Pedestrian detection method based on semantic segmentation information

A technology of semantic segmentation and pedestrian detection, which is applied in the field of pedestrian detection based on neural network, can solve the problems of blurred semantic information segmentation, difficulty in distinguishing pedestrians from the background, and low detection accuracy, so as to improve positioning accuracy, improve performance, and avoid The effect of false detection and missed detection

Inactive Publication Date: 2018-08-24
UNIV OF ELECTRONICS SCI & TECH OF CHINA
View PDF6 Cites 21 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is: the present invention provides a pedestrian detection method based on semantic segmentation information, which solves the problem of low detection accuracy caused by the difficulty of distinguishing pedestrians from the background in the case of low-resolution existing pedestrian detection and the use of convolutional neural networks. Semantic Information Segmentation Fuzzy Adjacent Target Boundary Leads to Missed Detection and False Detection Problems

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
  • Pedestrian detection method based on semantic segmentation information
  • Pedestrian detection method based on semantic segmentation information
  • Pedestrian detection method based on semantic segmentation information

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0043] A pedestrian detection method based on semantic segmentation information, comprising the following steps

[0044] Step 1: Input the original RGB image in the training set sample into the backbone network in the overall network, input the corresponding semantic segmentation image into the branch network in the overall network, and set the loss function of the overall network to complete the training;

[0045] Step 2: Input the original RGB image in the test set sample into the backbone network in the overall network that has been trained to perform convolutional feature extraction to generate a multi-layer feature map;

[0046] Step 3: Input the multi-layer feature map into the area generation network in the trained overall network to extract the pedestrian candidate frame and generate the pedestrian candidate area;

[0047] Step 4: After the classification and regression network in the trained overall network classifies and locates the pedestrian candidate area, it outp...

Embodiment 2

[0049] Step 1 includes the following steps:

[0050] Step 1.1: Initialize the branch network of the backbone network in the overall network to determine the loss proportion λ i ;

[0051] The initialization is as follows: the parameter initialization of the backbone network adopts the pre-training initialization method, the initialization of the branch network adopts random initialization, the first 60,000 iterations, the learning rate is 0.001, and the last 20,000 iterations, the learning rate is 0.0001, the momentum is set to 0.9, and the weight The attenuation is set to 0.0005, and the loss specific gravity λ i Take 1, the loss proportion is determined according to different training sets.

[0052] Step 1.2: Input the original RGB image into the backbone network and the corresponding semantic segmentation image into the branch network to complete the selection of foreground samples and background samples and generate multi-layer feature maps;

[0053]Select the foregroun...

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 present invention discloses a pedestrian detection method based on semantic segmentation information, and relates to the field of a pedestrian detection method based on a neural network. The pedestrian detection method based on semantic segmentation information comprises the steps of: 1: inputting original RGB images in training set samples into a backbone network, inputting corresponding semantic segmentation images into branch networks, and setting a loss function of the whole network to complete training; 2: inputting the original RGB images in THE training set samples into the backbonenetwork which completes training to perform convolutional feature extraction and generate a multi-layer feature map; 3: inputting the multi-layer feature map into an area generation network which completes training to perform pedestrian candidate frame extraction and generate a pedestrian candidate area; and 4: performing classification and location of the pedestrian candidate area by a classification regression network which completes training to output detection result images including a pedestrian position surrounding frame. The problem is solved that pedestrians and backgrounds are difficult to be distinguished in a low-resolution condition to cause low detection precision in the current pedestrian detection, and the precision of pedestrian detection is improved in the low-resolutioncondition.

Description

technical field [0001] The invention relates to the field of pedestrian detection methods based on neural networks, in particular to a pedestrian detection method based on semantic segmentation information. Background technique [0002] Pedestrian detection technology is the most basic and common target detection technology in practical applications. It is the basis of human behavior analysis, gait recognition, intelligent video surveillance and automatic driving technology. With the rise of convolutional neural networks in recent years, great progress has been made in the field of target detection, but in the field of pedestrian detection, there are still two major challenges: [0003] 1. Compared with general target detection, it is more difficult to distinguish pedestrian targets from the background. For example, in the case of low resolution, pedestrian targets have very similar surface features to traffic lights and columnar targets, and the pixel distribution of pedes...

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): G06K9/00G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06V40/10G06N3/045G06F18/2413
Inventor 杨昕梅李耀斌高原杨承李绍荣
Owner UNIV OF ELECTRONICS SCI & TECH OF CHINA
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