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

SURF improved method based on adaptive fractional-order differentiation

A fractional differential and adaptive technology, applied in the field of image processing, can solve the problem of unsatisfactory texture description in smooth areas

Inactive Publication Date: 2016-06-15
SUZHOU UNIV OF SCI & TECH
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

General image descriptors usually use integer-order differences (such as direction histograms, etc.) to describe local texture features, but in fact integer-order differences are not ideal for texture descriptions in smooth areas

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
  • SURF improved method based on adaptive fractional-order differentiation
  • SURF improved method based on adaptive fractional-order differentiation
  • SURF improved method based on adaptive fractional-order differentiation

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0034] This embodiment discloses an improved method based on adaptive fractional differential SURF. The dimension of the feature descriptor has the most direct impact on the real-time application of subsequent point features, and the feature descriptor with a lower dimension can quickly match the feature points. is more ideal. However, the uniqueness of low-dimensional descriptors, that is, the degree of recognition, is not as good as high-dimensional feature descriptors (such as SIFT descriptors, etc.). While SpeedUpRobustFeatures (SURF) ensures the accuracy, high recognition and uniqueness of the descriptor, it effectively reduces its dimension and meets the real-time requirements.

[0035] In this embodiment, the SURF descriptor uses the integral image as the basis for subsequent feature point representation, and uses the Haar wavelet response in the moving fan-shaped area when calculating the main direction. The response in the horizontal direction is shown in formula (1),...

Embodiment 2

[0061] In order to illustrate the effectiveness of the algorithm in Example 1, after adding noise in Example 2, a comparative analysis is performed, and Gaussian noise σ=0.02 is added to the test image as the image to be matched. The specific experimental data comparison is given in Table 1:

[0062] In Table 1 Example 2, the comparison results after adding noise

[0063] local feature descriptor number of matches correct match rate SURF 627 96.0％ Improved SURF 694 96.7％

[0064] It can be seen from the results in Table 1 that the improved SURF algorithm feature descriptor of the present invention, compared with the original algorithm, can better describe the texture features of the smooth area, so more point pairs with the same name can be obtained. And improve the correctness of the final matching.

Embodiment 3

[0066] In order to improve the robustness of the descriptor, it is necessary to consider the influence of the illumination on the feature descriptor. In the method of the embodiment 1, the influence of the illumination change on the result is added in this embodiment. Due to the influence of camera saturation, non-linear lighting changes often exist in the image, which makes it difficult to locate texture key points in smooth areas, especially in dark areas.

[0067] Therefore, in order to effectively match and apply subsequent feature points, this embodiment performs preprocessing on the image to be matched, and through the algorithm of the present invention, the image to be matched is enhanced, especially the texture enhancement of the smooth area or the unobviously illuminated area.

[0068] The data contrast result of table 2 embodiment 3

[0069]

[0070] It can be seen from Table 2 that after preprocessing by the algorithm of the present invention, the number of key p...

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 relates to an SURF improved method based on fractional-order differentiation. An SURF descriptor takes an integral image as the basis of subsequent feature point representation, Haar wavelet response in a mobile sector area is utilized in calculation of the principal direction, the horizontal response is shown by an expression (1): Vx=([L<integral>(x, y+1)-L<integral>(x, y)]+[L<integral>(x+1, y+1)-L<integral>(x+1, y)])*wd, and the vertical response is shown by an expression (2): Vy=([L<integral>(x+1, y)-L<integral>(x, y)]+[L<integral>(x+1, y+1)- L<integral>(x, y+1)])*wd, wherein L<integral>(x, y) represents the numerical value of an integral image L(x, y) in a position (x, y), and wd represents the Gaussian weight; and the horizontal response and vertical response are recalculated through fractional-order differencing, wherein the horizontal response is shown by an expression (3) (as shown in the description), and the vertical response is shown by an expression (4) (as shown in the description). The SURF improved method reduces the dimension effectively and meets the requirement for real-time performance while ensuring the accuracy, high degree of recognition and uniqueness of the descriptor. Under the conditions of noise, rotation and light change, the number of detected feature points and the number of matched points are greater than those of the traditional SURF, and the accuracy rate of matching is improved.

Description

technical field [0001] The invention belongs to the field of image processing, and in particular relates to an improved method based on adaptive fractional differential SURF. Background technique [0002] As one of the representations of images, image features are widely used in the field of image processing. The representation of image features can be divided into: point features, line features and surface features, among which point features are the most widely used in feature representation; and artificial design-based Point features can be divided into four categories: moment feature, difference feature, space-frequency domain feature and space domain distribution feature, among which space-frequency domain feature and space domain distribution feature are widely used. [0003] Lowe proposed a scale-invariant feature extraction algorithm (SIFT), using DOG to fit fine models in each scale space, screening and determining key points; Irrelevant features such as scaling, r...

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): G06K9/46
CPCG06V10/462
Inventor 胡伏原李林燕姒绍辉付保川李宏
Owner SUZHOU UNIV OF SCI & TECH