[Retracted] Wireless Network Sensing of Urban Surface Water Environment Based on Clustering Algorithm

Zhao, Qichao; Yang, Xiufeng; Dong, Xuxin; Li, Huairui

doi:https://doi.org/10.1155/2021/2829195

Wireless Communications and Mobile Computing

On this page

Abstract Introduction Methods Results and Analysis Conclusions Data Availability Conflicts of Interest Acknowledgments References Copyright Related Articles

Research Article Retraction

!

This article has been Retracted. To view the article details, please click the ‘Retraction’ tab above.

Special Issue

Artificial Intelligence Applications in Mobile Virtual Reality Technology

View this Special Issue

Research Article | Open Access

Volume 2021 | Article ID 2829195 | https://doi.org/10.1155/2021/2829195

[Retracted] Wireless Network Sensing of Urban Surface Water Environment Based on Clustering Algorithm

Qichao Zhao,^1,2Xiufeng Yang,^1,2Xuxin Dong,^2,3and Huairui Li⁴

Academic Editor: Balakrishnan Nagaraj

Received08 Oct 2021

Revised26 Oct 2021

Accepted15 Nov 2021

Published06 Dec 2021

Abstract

To improve the wireless sensing image extraction technology of urban surface water environment, a regional FCM clustering method combined with water index was proposed in this paper. The normalized water index (NDWI) was obtained by calculating the fusion multispectral wireless sensing image. Through the combination with normalized water index, fuzzy clustering results were obtained by RFCM algorithm proposed in this paper. The optimal threshold was selected to defuzzify the fuzzy clustering results, and finally, the extraction results of urban surface water were obtained. The accuracy of the proposed algorithm was compared with that of the traditional surface water extraction algorithm. The experimental results showed that the size of different neighborhood regions affected the water extraction accuracy. In W city, the kappa coefficient of MFCM16 was 0.41% higher than that of MFCM8, and the overall classification accuracy of MFCM16 was 1.33% higher than that of MFCM. In G city area, the kappa coefficient of MFCM16 was 1.81% higher than that of MFCM8, and the overall classification accuracy of MFCM16 was 1.7% higher than that of MFCM. Comparing the RFCM algorithm with other algorithms, the RFCM algorithm obtained the best experimental results, to reduce the “salt-and-pepper phenomenon” effect.

