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

Target counting method and system based on double-attention multi-scale cascade network

A multi-scale cascading and target counting technology, which is applied in the field of image processing, can solve problems such as counting deviation, achieve the effect of improving robustness and improving target counting deviation

Active Publication Date: 2019-08-30
YANSHAN UNIV
View PDF2 Cites 51 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a target counting method and system based on a double-attention multi-scale cascaded network, which can effectively solve the problem of counting deviation caused by multi-scale target appearance in the image by extracting the multi-scale features of the image through the spatial attention pyramid structure. Extracting global feature information through the channel attention weighted fusion structure can effectively improve the robustness of the counting method, so as to complete the accurate target counting task

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
  • Target counting method and system based on double-attention multi-scale cascade network
  • Target counting method and system based on double-attention multi-scale cascade network
  • Target counting method and system based on double-attention multi-scale cascade network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0077] figure 1 It is a flow chart of the object counting method based on the dual-attention multi-scale cascaded network of the present invention. Such as figure 1 As shown, an object counting method based on dual-attention multi-scale cascade network, including:

[0078] Step 101: Acquire a grayscale-processed image.

[0079] Step 102: Input the grayscaled image to the initial module of the double-attention multi-scale cascade network for initial feature extraction to obtain an initial feature map.

[0080] The inception module consists of two convolutional layers. Input the grayscaled image into the first convolutional layer in the cascaded network to obtain a first initial feature map; input the first initial feature map into the second convolutional layer in the cascaded network layer to get the initial feature map. The convolution kernel sizes of the first convolution layer and the second convolution layer are both 9×9. The image generates 16 feature maps through t...

Embodiment 2

[0116] In addition to the steps described in Embodiment 1, Embodiment 2 of the present invention also includes after step 108:

[0117] Get loss functions for fully connected layers and feature extraction operations.

[0118] Weighting the loss function of the fully connected layer and the feature extraction operation to obtain the overall loss function of the cascaded network; the loss function of the fully connected layer adopts a cross entropy function, and the loss function of the feature extraction operation is an estimated target The Euclidean distance between the distribution density map and the true target distribution density map.

[0119] An error of the cascaded network is determined from an overall loss function of the cascaded network.

[0120] The error is backpropagated, the weight parameters of the cascaded network are updated, and a trained model for object counting is obtained through multiple iterations.

Embodiment 3

[0122] An object counting system based on a dual-attention multi-scale cascaded network, including:

[0123] The acquisition module is used to acquire the grayscale processed image.

[0124] The first feature extraction module is used to input the image after the grayscale processing to the initial module of the double-attention multi-scale cascade network for initial feature extraction to obtain an initial feature map; the initial module includes two convolutions layer.

[0125] The low-level detail feature map and the high-level semantic feature map determination module are used to input the initial feature map to the first branch network of the double-attention multi-scale cascaded network to obtain the low-level detail feature map and the high-level semantic feature map; The first branch network includes multiple convolutional layers and multiple pooling layers.

[0126] A transformation module, configured to perform channel attention transformation on the low-level deta...

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 target counting method and system based on a double-attention multi-scale cascade network. The target counting method comprises the following steps: inputting an image subjected to graying processing into an initial module of the double-attention multi-scale cascade network for initial feature extraction to obtain an initial feature map; inputting the initial feature mapinto a first branch network of the cascade network to obtain a low-level detail feature map and a high-level semantic feature map; performing channel attention transformation on the feature map to generate global feature information; inputting the initial feature map into a spatial attention pyramid structure of a second branch network in the cascade network to generate multi-scale features of theimage; fusing the multi-scale features and the global feature information to obtain a fusion result; performing feature extraction on the fusion result to generate an estimated target distribution density map; and carrying out pixel summation on the target distribution density map to obtain an estimated target counting result. The target counting method can effectively realize accurate target counting in a complex scene, and has good robustness and generalization.

Description

technical field [0001] The invention relates to the field of image processing, in particular to an object counting method based on a double-attention multi-scale cascade network. Background technique [0002] Object counting is the estimation of the number of objects of a certain class in a scene. The object counting problem is crucial to the process of building high-level cognition in tasks such as scene understanding and visual reasoning. The mature target counting method can be used for real-world applications such as crowd counting and vehicle counting in surveillance video, cell counting under a microscope, and animal and plant counting in open scenes in the wild. [0003] Most of the existing object counting methods are designed for crowd counting and vehicle counting, which is mainly due to the wide application of security monitoring technology. The existing object counting methods mainly adopt the method based on density distribution map estimation, that is, estima...

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): G06K9/00G06K9/46G06K9/62G06N3/04G06N3/08
CPCG06N3/084G06V20/54G06V10/464G06V2201/08G06V2201/07G06N3/045G06F18/2413G06F18/253
Inventor 张世辉李贺孔维航何欢王爽
Owner YANSHAN UNIV
Features
  • Generate Ideas
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More

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

© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com