Vehicle target detection method for remote sensing application scene

A technology for vehicle detection and application scenarios, which is applied in the fields of computer vision and artificial intelligence, and can solve problems such as high cost and high data set requirements

Pending Publication Date: 2020-11-06
WUHAN UNIV
View PDF4 Cites 19 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, low/high-resolution paired images are required to train th...

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
  • Vehicle target detection method for remote sensing application scene
  • Vehicle target detection method for remote sensing application scene
  • Vehicle target detection method for remote sensing application scene

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0067] In order to better understand the technical solution of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings. Implement the main processes such as figure 1 As shown, the specific implementation process is as follows:

[0068] Step 1. Use the data enhancement method based on homography transformation to create a diverse remote sensing image vehicle detection dataset. Methods as below:

[0069] Step 1.1, collect representative datasets Potsdam and VEDAI for vehicle detection tasks in remote sensing images, organize and mark them as horizontal outsourcing frames, cut the image size into 600×600 pixels, and convert the semantic annotation of "vehicle" in the image into two groups: upper left and lower right Position information represented by coordinates; collect representative data sets DLR Munich and UCAS-AOD for remote sensing image vehicle detection tasks, make low-resolution vehicle detec...

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 vehicle detection method for a remote sensing application scene, and the method comprises the steps: making a diversified remote sensing image vehicle detection data set through a homography-based data enhancement method; determining candidate area positions and analogy based on a target detection model considering target area characteristics of a remote sensing image vehicle; improving the resolution of the low-resolution remote sensing image based on a super-resolution reconstruction model of the cyclic generative adversarial network; and completing super-resolutionreconstruction and vehicle detection tasks at the same time based on a synchronous super-division and target detection network model of a task driving thought. Compared with a traditional vehicle detection method, the method has the advantages of being high in efficiency, rich in depth feature semantic information and the like; compared with other vehicle detection algorithms based on deep learning, remote sensing image characteristics are considered, and the method has the advantages of being accurate in target positioning, good in generalization performance and the like. In addition, the multi-task learning model improves the vehicle detection precision of the low-resolution remote sensing image.

Description

technical field [0001] The present invention relates to the fields of computer vision and artificial intelligence, in particular to a vehicle target detection method for remote sensing application scenarios based on a deep convolutional network and a multi-task learning model. Taking into account the characteristics of remote sensing images, it can improve the resolution and Accurately locate vehicle targets. Background technique [0002] Optical remote sensing images can be captured with cameras mounted on UAVs, aircraft or satellites. The target detection task in the field of remote sensing is to determine whether an object of interest is contained in a remote sensing image, and to locate the object. Among them, vehicle detection plays an important role in rescue missions, traffic safety, military defense and other fields, and has important theoretical research significance and practical application value. [0003] Commonly used vehicle detection algorithms can be divide...

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): G06T3/40G06T7/90G06K9/62G06K9/32G06N3/04G06F17/16
CPCG06T3/4053G06F17/16G06T7/90G06T2207/10032G06T2207/20016G06V10/25G06V2201/08G06N3/045G06F18/241G06F18/214
Inventor 梅天灿高智冀虹
Owner WUHAN 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
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