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

An image classification method based on adaptive coding and geometric smooth fusion

A classification method and adaptive technology, applied in character and pattern recognition, instruments, computer parts, etc., can solve the problems of ignoring the geometric distribution of significant coding features, not making full use of spatial geometric position information, and loss of image spatial information.

Active Publication Date: 2017-04-26
NANJING UNIV
View PDF6 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the LLC method generates codebooks through the standard K-means algorithm. When the data set is very large, the codebook training time complexity is relatively high. In addition, most of the above-mentioned coding methods only consider the significant feature information in the coding process. while ignoring the geometric distribution of local features around salient coding features
Therefore, the feature descriptor only retains the maximum response value of each codeword, and does not make full use of the spatial geometric position information of the local features around the salient features, resulting in the loss of image spatial information

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
  • An image classification method based on adaptive coding and geometric smooth fusion
  • An image classification method based on adaptive coding and geometric smooth fusion
  • An image classification method based on adaptive coding and geometric smooth fusion

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0098] This embodiment includes the following parts:

[0099] 1. Extract the underlying local features of the image:

[0100] Scale-invariant feature transform (SIFT) is used to extract local features in the image. In this module, the scale-invariant feature transform is mainly used.

[0101] The Scale Invariant Feature Transformation (SIFT) feature is a computer vision algorithm used to detect and describe local features in images. It looks for extreme points in the spatial scale and extracts its position, scale, and rotation invariants. , this algorithm was published by David Lowe in 1999 and perfected in 2004.

[0102] 2. Code book learning:

[0103] Using the fast k-means method for codebook training on the bottom layer local features obtained by sampling, it is expected to learn a complete codebook. This module mainly includes two steps: initializing the codebook and updating the central features.

[0104] Initialization codebook: d-dimensional local underlying featur...

Embodiment 2

[0132] image 3 is an example image sourced from a database of 15 scene classes. Figure 4 For the classification accuracy rate on 15 scene classes with the inventive method, Figure 5 Comparison of classification accuracy on the Caltech-101 database for different feature pooling strategies.

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 an image classification method based on adaptability coding and geometrical smooth convergence. The image classification method based on adaptability coding and geometrical smooth convergence includes the steps of partial feature extraction of an image bottom layer, study of a code book, feature coding, feature convergence, segmentation of training data, building of a model and image classification. The image classification method is applied to the field of image classification, can keep structural information of the code book and feature coding on the aspect of image representation, can largely reduce the time complexity of code book generation, generates image feature representation with structural features, makes full use of the space position information rich in image features and has an obvious effect on the aspect of image classification, thereby having high use value.

Description

technical field [0001] The invention belongs to the field of image classification, and is an image classification method based on adaptive coding and geometric smooth fusion. Background technique [0002] Image classification tasks mainly include object and scene classification, which is one of the important research fields in the field of computer vision and pattern recognition. In recent years, because some local features can reveal unique information in the image, they are widely used to represent images, such as SIFT features and HOG features. These features mainly represent the image content through some small, possibly overlapping and independent local blocks. Due to computational complexity and sensitivity to noise, these low-level local features are not directly used for image classification. A general strategy is to encode these features into a global image feature representation, so the codebook-based model (BoW model) and its extension methods are proposed and s...

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): G06K9/66G06K9/46
Inventor 杨育彬王喆正毛晓蛟李亚楠
Owner NANJING UNIV