Radar remote sensing image retrieval algorithm based on improved Sobel operator

https://doi.org/10.1016/j.jvcir.2019.102720Get rights and content

Abstract

Aiming at the time-consuming problem caused by large computational load of radar image retrieval, based on blocking histogram, Sobel edge detection operator and gray level co-occurrence matrix (GLCCM), new radar remote sensing image retrieval algorithm based on improved Sobel operator is proposed. Firstly, the Sobel edge detection algorithm is used to process the image, the edge image is acquired, the radar remote sensing image is analyzed from different angles, and then the different radar remote sensing images are transformed. Then, based on the above processing, Radar Remote Sensing Image Retrieval Algorithm is acquired; finally, the plurality of statistic of the matrix is recorded as a feature vector describing the radar image, and the image is retrieved according to the feature vector of the radar image. Through a large number of experiments, Radar Remote Sensing Image Retrieval algorithm can greatly reduce the retrieval time, and it also has a good retrieval effect for images with rich texture.

Introduction

With the wide spread of information media technology, the current methods and methods of information processing have been further developed. Nowadays, we are entering an era of rapid development of information technology. With the expansion of the research field, people gradually realized the importance of using multimedia information, and proposed more solutions [1]. How to quickly and accurately organize, manage and retrieve massive remote sensing image data, and to meet the needs of users for fast browsing and query of images of interest, has become the focus of researchers at home and abroad. The key of Content-based image retrieval (CBIR) [2] provides a new opportunity to solve the problem of information extraction and sharing in image retrieval, and has been widely studied and applied as a mainstream method. The traditional CBIR system mainly searches for color, texture, shape of image [3], [4].

Considering image content retrieval can be divided into multiple detection methods according to different forms of content [5], [6]. Effectively combines the texture characteristics of the image to illustrate the specific detection method. In the single color-based retrieval method, the color histogram method has the characteristics of fast operation speed and low storage space requirement, and has the characteristics of image scale and rotation insensitivity [7]. However, it loses the spatial distribution information of the image during the retrieval. Therefore, some methods have been proposed to make up for this deficiency. For example, in the literature [8], [9], the image is uniformly segmented first, and the histogram method is used to obtain the color feature. Although the spatial information of the image is compensated, people generally are interested in the intermediate target area of the image. In view of this paper, based on the literature, a method of overlapping partitioning is proposed and obtained. Better retrieval results.

Section snippets

Color space conversion and quantization

Compared with the color space model HSV and RGB, the former is more intuitive and closer to people's subjective consciousness of color [10], [11], [12], [13], [14], [15], [16], [17], so this paper first converts RGB into HSV color model, and then performs related operations. R, G, and B values (both in the interval [0, 255]) at any point in the RGB space can be converted into the HSV color space to obtain the corresponding H (color) Degree), S (saturation), V (brightness) values, thus the

Improve Sobel operators and transforms

The following is a redefinition of Sobel edge detection as follows:F(m,n)=fm1(m,n)2+fn1(m,n)2

The gradient q is calculated using the following formula.q=arctanfn1fm1

The operation theory of the Sobel operator is the sum of the gray values around any pixel of the image, and τ is determined accordingly. If F(m,n)>τ, the pixel F(m,n) is considered to be the edge point.

The Sobel edge detection operator is very easy to implement in space, which is based on the method of summing the pixel points of the

GLCCM model

A pair of pixels in the θ direction and a pixel distance d is represented by a GLCCM, and the gray values are frequency of i and j, respectively, and the element is denoted by P(i,jd,θ). In the case, where θ and d are constant, a brief is Pij. It can be seen that the matrix is symmetrical, and the order size is determined by gray levels number. The method for solving the element value sees:P(i,jd,θ)=P(i,jd,θ)ijP(i,jd,θ)

GLCCM method generally constructs the feature matrix in four directions of

Experimental analysis

The test picture in this paper consists of 1000 RGB color images in the radar image library, from UC Merced land-use dataset [12], which is a common data set for radar remote sensing image scene classification, including 21 categories of scene images. For example: airplanes, beaches, forests, buildings, etc. There are 100 images in each category, and the size of the image is 256×256; the resolution of the image is 1 foot. The image is accessed in JPEG file format in order to demonstrate the

Conclusion

The radar remote sensing image retrieval algorithm based on improved Sobel operator is described. The method is systematically described. It can be known that when this method is used to acquire color features, the calculation amount of the similarity measure of the image is reduced. At the same time, this method does not artificially set the weight to achieve the retrieval of the intermediate target area, which reduces the influence of human factors on the retrieval, and does not destroy the

Author Contributions

(1) Based on the above processing, Radar Remote Sensing Image Retrieval Algorithm is acquired.

(2) Considering image content retrieval can be divided into multiple detection methods according to different forms of content.

Declaration of Competing Interest

There is no conflict of interest.

Acknowledgements

This work was supported in part by the Science and Technology Research Program of Chongqing Municipal Education Commission (KJZD-K201902101); Key R & D plan of Guilin scientific research and technological development plan (20170101-3), Guilin science research and technology development plan (2016012006) and Basic Competence Promotion Project for Young and Middle-aged Teachers in Guangxi Universities (2017KY0862); Jouf University, Sakaka, Al-Jouf, KSA (40/168).

References (17)

There are more references available in the full text version of this article.

Cited by (30)

  • MFS enhanced SAM: Achieving superior performance in bimodal few-shot segmentation

    2023, Journal of Visual Communication and Image Representation
  • Quantum SUSAN edge detection based on double chains quantum genetic algorithm

    2022, Physica A: Statistical Mechanics and its Applications
    Citation Excerpt :

    Edge detection technology can extract the most meaningful edge features in the image, which can further provide the basis for image recognition and image detection analysis in real life [1]. The early edge detection methods mainly involve the classical Sobel operator, Roberts operator, Prewitt operator, and Laplacian operator [2–5]. The above operators are unsatisfactory for detecting oblique edges and some real edge points with small amplitudes, even though their processes are simple and fast.

  • A novel trust-based security and privacy model for Internet of Vehicles using encryption and steganography

    2022, Computers and Electrical Engineering
    Citation Excerpt :

    The proposed security method employs a minimum threshold of the ciphertext, although it takes an additional expense of alteration. The suggested methodology in this study does have many distortions [25,42–55]. Research [26] introduced a data encryption optimization method based on equality of the symmetric cryptography labeled "DJMNA."

  • Research on statistical detection method of micro bubbles in transparent layer of quartz crucible based on image processing

    2021, Journal of Crystal Growth
    Citation Excerpt :

    In traditional edge detection algorithms, Roberts, Sobel, and Prewitt operators are commonly used. In this paper, Sober detection operator is selected to detect the edge of crucible bubbles [19]. In the Sobel operator, for each pixel in the image, the weighted difference of the gray levels of the upper, lower, left, and right neighbors is examined, which is defined as follows:

View all citing articles on Scopus

This paper has been recommended for acceptance by Luming Zhang.

View full text