1. Introduction

With the development of wireless sensing technology, especially the improvement of wireless sensing image processing level, wireless sensing is increasingly widely used in various social fields. In terms of urban planning, wireless sensing can realize dynamic monitoring of land use, supervision and control of air quality, urban ecological environment planning and construction, etc. [1]. Urban surface is a subsystem of urban ecosystem. As an important part of urban ecosystem, it plays an important role in improving urban ecological environment quality and improving residents’ living standards [2]. In recent years, many cities at home and abroad have applied wireless sensing technology to surface information extraction to dynamically master the coverage area, optimize the spatial structure of green space, improve the potential of sustainable development of the city, and realize the overall planning [3]. Compared with the traditional way, the extraction of green space information by aerial wireless sensing image has the advantages of wide field of view, strong macroscopic view, clear and realistic image, large amount of information, short repetition cycle, and convenient data collection. It is very economical in terms of manpower, material resources, and financial resources and has a short time and high efficiency. Wireless sensing is a comprehensive earth detection technology developed in the century [4]. That is, collect information about an object without directly touching it. It usually refers to the acquisition of various ground object information from the air or space by some kind of sensor and the extraction and analysis of this information, so as to measure and judge the nature or characteristics of the ground object. With the development of space technology, optical technology, sensor technology, computer technology, and modern communication technology, wireless sensing technology has made great progress [5]. Since the rise of the century, the development of wireless sensing technology is increasingly rapid, on the basis of aerial photogrammetry, with the rapid development of modern science and technology such as space technology and electronic computer, as well as the needs of the development of geoscience, biology, and other disciplines; as an emerging technical discipline gradually developed, it has formed a relatively complete basic theoretical system and a series of technical support [6]. As a means of information acquisition, wireless sensing technology has penetrated into various fields of the national economy, such as agriculture, forestry, geology, meteorology, oceanography, environment, urban planning and land management, and other professional fields and departments [7]. Zhao et al. proposed the main methods of urban wireless sensing summarized as postclassification comparison method, multitemporal complex method, image difference ratio method, vegetation index method, principal component analysis, and transformation vector analysis [8]. Song et al. proposed a new urban land use classification method, which determines urban building density based on image texture to determine land use type. They used panchromatic spectral images to conduct experiments in Athens, Greece, which is higher than the traditional maximum likelihood method. This method can be used for urban wireless sensing monitoring such as urban land use change, urban expansion, and illegal building monitoring [9]. Awad et al. adopted the difference method, used spatial texture information and set certain constraints to eliminate the change information of agricultural land, and accurately extracted the annual urban expansion area [10]. Based on the current research, MFCM algorithm only considers the fixed neighborhood of each pixel to improve the robustness of the algorithm [11]. Most of the mature monitoring system instruments are imported. The imported instruments have high precision and many measuring indexes, but they are expensive and cumbersome to operate. The wireless frequency domain used in China needs to be specially applied to the radio department. There are also some independent research and development monitoring systems, but compared with foreign countries, there are still some problems, such as the number of monitoring stations and limited scope. In particular, in the wireless sensor network, the application of wireless sensor network technology is less. Real-world objects, however, are irregular neighborhood areas. On the basis of the current research, this paper proposes a regional FCM clustering method combined with water index, which calculates the normalized water index (NDWI) through the fusion of multispectral wireless sensing images. Combined with normalized water index, fuzzy clustering results were obtained by RFCM algorithm proposed in this paper. The optimal threshold was selected to defuzzify the fuzzy clustering results, and finally, the extraction results of urban surface water were obtained. The accuracy of the proposed algorithm was compared with that of the traditional surface water extraction algorithm. The experimental results showed that the size of different neighborhood regions affected the water extraction accuracy. In W city, the kappa coefficient of MFCM16 was 0.41% higher than that of MFCM8, and the overall classification accuracy of MFCM16 was 1.33% higher than that of MFCM. In G city area, the kappa coefficient of MFCM16 was 1.81% higher than that of MFCM8, and the overall classification accuracy of MFCM16 was 1.7% higher than that of MFCM. Comparing the RFCM algorithm with other algorithms, the RFCM algorithm obtained the best experimental results, to reduce the “salt-and-pepper phenomenon” effect [12]. The innovation of sensor technology and communication technology makes wireless sensor network technology more perfect, which is often used in environmental monitoring. Using wireless sensor network to monitor water quality and obtain water quality data for storage and remote transmission can well solve the shortcomings of traditional monitoring methods and achieve real-time, low-cost, and long-term measurement of the environment.

2. Methods

2.1. Wireless Sensing Data Preprocessing

Two-scene gF-2 domestic high-resolution wireless sensing images (Guangzhou and Wuhan) were used to extract urban surface water. (1) Firstly, radiometric calibration and atmospheric correction are performed on the multispectral data and panchromatic data, and then, a 1 m resolution multispectral wireless sensing image is obtained by using NND (Nearest Neighbor Diffusion) image fusion algorithm. (2) Use morphological shadow index to remove building shadow. (3) Normalized water index (NDWI) was obtained by using fusion multispectral wireless sensing images. (4) Combined with normalized water index, fuzzy clustering results were obtained by using the RFCM algorithm proposed in this paper. (5) Select the optimal threshold to defuzzify the fuzzy clustering results, and finally, obtain the extraction results of urban surface water. (6) Artificial vectorized real surface water data is used to verify the effectiveness of the algorithm, and the accuracy of the algorithm is compared with that of the traditional surface water extraction algorithm [13].

