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

Camouflage object detection method based on boundary alternate guidance

A technology of object detection and boundary, applied in neural learning methods, biological neural network models, image data processing, etc., can solve problems such as blurred boundaries, wrong target areas in model learning, lack of boundaries, etc., to improve accuracy, improve precision, The effect of expanding the receptive field

Pending Publication Date: 2022-03-22
YANGZHOU UNIV
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Among them, the method with better effect is the boundary fusion method, but most of the existing target object detection models use boundary fusion to extract the target object (such as EGNet, SCRN, etc.), but this type of method still has the following disadvantages: the lack of area for the boundary Further refinement may lead to blurred and missing boundaries, leading the model to learn the wrong target area

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
  • Camouflage object detection method based on boundary alternate guidance
  • Camouflage object detection method based on boundary alternate guidance
  • Camouflage object detection method based on boundary alternate guidance

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0091] refer to Figure 1~5 , which is an embodiment of the present invention, provides a method for detecting camouflaged objects based on alternate boundary guidance, including:

[0092] S1: Construct a camouflage detection model based on alternate boundary guidance. On the basis of the backbone network, add and use the initial positioning module to extract the multi-scale deep features of the backbone network, and obtain a rough positioning map of the camouflaged object;

[0093] Furthermore, the backbone network adopts ResNet-34 network.

[0094] Furthermore, the initial positioning module includes:

[0095] Such as figure 1 , remove the last fully connected layer in the ResNet-34 network, and output S in the deep layer 2 -S 5 ;

[0096] Such as figure 2 , the initial positioning module outputs the fifth layer measurement output S 5 Enter the average pooling branch and the maximum pooling branch respectively;

[0097] S in the average pooling branch 5 After the g...

Embodiment 2

[0170] With reference to Table 1, it is another embodiment of the present invention. In order to verify and illustrate the technical effect adopted in the method, the present embodiment adopts the traditional technical scheme and the method of the present invention to carry out comparative tests, and compares the test results with the means of scientific demonstration, to verify The real effect of this method.

[0171] In Table 1, the first column is the method model proposed in recent years, the subscript indicates the year proposed, such as NLDF17 indicates the method proposed in 2017; the second column indicates the backbone network of the model, and the complexity is ResNet-50>ResNet -34>VGG-16; the first line indicates the name of the dataset, such as CAMO, CHAMELEON, COD10K and NC4K, where CAMO and CHAMELEON are small datasets with only 250 and 76 pictures respectively, while COD10K and NC4K are large datasets, respectively By 2026, there are 4121 pictures; the second li...

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 camouflage object detection method based on boundary alternate guidance, which comprises the following steps: constructing a camouflage detection model based on boundary alternate guidance, adding an initial positioning module to extract multi-scale deep features of a backbone network, and obtaining a rough positioning map of a camouflage object; a multi-scale receptive field module is added to enlarge the receptive field of the rough positioning map and improve the semantic information extraction capability; a boundary alternating guide module is added to extract a region graph and a boundary graph layer by layer, and constraint refinement is carried out on the region graph and the boundary graph to obtain an accurate camouflage object prediction graph; according to the method, the camouflage object can be accurately positioned under similar backgrounds, a smoother camouflage object prediction image can be obtained by utilizing boundary constraint based on a boundary alternate guide structure, a continuous and clear boundary image can also be obtained by utilizing the prediction image constraint boundary, and the camouflage object detection precision is effectively improved.

Description

technical field [0001] The invention belongs to the field of computer vision and digital image processing, in particular to a detection method for a camouflaged object based on alternate boundary guidance. Background technique [0002] In nature, camouflage is an important skill that allows creatures to blend with the background and avoid being attacked by natural enemies. Camouflaged object detection aims to identify objects that closely resemble their surroundings. The low contrast between camouflaged objects and the background can greatly deceive the human visual system, therefore, camouflaged object detection is more challenging than traditional saliency detection. [0003] In recent years, many algorithms based on color and texture features have been proposed for camouflaged object detection. When the color of the camouflage object is similar to that of the background, the texture feature is used to distinguish the camouflage object from the background; when the textu...

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): G06V20/10G06V10/82G06T7/13G06T5/00G06N3/04G06N3/08
CPCG06T7/13G06N3/084G06T2207/20192G06N3/047G06N3/048G06N3/045G06T5/73
Inventor 俞锦豪陈舒涵徐秀奇陆露陈泽宇
Owner YANGZHOU 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