Eureka AIR delivers breakthrough ideas for toughest innovation challenges, trusted by R&D personnel around the world.

Dense group counting and positioning method based on distance transformation label

A technology of distance transformation and positioning method, which is applied in the field of computer vision and can solve the problems of unable to provide positioning information and unable to apply to dense group scenes.

Active Publication Date: 2020-04-24
HUAZHONG UNIV OF SCI & TECH
View PDF4 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The present invention proposes a dense group counting and positioning method based on distance transformation labels, which can provide specific location information of each head while ensuring counting accuracy, so as to solve the problem that the method based on regression density map in dense group scenes cannot provide Accurate positioning information, the problem that detection-based methods cannot be applied to dense crowd scenes

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
  • Dense group counting and positioning method based on distance transformation label
  • Dense group counting and positioning method based on distance transformation label
  • Dense group counting and positioning method based on distance transformation label

Examples

Experimental program
Comparison scheme
Effect test

example 1

[0069] (1) Make a data set. For each head, due to the innovativeness of this method, only the center coordinates of each head need to be provided, and no marking box for each head is required. Compared with the detection-based method, the labor marking cost is greatly reduced. Online expansion of the prepared data set, the specific operation is random horizontal flip, random scaling between 0.8-1.5 times, and random superimposition of salt and pepper noise. The number of noise points of salt and pepper noise is 0.001-0.015 times the number of picture pixels.

[0070] (2) Generate Binary map. Set up a two-dimensional zero matrix in advance. For a given center coordinate (x, y) of each human head, use 1 to represent the point on the matrix according to the coordinate position, such as figure 1 Shown represents the original picture, figure 2 The Binary map generated by the coordinates of the center of each human head is displayed. It should be pointed out that, for intuitive dis...

example 2

[0103] According to the method proposed by the present invention, the performance obtained on three common public data sets is given, specifically including ShanghaiTech_partA, ShanghaiTech_partB and UCF-QNRF.

[0104] The ShanghaiTech_partA data set contains 482 pictures, including 300 training pictures and 182 test pictures, which are mainly crawled from the Internet. The density of the data set is relatively moderate, with an average number of 501.

[0105] The ShanghaiTech_partB data set contains 716 pictures, 400 of which are used for training and 316 are used for testing. The data set is mainly from the streets of Shanghai. The density of the data set is relatively sparse, and the average number of people is 123.

[0106] UCF-QNRF contains 1535 pictures, of which 1201 are used for training and 334 are used for testing. The data set is very dense and the average number of people is 815, and the maximum number of people has reached 12,865.

[0107] The evaluation indicators used ar...

example 3

[0121] Expandability. This method can not only be used to complete the counting and positioning of people, but also to complete the counting and positioning of other objects, such as vehicles and cells. Here, the present invention uses the disclosed vehicle data set trancos as an example to illustrate its capabilities in vehicle counting and positioning.

[0122] First, the generated Distance label map is also used as the real label D. Since the data set provides a mask label, which is the region of interest, the input image dot multiplication mask is used as the input, and the output dot multiplication mask is also used. Using the same FPN structure as crowd counting and positioning, the loss function remains unchanged, and the test results after training are shown Picture 9 , Where GT represents the number of vehicles in the real label, and Count represents the number of predicted vehicles, showing excellent counting and positioning accuracy.

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 dense group counting and positioning method based on a distance transformation label. According to the invention, a novel real label generation mode is provided; the real label is named as a Distance label map. Through Distance label map, the estimated number of people can be obtained only by counting the number of the local minimum value areas, and besides, the specificposition information of the predicted human head can be obtained by calculating the center coordinates of the local minimum value areas. A feature pyramid network is adopted for feature extraction, and the picture is input into the network to obtain a plurality of different fusion features, and loss is calculated with the real label and summed; according to the method, a new function for calculating the loss is provided, the function is called as the self-adaptive cross entropy, the label of the pixel point can be used for representing the distance information between the pixel point and the nearest pixel point, and the traditional cross entropy is subjected to weighted improvement. Compared with the related work of the previous group counting, the positioning information of each head canbe provided while the counting accuracy is ensured.

Description

Technical field [0001] The invention relates to the field of computer vision, in particular to a method for counting and positioning dense groups based on distance transform tags. Background technique [0002] In recent years, there have been frequent stampede incidents when holding large-scale events at home and abroad. The main reason is that the specific number and location of the crowd cannot be obtained in time, and the flow of people cannot be effectively evacuated. This makes crowd counting work in video surveillance, crowd understanding, and public safety. Prevention and other fields have gradually played an important role, and have become a hot research issue in the field of computer vision. [0003] With the development of deep learning, researchers can use deep neural networks to achieve dense crowd counting. The existing crowd counting methods can generally be divided into two categories. The first is to use crowd counting as a target detection task, and view pedestrian...

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
IPC IPC(8): G06K9/00G06K9/62
CPCG06V20/41G06V20/53G06F18/214
Inventor 许永超徐晨丰梁定康白翔
Owner HUAZHONG UNIV OF SCI & 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
Eureka Blog
Learn More
PatSnap group products