, device and storage medium for detecting targets in extremely small remote sensing images. Background technique [0002] In recent years, target detection has become a research hotspot in computer vision. It is widely used in many fields such as robot navigation, intelligent video surveillance, industrial inspection, aerospace, etc. It not only needs to determine the category of the target in the image, but also needs to give its precise Location. [0003] In the past few years, researchers have found that designing a deep convolutional neural network can greatly improve the performance of image classification and object detection, because it can not only extract strong semantic high-level features of objects from images, but also combine feature extraction, feature Selection and feature classification are fused into one model, and thro...

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): G06T7/00G06N3/04G06N3/08
CPCG06T7/0002G06N3/08G06T2207/10032G06T2207/20016G06T2207/20081G06N3/045
Inventor 张孙杰陈磊肖寒臣
Owner UNIV OF SHANGHAI FOR SCI & 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