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

Method and device for feature detection

A feature detection and feature value technology, applied in the field of image processing, can solve the problems of feature dislocation and low reliability.

Active Publication Date: 2020-07-28
ZHEJIANG UNIVIEW TECH CO LTD
View PDF7 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] Since the above method is directly based on the original video image to generate the corresponding high-dimensional feature vector, its reliability is low in the case of rotation, translation and large noise
On the other hand, due to the direct use of cascading methods to generate high-dimensional feature vectors, feature misalignment will occur during feature detection

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
  • Method and device for feature detection
  • Method and device for feature detection
  • Method and device for feature detection

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0037] Aiming at the problems existing in the prior art, an embodiment of the present invention proposes a method for feature detection, which can be applied to a system for feature detection in a Fisher Vector (Fisher Vector) manner. Among them, the FisherVector method is a feature extraction method based on a mixed Gaussian model. By using K Gaussian kernels to simulate the distribution of local features for video images, local features can be effectively fused, and it has strong reliability for changes in video images. Therefore, the Fisher Vector method is an effective feature encoding method. However, when the Fisher Vector method generates feature vectors, it needs to use more Gaussian kernels to generate a GMM (Gaussian Mixture Model, Gaussian mixture model) model. For example, K Gaussian kernels are usually required, and the value of K is generally 256- 512. Due to the large K value, the calculation complexity of generating the GMM model is high, the calculation speed ...

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 present invention provides a method and device for feature detection. For each video image to be detected, the method includes: performing feature extraction on the video image to obtain M feature values; performing N times sampling on the M feature values ​​to obtain N sampled results, each sampled result contains some eigenvalues ​​of M eigenvalues; for each sampled result, k Gaussian kernels are used for the sampled result, the GMM sub-model corresponding to the sampled result is generated, and N GMM sub-models; after sorting the N GMM sub-models, a corresponding GMM model is obtained, and a feature vector corresponding to the GMM model is obtained, and feature detection is performed using the feature vector. Through the technical solution of the present invention, fewer k Gaussian kernels can be used to generate the corresponding GMM sub-model, thereby reducing the computational complexity of the feature detection algorithm, improving the computational performance, improving the convergence of the algorithm, and effectively accelerating the generation of feature vectors.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a method and device for feature detection. Background technique [0002] Feature detection has a wide range of applications, such as video analysis, object detection, image recognition, etc. Typically, the process of feature detection includes: dividing a video image into several sub-regions, performing feature extraction on each sub-region, obtaining the feature value of each sub-region, and finally cascading the extracted feature values ​​to generate a set of high-dimensional A feature vector is used, and the video image is represented by the high-dimensional feature vector, so that the feature detection of the video image is performed based on the high-dimensional feature vector of the video image. [0003] Since the above method generates a corresponding high-dimensional feature vector directly based on the original video image, its reliability is low in the...

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 Patents(China)
IPC IPC(8): G06K9/46
CPCG06V10/50
Inventor 毛敏
Owner ZHEJIANG UNIVIEW TECH CO LTD