2.2. Water Index

Water index method is a combination of single-band threshold method and multiband spectral logic operation, so as to improve the difference between water body and other ground objects and effectively suppress the influence of background noise (shadow, black impervious surface, ice and snow, etc.). Water index method is to use the water body in visible light reflectance which is generally low, less than 10%, generally 4%-5%, and gradually decreases with the increase of wavelength, but in the near-infrared band, water body almost shows full absorption. In order to enhance the difference between water body and other ground objects, the ratio calculation of visible band and near-infrared band is carried out or different weights are given to the band based on the above principle. At present, the mainstream water index includes normalized water index (NDWI), improved normalized water index (MNDWI), enhanced water index (EWI), and automated water extraction index (AWEI). However, gF-2 domestic high-resolution wireless sensing image has only one panchromatic band, which consists of four multispectral bands, namely, red, green, blue, and near middle red, without middle infrared band and thermal infrared band. Therefore, normalized water index (NDWI) is adopted in this paper. The specific calculation is shown in Formula (1). This water index has a good suppression effect on background noise [14]. where is the surface reflectance of green band and is the surface reflectance of near-infrared band.

2.3. Regional FCM Clustering Algorithm

2.3.1. Principle of Regional FCM Clustering Algorithm

RFCM (regional fuzzy C-means) clustering algorithm is derived from traditional FCM algorithm and improved FCM algorithm. The traditional FCM algorithm only considers the information of the pixel itself, not the spatial information of the pixel neighborhood. The improved FCM algorithm MFCM (modified FCM) considers the spatial information of the pixel neighborhood, but it only considers the spatial information of the fixed window in the pixel regular neighborhood. Real objects are irregular neighborhood areas (image objects). The regional FCM clustering algorithm will determine the size of the neighborhood region according to the spatial heterogeneity of the pixel and the neighborhood pixel. The principle of the RFCM algorithm only adds spatial information of pixel neighborhood region on the basis of the FCM algorithm and considers the membership degree constraint of pixel neighborhood region. Finally, the membership degree was obtained by iterative optimization of RFCM objective function considering neighborhood region constraints, and the category of pixels was determined according to the optimal threshold of membership degree. The objective function of region FCM is shown in where represents the membership degree that pixel belongs to category ; represents a weighting coefficient; represents the feature space distance between pixel and cluster center . is the constraint parameter of neighborhood area; represents the feature space distance between the mean value of the feature space of the pixel in the neighborhood of pixel and of the cluster center. By introducing Lagrange theory to optimize the objective function of regional FCM, a constraint function is constructed, as shown in where stands for Lagrange multiplier; represents the sum of the membership degree of each category for the JTH pixel, where .

2.3.2. Calculation of Neighborhood Area

The key of RFCM algorithm is the calculation of neighborhood area. The size of neighborhood area is determined by the spectral difference between center pixel and neighborhood pixel. PSI index method is adopted in this paper to determine neighborhood area, which is more reasonable than fixed window neighborhood area [15]. The basic theory of the algorithm is a series of direction lines diverging from the center of the pixel in different directions as shown in Figure 1. The spectral heterogeneity of the pixel on each direction line and the central pixel is calculated. If the spectral heterogeneity is less than the threshold value and the length of the direction line is less than the threshold value, the pixel is determined as the neighborhood area of the central pixel. The calculation of spectral heterogeneity measure is shown in where represents the heterogeneity measure value of the current pixel and the neighborhood pixel on the direction line , and represent the spectral value of the center pixel and the current neighborhood pixel on band , respectively, and represents the number of bands. Under the condition of two thresholds, each direction line expands simultaneously from the center pixel. One threshold condition is that when the current pixel heterogeneity value is greater than the spectral constraint threshold , diffusion stops on this direction line. Another threshold condition is that the length of the direction line between the current pixel and the center pixel is less than the threshold , so the search in this direction is stopped to prevent the neighborhood on the direction line from being too large. Finally, the neighborhood region pixels that meet the conditions in each direction of the central pixel are counted, and the pixel set obtained is the neighborhood region of the pixel.

