Unlock instant, AI-driven research and patent intelligence for your innovation.

Image feature detection method, system, device and medium

A technology of image features and detection methods, applied in the field of image processing, can solve problems such as low efficiency and inaccurate image feature detection

Active Publication Date: 2020-08-25
恒睿(重庆)人工智能技术研究院有限公司
View PDF8 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] In view of the shortcomings of the prior art described above, the object of the present invention is to provide an image feature detection method, system, device and medium for solving the technical problems of inaccurate and low efficiency in image feature detection in the prior art

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
  • Image feature detection method, system, device and medium
  • Image feature detection method, system, device and medium
  • Image feature detection method, system, device and medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0083] see figure 1 , the present invention provides an image feature detection method, comprising:

[0084] S1: Perform multiple feature processing on the image information through one or more feature processing structures to obtain output images of various scales, wherein the feature processing structure includes at least one of the following: downsampling unit, feature extraction unit, channel The dimensionality reduction unit, for example, reduces the amount of parameters in the processing process through a feature processing structure to improve the speed of image feature detection, and for example, processes through multiple feature processing structures to extract more image feature information and improve image feature detection precision;

[0085] S2: Output or fuse the output images of multiple scales through one or more fusion processing structures to obtain feature maps of multiple scales, wherein the fusion processing structure includes at least one of the follow...

Embodiment 2

[0101] see Figure 10 , Embodiment 2 provides an image feature detection system, including:

[0102] The first network includes one or more feature processing structures, and performs multiple feature processing on image information through multiple feature processing structures to obtain output images of various scales, wherein the feature processing structures include at least one of the following One: down-sampling unit, feature extraction unit, channel dimensionality reduction unit, for example, reduce the amount of parameters in the processing process through a feature processing structure, improve the speed of image feature detection, and for example, process through a variety of feature processing structures, extract More image feature information to improve image feature detection accuracy;

[0103] The second network includes one or more fusion processing structures, and performs output or fusion processing on the output images of multiple scales through multiple fus...

Embodiment 3

[0123] see Figure 11 , the present embodiment provides an image feature detection system, which includes a first network 31 and a second network 32, wherein the first network 31 includes 12 convolutional layers, and the first network 31 includes 5 first feature Handle structure. The first network 31 down-samples the input image 5 times to output the minimum-scale feature map, and inputs output images of different scales to the second network 32 . The results of the downsampling are fused through the second network to obtain feature maps with three different scales and the number of image channels. In this embodiment, the first network reduces the number of convolutional layers, and the second network uses The feature map with fewer channels completes the subsequent upsampling and feature extraction, which further reduces the amount of calculation parameters and improves the detection speed faster. It can achieve FPS (Frames Per Second)>150 Real-time image feature detection....

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 an image feature detection method, a system, a device and a medium, and the method comprises the steps: carrying out the feature processing of image information for many times through one or more feature processing structures, and obtaining output images of various scales; and carrying out output or fusion processing on the output images of the multiple scales through one ormore fusion processing structures to obtain feature maps of the multiple scales. The receptive field is expanded in the down-sampling process, and more image feature information is extracted; characteristic maps of multiple scales are extracted through multiple fusion processing, the characteristic map with a large scale has a small receptive field and corresponds to a small target, and the characteristic map with a small scale has a large receptive field and corresponds to a large target, so that the model is more robust to targets with wide scale and proportional distribution.

Description

technical field [0001] The present invention relates to image processing technology, in particular to an image feature detection method, system, equipment and medium. Background technique [0002] In order to facilitate the recognition and detection of images, it is necessary to recognize and detect the targets in the images. The existing target detection methods have high requirements for hardware and are difficult to meet the needs of real-time detection. In order to reduce the dependence on hardware configuration, it is possible resulting in lower detection accuracy. For example, in some scene conditions, when the target size distribution is wide, if the detection model is simplified and the parameters are not fully utilized, there may be errors in the extraction of target feature information. If the detection model network is complex, the calculation speed of the model may be too slow , it is not convenient to achieve the purpose of real-time detection. Contents of th...

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): G06K9/62
CPCG06F18/21
Inventor 姚志强周曦周依梦
Owner 恒睿(重庆)人工智能技术研究院有限公司