Population density estimation method based on pyramid feature fusion

A technology of pyramid features and crowd density, applied in neural learning methods, computing, computer components, etc., can solve problems such as multi-scale target perception difficulties, and achieve the effect of enhancing perception, enhancing aggregation ability, and inhibiting activation.

Active Publication Date: 2022-07-15
QINGDAO SONLI SOFTWARE INFORMATION TECH
View PDF6 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to overcome the deficiencies of the prior art, and to design and provide a crowd density estimation method based on pyramid feature fusion, which is used to solve the problem of multi-scale target perception difficulties, and can be used for crowd density estimation in unrestricted actual scenes. In, the density estimation can be accurately realized

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
  • Population density estimation method based on pyramid feature fusion
  • Population density estimation method based on pyramid feature fusion
  • Population density estimation method based on pyramid feature fusion

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0036] In this embodiment, convolutional features are extracted through the backbone, and then four parallel branches are used to further extract multi-scale features. At the same time, attention branches are introduced to focus on the area where the crowd is located. Finally, the multi-scale features are processed by the density map regressor to obtain the final result. The density map includes the following steps:

[0037] (1) Dataset generation: Process the annotation information of the public data set of the crowd, and use the Gaussian kernel to blur the annotation points in the annotation information to form the true density map required for training. Among them, for crowd images with high crowd density, use The geometric adaptive Gaussian kernel uses a Gaussian kernel with a standard deviation of 15 for crowd images with low crowd density; then the dataset is augmented by cropping and flipping to obtain more training data images;

[0038] (2) Feature extraction of backbo...

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 belongs to the technical field of crowd density estimation, and relates to a crowd density estimation method based on pyramid feature fusion, the first thirteen layers of VGG-16 are used as a backbone network to perform preliminary feature extraction, and then a pyramid feature fusion module is designed for solving the problem that multi-scale target perception is difficult. Extracting multi-scale features by adopting hole convolution with different hole rates, and carrying out feature fusion on the multi-scale features from bottom to top and from top to bottom to enhance the scale perception capability of the network; meanwhile, for the problem that a complex background is misjudged as a target, channel attention branches are designed, global attention information is extracted, extraction of multi-scale features is supervised, finally, the multi-scale features are regressed into a final density map through combination of convolution and activation functions, crowd density estimation of an unconstrained scene can be carried out, and crowd density estimation accuracy is improved. The method can also be used for density estimation of various dense targets such as logs and cells.

Description

technical field [0001] The invention belongs to the technical field of crowd density estimation, and relates to a crowd density estimation method based on pyramid feature fusion. Background technique [0002] With the increasing progress of the world's medical technology, the population of various countries is generally increasing. At the same time, due to the continuous advancement of urbanization, the urban population is increasing day by day, and it appears more and more frequently in public places such as performance venues, tourist resorts, commercial streets, airports, and railway stations. Large-scale crowds gather. In areas with high crowd density, various security accidents are prone to occur, which brings great challenges to security work. Crowd density estimation can monitor the degree of crowd aggregation in the scene in real time and detect high-density crowds in time. crowd and take evacuation measures. [0003] Early detection-based crowd density estimation m...

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): G06V20/52G06V20/70G06V10/80G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06N3/045G06F18/253
Inventor 刘寒松王国强王永翟贵乾刘瑞焦安健
Owner QINGDAO SONLI SOFTWARE INFORMATION 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