2.3.3. The Specific Process of RFCM Algorithm

Set the number of clustering and parameter , and terminate the algorithm , the maximum number of iterations , spectral difference measure threshold, and direction line length threshold. According to the neighborhood region calculation method, the neighborhood region of each pixel of the image is determined, and of each pixel is calculated. Initialize of each cluster center. The membership function was calculated by using the current clustering center , and then, of various clustering centers was calculated according to the membership function . According to the threshold and the maximum number of iterations , if or the number of iterations , the iteration is terminated. When the algorithm converges, the membership degree of each cluster center and each pixel to each cluster center is obtained, and the fuzzy clustering result is completed. Finally, the fuzzy clustering results are defuzzized according to the optimal threshold value, and the fuzzy clustering results are transformed into deterministic classification to realize the final clustering of wireless sensing images.

Compared with the spectral features of wireless sensing image, the water index improves the separability of water body and other ground objects and has a good suppression effect on background noise. Therefore, the normalized water index and shadow index are stacked with spectral features as the input features of RFCM clustering algorithm in the experiment. At the same time, in RFCM clustering algorithm, the spatial information of homogeneous neighborhood region of pixels is considered. Based on the advantages of the above two aspects, an urban surface water extraction algorithm integrating water index and RFCM was designed and applied to surface water extraction from domestic gF-2 wireless sensing images in complex urban environment.

2.4. Design Objectives of Water Environment Monitoring System Based on Wireless Sensor Network

(1)The construction of intelligent system and the application of wireless sensing technology to achieve the measurement and monitoring of water quality(2)To achieve reliable, high-speed, and low-power wireless transmission of water monitoring data(3)Collect data and monitor data: the host computer and the web page display the location of specific water areas and sensor nodes in water and display the latest time recorded data of various water quality parameters measured by sensor nodes(4)Under low power consumption, it can work normally for a long time

3. Results and Analysis

The RFCM algorithm proposed in this paper is compared and analyzed with the improved MFCM algorithm ( and were considered as MFCM8 and MFCM16 in the experiment), -means clustering algorithm, NDWI threshold algorithm (TH), and object-oriented method (OBIA). In the experiment, OA and kappa coefficients of overall classification accuracy were used for statistical and quantitative evaluation of accuracy, and visual discrimination was used for qualitative evaluation [16].

The classified kappa coefficients and overall accuracy of the two research areas in G city and W city are shown in Figures 2 and 3.

The kappa coefficient in G city is 89.88%, the kappa coefficient in W city is 92.49%, and the overall classification accuracy in G city is 89.14%. The overall classification accuracy of W city is 92.58%, and the classification accuracy of the method considering regional spatial information (MFCM and RFCM) and the object-oriented method (OBIA) is higher than that of the pixel-based method (-means) and the water index threshold method (TH) and has a lower misclassification error rate and missed classification error rate [17]. Among them, in W city, the kappa coefficient of RFCM was 1%, 1.5%, and 1.3% higher than that of OBIA, MFCM8, and MFCM16, respectively. The OA of RFCM was 3.86%, 5.82%, and 4.48% higher than that of OBIA, MFCM8, and MFCM16, respectively. In city G, the kappa coefficient of RFCM was 0.2%, 2.3%, and 0.7% higher than that of OBIA, MFCM8, and MFCM16, respectively, and the overall classification accuracy (OA) of RFCM was 3.2%, 5.83%, and 4.12% higher than that of OBIA, MFCM8, and MFCM16, respectively [18]. The OBIA method maintains the integrity of ground object better than -means and TH method. RFCM and MFCM algorithms not only maintain the integrity of ground object but also better retain local details of ground object. Compared with other OBIA, -means, TH, MFCM8, and MFCM16 algorithms, the RFCM algorithm can not only maintain the integrity of ground object but also better retain local details of ground object. The fine surface water bodies are effectively identified, the boundary information of surface water bodies is maintained well, and the influence of shadow of urban buildings is eliminated at the same time [19]. The improved FCM algorithm also considered the neighborhood regions with different window sizes of rules. Experimental results showed that the size of different neighborhood regions affected the water extraction accuracy. In W city, the kappa coefficient of MFCM16 was 0.41% higher than that of MFCM8, and the overall classification accuracy of MFCM16 was 1.33% higher than that of MFCM. In G city area, the kappa coefficient of MFCM16 was 1.81% higher than that of MFCM8, and the overall classification accuracy of MFCM16 was 1.7% higher than that of MFCM.

