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

Crowd counting method and system based on multi-scale feature information

A multi-scale feature and crowd counting technology, applied in the field of computer vision, can solve the problems of poor algorithm's ability to count small objects, loss, and multi-detail information, so as to improve accuracy, good accuracy and robustness, and improve perception. effect of effect

Pending Publication Date: 2020-08-04
SHANDONG NORMAL UNIV
View PDF3 Cites 12 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] However, the inventors of the present disclosure found that the existing multi-column CNN generally uses the highest-level feature map to regress to generate a density map, and the high-level feature map will be lost after being abstracted layer by layer and down-sampled by the pooling layer. More detailed information even filters out some small-scale targets, resulting in poor counting ability of the algorithm for small targets

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
  • Crowd counting method and system based on multi-scale feature information
  • Crowd counting method and system based on multi-scale feature information
  • Crowd counting method and system based on multi-scale feature information

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0035] Such as figure 1 As shown, Embodiment 1 of the present disclosure provides a crowd counting method based on multi-scale feature information, including the following steps:

[0036] Step 1: Image preprocessing

[0037] Convolve the image data set with head position annotation through a two-dimensional Gaussian convolution kernel to generate the crowd density map label corresponding to each image in the data set, and form a training sample set.

[0038] In the crowd counting task, the convolutional neural network needs to be trained. This implementation chooses to use the density map as the data label. Since the crowd counting database only provides the coordinate points marked by the head, it is necessary to generate the density map of the training picture before the network training.

[0039] Then the generation density map equation can be expressed as:

[0040]

[0041] Among them, N represents the number of people in the crowd image, x represents the position of ...

Embodiment 2

[0065] Embodiment 2 of the present disclosure provides a crowd counting system based on multi-scale feature information.

[0066] A crowd counting system based on multi-scale feature information, characterized in that it includes:

[0067] The data preprocessing module is configured to: preprocess the acquired image to obtain a crowd density map corresponding to the image;

[0068] The feature extraction module is configured to: input the obtained crowd density map into the preset multi-feature extraction network model, calculate layer by layer to obtain multiple feature maps with gradually reduced scales, and reverse the obtained feature maps with multiple scales. Connect to the layer-by-layer side to get multiple connected feature maps;

[0069] The crowd counting module is configured to: perform feature fusion on the obtained multiple connected feature maps to obtain multi-scale feature maps, perform density map regression to obtain the final crowd density map, and then ob...

Embodiment 3

[0072] Embodiment 3 of the present disclosure provides a medium on which a program is stored. When the program is executed by a processor, the steps in the crowd counting method based on multi-scale feature information as described in Embodiment 1 of the present disclosure are implemented.

[0073] The steps are specifically:

[0074] Preprocessing the acquired image to obtain the crowd density map corresponding to the image;

[0075] Input the obtained crowd density map into the preset multi-feature extraction network model, calculate layer by layer to obtain multiple feature maps with gradually decreasing scales, and perform reverse side-by-side connection of the obtained multi-scale feature maps to obtain Multiple connected feature maps;

[0076] The feature fusion of the obtained multiple connected feature maps is performed to obtain a multi-scale feature map, and the final crowd density map is obtained after density map regression, and then the final crowd count value is...

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 crowd counting method based on multi-scale feature information, and belongs to the technical field of computer vision, and the method comprises the steps: carrying out the preprocessing of an obtained image, and obtaining a crowd density map corresponding to the image; inputting the obtained crowd density map into a preset multi-feature extraction network model, performinglayer-by-layer calculation to obtain a plurality of feature maps of which the scales are gradually reduced, and performing reverse layer-by-layer side edge connection on the obtained feature maps ofthe scales to obtain a plurality of connected feature maps; performing feature fusion on the plurality of obtained connected feature maps to obtain a multi-scale feature map, performing density map regression to obtain a final crowd density map, and further obtaining a final crowd counting value. According to the method and the device, the influence of large-scale change of crowds on crowd counting is solved, the loss caused by detail information loss after layer-by-layer abstract expression and down-sampling of a pooling layer is further reduced by adopting a feature map fusion mode, and moredetail information can be reserved, so counting of multi-scale dense crowds with a relatively good effect is realized.

Description

technical field [0001] The present disclosure relates to the technical field of computer vision, in particular to a crowd counting method and system based on multi-scale feature information. Background technique [0002] The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art. [0003] Focusing on the current era theme of peace and development, the world today has entered a stage of rapid peaceful development. The population has grown exponentially, and the resulting dense crowd gatherings in public places have become more and more frequent, such as large-scale concerts, sports competitions, and large-scale crowd gatherings and parades that are commonplace today. Under such circumstances, crowd stampedes and riots have become more and more frequent. For example, the stampede on the Bund in Shanghai in 2015 has reached the level of major casualty accidents stipulated by our country. ...

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): G06K9/00G06K9/46G06K9/62G06N3/04G06N3/08
CPCG06N3/08G06V20/53G06V10/40G06N3/045G06F18/253
Inventor 吕蕾谢锦阳顾玲玉陈梓铭
Owner SHANDONG NORMAL 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