Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Population counting method based on sub-module technology and semi-supervised learning

A semi-supervised learning and crowd counting technology, applied in the field of image and video-based crowd counting, which can solve problems such as inability to select samples, redundant information, and failure to consider the amount of information.

Active Publication Date: 2017-06-20
FUDAN UNIV
View PDF7 Cites 10 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, this method only avoids the information redundancy between samples, but it does not take into account that the amount of information contained in different samples is also different, so it is impossible to select the most representative sample from each cluster.

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 counting method based on sub-module technology and semi-supervised learning
  • Population counting method based on sub-module technology and semi-supervised learning
  • Population counting method based on sub-module technology and semi-supervised learning

Examples

Experimental program
Comparison scheme
Effect test

experiment example 1

[0072] Experimental Example 1: Regression Algorithm Prediction Performance

[0073] This part of the experiment shows the error of different regression methods when randomly choosing labeled samples. As comparison methods, we choose Gaussian process regression with supervised learning, semi-supervised elastic net based and semi-supervised ridge regression. Table 1 shows the comparison of the regression algorithm of the present invention and the three comparison methods on the three datasets. It can be seen that on the UCSD and Fudan datasets, the performance of the present invention is the best, and on the Mall dataset, the performance of the present invention is very close to that of the semi-supervised elastic net.

experiment example 2

[0074] Experimental example 2: The impact of choosing different similarity measures on the model

[0075] Table 2 shows the performance changes of the regression model on the three data sets when using different similarity measures (ie, the weight ∈ of temporal similarity takes different values). It can be seen that on all datasets, using unlabeled samples can improve the prediction performance of the model; when using different similarity measures, it will have different effects on performance. If both temporal and spatial similarity are used (weight ∈ via Cross-validation obtained), the model can obtain the lowest prediction error.

experiment example 3

[0076] Experimental Example 3: Subplot Sample Selection Algorithm

[0077] Table 3 shows the effect of the subtype sample selection algorithm of the present invention. For comparison, k-means clustering and m-center point methods were selected. The k-means clustering method firstly clusters all samples, and then randomly selects samples from each cluster; the m-center point method first calculates the Laplacian matrix of the samples, and then obtains multiple clusters through spectral clustering, Finally select center point samples from each cluster.

[0078] Compared with Table 1, it can be found that for the UCSD data set, the samples selected by the semi-supervised elastic net and the sub-model technology of the present invention greatly improve the model performance; for the rest of the data sets, only the sub-model technology of the present invention can significantly improve the performance. In particular, the Mall dataset illustrates the applicability of the present i...

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, which belongs to the field of computer vision and machine learning, specifically relates to a population counting method based on a sub-module technology and semi-supervised learning. With the population counting method, a problem of shortage of samples for population counting during a model training process can be solved. For continuous high-frequency population monitoring images, pretreatment is carried out on each frame of image and features like a population area and a texture are extracted; for an image sample set, a few of optimal samples are extracted by using a sub-module sample selection algorithm and then artificial marking is carried out; and then a model is trained by using a semi-supervised regression algorithm and the model performance is improved by using lots of unmarked samples. Compared with random-sampling cluster-based sample selection method, the population counting method enables the training sample number to be reduced obviously and the prediction performance of the model to be improved.

Description

technical field [0001] The invention belongs to the technical fields of computer vision, machine learning, and intelligent transportation, and particularly relates to a crowd counting method based on images and videos. Background technique [0002] The problem of crowd counting based on video images is one of the research problems in the field of computer vision and machine learning. When a crowd image in a scene is given, it is required to output the number of pedestrians according to a computer vision or machine learning algorithm. At present, there have been many previous works in this field, and the main methods can be divided into two categories: methods based on pedestrian detection and regression methods based on image features. Here are some references for both types of methods: [0003] [1] A.B.Chan, Z.J.Liang, and N.Vasconcelos.Privacy pre-serving crowdmonitoring: Counting people without people models or tracking.In Conference on Computer Vision and Pattern Recog...

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): G06K9/00G06K9/62
CPCG06V20/53G06F18/23213
Inventor 周齐张军平
Owner FUDAN UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com