The “salt-and-pepper” phenomenon of the experimental results was quantitatively counted, and the speckle noise was defined as the water area less than the water area. Table 1 shows that compared with other algorithms, the RFCM algorithm can obviously eliminate the “salt-and-pepper phenomenon” of surface water extraction. In the comparison experiment of the six algorithms, TH has the worst effect and has a large number of noise spots. Due to the phenomenon of “same object, different spectrum, foreign object in the same spectrum” of high-resolution data, a large number of isolated noise points are caused. -means algorithm is better than TH algorithm and MFCM8 and MFCM16 algorithms. Thirdly, MFCM algorithm significantly reduces the “noise” spots of surface water, which proves that the spatial information of pixel neighborhood has a certain inhibitory effect on the “noise” spots of water. The OBIA algorithm is better than MFCM8 and MFCM16. The RFCM algorithm achieves the best experimental results and has a significant effect on reducing “salt-and-pepper phenomenon” [20].

4. Conclusions

In this paper, a regional FCM clustering method combined with water index is proposed, which calculates the normalized water index (NDWI) from the fusion of multispectral wireless sensing images. Through the design of wireless sensor network water environment monitoring system, combined with normalized water index, fuzzy clustering results were obtained by the RFCM algorithm proposed in this paper. The optimal threshold was selected to defuzzify the fuzzy clustering results, and finally, the extraction results of urban surface water were obtained. The accuracy of the proposed algorithm was compared with that of the traditional surface water extraction algorithm. Comparing the RFCM algorithm with other algorithms, the RFCM algorithm obtained the best experimental results, to reduce the “salt-and-pepper phenomenon” effect. The algorithm in some details of surface water extraction was not taken into account and will continue to be studied in the future.

Data Availability

The data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest

The authors declare no conflicts of interest.

Acknowledgments

The study was supported by the (1) Natural Science Foundation of Hebei Province, spectral characteristics and remote sensing identification method of black and odorous water bodies in urban and rural areas at county scale, Project No. D2020409005; (2) Scientific Research Foundation of Higher Education Institutions of Hebei Province, remote sensing recognition model of black and odorous water in Beijing-Tianjin-Hebei region based on high-resolution image, Project No. ZD2019138; and (3) National Defense Basic Scientific Research Program of China, Project No. JCKY2019407D004.

