Single Image Crowd Counting Algorithm Based on Multi-column Convolutional Neural Network

A convolutional neural network and crowd counting technology, applied in computing, computer components, instruments, etc., can solve problems such as the ineffective application of crowd information processing, low image counting accuracy, and large changes in the number of people

CN105528589BActive Publication Date: 2019-01-01SHANGHAI TECH UNIV
4 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Publication Date
2019-01-01

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The present invention provides a single image crowd counting algorithm based on a multi-column convolutional neural network, wherein the multi-column convolutional neural network has three sub-networks, each sub-network uses a different size of the convolution kernel, and the input of each sub-network For the same image, after four convolutions and two poolings, the feature maps output by the three sub-networks are linked together in the "channel" dimension, and then a 1×1 kernel convolution is used to obtain the density map of the crowd . The crowd density obtained by the invention is better than the existing algorithm.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention relates to an algorithm for accurate crowd counting or crowd density estimation based on a single image. Background technique

[0002] Existing crowd counting algorithms have many limitations such as large dependence on image segmentation technology, small counting scale, and fixed input image size. Many algorithms have low counting accuracy for images with large changes in the number of people or images with complex backgrounds. Today, outdoor squares and streets are basically equipped with cameras, but crowd information processing has not been effectively applied. Therefore, accurate crowd counting or crowd density estimation algorithms are of great significance for the detection of abnormal crowd events in monitoring. Contents of the invention

[0003] The purpose of the present invention is to provide an algorithm for accurate crowd counting or crowd density estimation based on a single image.

[0004] In order to achieve the above...

Examples

Embodiment Construction

[0024] In order to make the present invention more comprehensible, preferred embodiments are described in detail below with accompanying drawings.

[0025] The present invention needs to solve a given image of a crowd or a frame in a video, and then estimate the density and total number of people in each area of ​​the image.

[0026] It is known that the input image can be represented as an m×n matrix: x∈R m×n , then the actual crowd density corresponding to the input image x can be expressed as: In the formula: N is the number of people in the image, Indicates the position of each pixel in the image, x i is the position of the ith head in the image, δ( ) is the unit impact function, * is the convolution operation, is the standard deviation σ i Gaussian kernel. The goal of the single image crowd counting algorithm based on multi-column convolutional neural network is to learn a crowd density from the input image x to the image (such as figure 2 The mapping function F...