Check patentability & draft patents in minutes with Patsnap Eureka AI!

A method and system for generating a subway environment target training set

A technology for generating systems and training sets, applied in neural learning methods, biological neural network models, instruments, etc.

Active Publication Date: 2021-09-17
JIANGSU BIDE SCI & TECH CO LTD
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0002] Due to the complex environment in the subway, a wide range of illumination conditions, and many types of targets, the device is installed on a moving train, which is difficult to detect using traditional background-based learning or morphological 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
  • A method and system for generating a subway environment target training set
  • A method and system for generating a subway environment target training set
  • A method and system for generating a subway environment target training set

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0076] figure 1 It is a flow chart of the method for generating the subway environment target training set of the present invention. Such as figure 1 As shown, a subway environment target training set generation method, including:

[0077] Step 101: Obtain and label the subway environment pictures with targets and various non-metro environment scene pictures.

[0078] Step 102: According to the subway environment picture, train to obtain the first generation network, specifically including:

[0079] Randomly extract a picture of the subway environment with the target and add random noise, substitute the picture of the subway environment with the target into the first generation network as a condition, and train to obtain the first generation network. The preliminary model of the first generation network uses PixelRNN model and weights.

[0080] Step 103: Generate a first random graph according to the first generation network.

[0081] Step 104: Train the first judgment ne...

Embodiment 2

[0111] figure 2 Generate a system structure diagram for the subway environment target training set of the present invention. Such as figure 2 As shown, a subway environment target training set generation system includes:

[0112] The first obtaining module 201 is used to obtain and mark the subway environment picture with the target and various non-metro environment scene pictures;

[0113] The first generation network determination module 202 is used to train the first generation network according to the subway environment picture;

[0114] The first random graph generation module 203 is configured to generate a first random graph according to the first generation network;

[0115] The first decision network determination module 204 is used to train the first decision network according to the subway environment picture and the first random graph, to obtain the trained first decision network;

[0116] A first update module 205, configured to update parameters of the firs...

Embodiment 3

[0139] Step 1: Prepare the target data of concern in various scenarios, including target data in office, field, day, night and other scenarios, and prepare conventional target data in the subway environment, including tracks, communication equipment, etc., and prepare for the non-metro environment routine target data, such as train tracks during the day, communication equipment in an office environment, etc.;

[0140] Step 2: The first stage of training the generative network G 0 , classify and label various environmental scene data (including subway data), and substitute various labels of the generated network and subway environment into the generator as condition c, and the generator first passes in a fully connected layer to generate a Gaussian distribution N[ mu 0 (φ c ), ∑ 0 (φ c )], and then upsample the Gaussian distribution to generate a random graph

[0141] Step 3: The first stage of training the decision network D 0 , for G 0 generated random graph is spa...

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 a method and system for generating a subway environment target training set. A game confrontation generation network is used to automatically generate possible obstacles into the natural environment of rail transit by learning the characteristics of normal objects in urban rail transit. At the same time, a decision network is used to distinguish generated data from real data. Through the game process of generative network and decision network, the real simulation of urban traffic obstacles is realized, and the target training set of subway environment is obtained. The accuracy of target training can be improved by adopting the present invention.

Description

technical field [0001] The invention relates to the field of subway traffic, in particular to a method and system for generating a subway environment target training set. Background technique [0002] Due to the complex environment in the subway, a wide range of illumination conditions, and many types of targets, the device is installed on a moving train, which is difficult to detect with traditional background-based learning or morphological methods. The artificial intelligence target training method requires a large number of data sets to support, and the installation conditions of the subway train restrict the computing power of the underlying processing platform, making it difficult to use a large-scale network with strong generalization ability for training. The target training set is especially important to improve the accuracy of target training. Contents of the invention [0003] The purpose of the present invention is to provide a method and system for generating...

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 Patents(China)
IPC IPC(8): G06K9/00G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06V20/58G06N3/045G06F18/241G06F18/214
Inventor 蒋怡亮张奔宇顾青松姚伟宇陈皓晞张道
Owner JIANGSU BIDE SCI & TECH CO LTD
Features
  • R&D
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More

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

© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com