References

C. Yu, L. Wang, J. Zhao, L. Hao, and Y. Shen, “Remote sensing image classification based on rbf neural network based on fuzzy c-means clustering algorithm,” Journal of Intelligent and Fuzzy Systems, vol. 38, no. 4, pp. 3567–3574, 2020.
View at: Publisher Site | Google Scholar
L. Zhang and A. Li, “Region-of-interest extraction based on saliency analysis of co-occurrence histogram in high spatial resolution remote sensing images,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 8, no. 5, pp. 2111–2124, 2015.
View at: Publisher Site | Google Scholar
Y. Cao, Y. Xun, Y. Han et al. et al., “Feature extraction of wireless sensing images based on bat algorithm and normalized chromatic aberration,” IFAC-PapersOnLine, vol. 52, no. 24, pp. 318–323, 2019.
View at: Google Scholar
B. Fha, Y. Ying, and C. Tf, “Automatic extraction of impervious surfaces from high resolution wireless sensing images based on deep learning,” Journal of Visual Communication and Image Representation, vol. 58, pp. 453–461, 2019.
View at: Google Scholar
W. Duo, H. GuoJin, and J. Weili, “Research on wireless sensing image fusion based on compressive sensing algorithm,” Journal of Computers, vol. 13, no. 5, pp. 519–526, 2018.
View at: Google Scholar
Y. Du, N. Yang, and M. Dong, “Application of target detection algorithm based on depth learning in wireless sensing image classification,” Revista de la Facultad de Ingenieria, vol. 32, no. 4, pp. 179–187, 2017.
View at: Google Scholar
D. Li, F. Yang, and X. Wang, “Study on ensemble crop information extraction of wireless sensing images based on svm and bpnn,” Journal of the Indian Society of Wireless Sensing, vol. 45, no. 2, pp. 1–9, 2016.
View at: Google Scholar
W. Zhao, X. Ma, L. Liang et al., “Remote sensing image registration based on dynamic threshold calculation strategy and multiple-feature distance fusion,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 12, no. 10, pp. 4049–4061, 2019.
View at: Publisher Site | Google Scholar
Y. Song, L. Yang, J. Xu, and A. Wei, “Water body information extraction based on landsat-8 satellite oli wireless sensing image and adaboost algorithm,” Journal of Geomatics, vol. 42, no. 3, pp. 44–47, 2017.
View at: Google Scholar
M. M. Awad, “Toward robust segmentation results based on fusion methods for very high resolution optical image and lidar data,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 10, no. 5, pp. 2067–2076, 2017.
View at: Publisher Site | Google Scholar
Z. Xi, J. Liu, and D. Liu, “Research on the water area extraction in Suzhou based on remote sensing,” IOP Conference Series Materials Science and Engineering, vol. 472, no. 1, article 012078, 2019.
View at: Publisher Site | Google Scholar
F. Wang, Y. Cheng, S. Li, H. Mu, and L. Li, “Wireless sensing image fusion algorithm based on multi-feature,” Xibei Gongye Daxue Xuebao/Journal of Northwestern Polytechnical University, vol. 33, no. 3, pp. 489–494, 2015.
View at: Google Scholar
S. Wang, J. Li, Q. Shen, B. Zhang, F. Zhang, and Z. Lu, “Modis-based radiometric color extraction and classification of inland water with the forel-ule scale: a case study of lake taihu,” IEEE Journal of Selected Topics in Applied Earth Observations & wireless sensing, vol. 8, no. 2, pp. 907–918, 2017.
View at: Google Scholar
N. Shrivastava and P. K. Rai, “Automatic building extraction based on multiresolution segmentation using remote sensing data,” Geographia Polonica, vol. 88, no. 3, pp. 407–421, 2015.
View at: Publisher Site | Google Scholar
Q. Z. Ye, P. Wu, and M. L. Zhang, “Research on automatic highway extraction technology based on spectral information of wireless sensing images,” Journal of Information Hiding and Multimedia Signal Processing, vol. 8, no. 2, pp. 368–380, 2017.
View at: Google Scholar
D. Velásquez, A. Sánchez, S. Sarmiento, M. Toro, M. Maiza, and B. Sierra, “A method for detecting coffee leaf rust through wireless sensor networks, remote sensing, and deep learning: case study of the caturra variety in Colombia,” Applied Sciences, vol. 10, no. 2, p. 697, 2020.
View at: Publisher Site | Google Scholar
O. Kanoun, S. Marcuccio, S. L. Ullo, and M. Carminati, “From the editors of the special issue on wireless sensor networks and remote sensing for environmental applications,” IEEE Aerospace and Electronic Systems Magazine, vol. 34, no. 6, pp. 3–3, 2019.
View at: Publisher Site | Google Scholar
P. J. Rao, K. N. Rao, and S. Kubo, Proceedings of International Conference on Remote Sensing for Disaster Management: Issues and Challenges in Disaster Management,, Springer Series in Geomechanics and Geoengineering, Springer, 2019.
View at: Publisher Site
W. T. Honeycutt, T. Kim, M. T. Ley, and N. F. Materer, “Sensor array for wireless remote monitoring of carbon dioxide and methane near carbon sequestration and oil recovery sites,” RSC Advances, vol. 11, no. 12, pp. 6972–6984, 2021.
View at: Publisher Site | Google Scholar
B. Yadegar, M. S. Wireless, and C. Whitehurst, “Improving the quality of life for skilled nursing residents through remote sensing of incontinence event,” Nonwovens Industry, vol. 50, no. 3, pp. 34-35, 2019.
View at: Google Scholar

Copyright

Copyright © 2021 Qichao Zhao et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

PDF Download Citation

Download other formats

Order printed copies

Views

264

Downloads

412

Citations