ABSTRACT
This study aims to automatically recognize drainage network patterns in a part of the Lahijan River using high-resolution panchromatic satellite imagery (HR-PRS) and fuzzy clustering algorithms. This study evaluates the performance of these methods in segmenting GeoEye-1 satellite images for detecting geomorphic features and drainage networks. After radiometric and geometric preprocessing, fuzzy segmentation of HR-PRS panchromatic images was performed using FWS, MSA, IDF, and CFM algorithms in MATLAB software. Fuzzy clustering algorithms were applied to the input HR-PRS images. The results show that the CFM (Classical Fusion Method and FCM) algorithm achieves superior performance in fuzzy segmentation and feature detection. This algorithm reduces segmentation errors caused by spectral feature overlap between classes and identifies spatial phenomena and clusters of various sizes, shapes, and densities, resulting in well-defined image boundaries. This superior performance is attributed to the use of fuzzy numbers and efficient clustering methods. Remote sensing technology provides multi-temporal imagery, offering a suitable foundation for monitoring environmental changes, detecting features, and accurately extracting information. The use of clustering algorithms and fuzzy features represents an optimal method for integrating information from HR-PRS satellite images for segmentation purposes.


1. INTRODUCTION
In recent decades, significant progress has been made in using multi-temporal, high-resolution satellite data as an important data source for detecting and monitoring land surface features and phenomena. Remote sensing technology provides new opportunities for studying the spatial organization of landscapes and drainage network characteristics due to its ability to acquire information under different climatic conditions, its multi-temporal and multi-spectral imaging capabilities, data availability, and increasing spatial and temporal resolution (Benincasa et al., 2019). Access to high-resolution panchromatic images enables automatic processing without human intervention for studying, identifying, and monitoring features, as well as classifying landforms (Chang et al., 2015; Wang et al., 2018). The presence of obstacles such as cloud cover and shadows is one of the most common sources of error in remote sensing imaging, causing confusion and misclassification in land surface information extraction (Swetnam et al., 2018; Jurado et al., 2020). Therefore, image segmentation is usually modeled as a clustering process or unsupervised classification. The advantages of image segmentation include algorithm simplicity, the ability to model, the elimination of uncertainty effects, and high implementation speed. Recent research indicates increasing use of fuzzy clustering algorithms for segmenting high-resolution panchromatic images in digital remote sensing image processing, with acceptable performance (Adachi et al., 2017; Hua, 2017; Arai et al., 2018). This paper aims to automatically recognize formal patterns of the drainage network using HR-PRS images and fuzzy clustering algorithms, while evaluating their performance in segmenting GeoEye-1 satellite images for detecting geomorphic features in the study area.
2. MATERIALS AND METHODS
Fuzzy segmentation using clustering algorithms and remote sensing techniques was applied to a part of the Lahijan River. The study area was selected due to its location in an area with human settlements and agricultural textures, offering good separability and diverse spatial and radiometric criteria. GeoEye-1 satellite panchromatic images with a spatial resolution of 0.5 meters were used, acquired from an altitude of 681 km. After radiometric and geometric pre-processing, fuzzy segmentation was performed using FWS, MSA, IDF, and CFM algorithms in MATLAB software (Bayram et al., 2018). For fuzzy clustering, input triangular fuzzy numbers for clustering in p dimensions were defined as x̃_k = {x̃_{k,1}, …, x̃_{k,p}} for k = 1, …, n (where n is the number of image pixels, p = 1). Cluster centers were represented as ṽ_i = {ṽ_{i,1}, …, ṽ_{i,p}} for i = 1, …, c (where c = 4 clusters). The processing was performed in four stages with parameters l_Max, θ > 1, K ≥ 0, and initial values for m_{ṽ_{i,j}}^{(0)}. The membership degrees u_{i,k}^{(l)} and cluster centers were iteratively updated until convergence. Defuzzification was then applied, assigning each pixel to the cluster with the highest membership degree, thereby producing the segmented image with clearly defined region boundaries. To improve fuzzy segmentation performance, digital number (DN) values and textural features (contrast, entropy, energy, and homogeneity from the GLCM matrix) were used alongside radiometric features (Ben Salah et al., 2010).

3. RESULTS
The FWS, CFM, IDF, and MSA algorithms were applied for fuzzy clustering and segmentation of HR-PRS panchromatic images in the study area. Performance was evaluated using three indicators:
1. Spatial indicator: Accurate detection of the river boundary and its distinction from surrounding features.
2. Spatial indicator: Correct detection of human settlement textures.
3. Spatio-radiometric indicator: Main and sub-boundaries between agricultural fields.
The CFM algorithm showed excellent performance in assigning pixels to different clusters and in finding optimal cluster numbers and centers. Image boundaries were well detected, and the algorithm provided higher accuracy in identifying clusters of various shapes, sizes, and densities, as well as in detecting spatial features. The CFM method performed best in the first indicator (river detection and distinction of drainage network boundaries from the surrounding environment). The IDF (Interval-valued Data Fuzzy c-means) method, which incorporates image uncertainty, performed second best after CFM for all three indicators. The MSA (Mean-Shift Algorithm) showed good performance in the first indicator and, along with IDF, showed the best performance in detecting rural settlement textures compared to the other methods.

4. DISCUSSION AND CONCLUSION
The results confirm the effectiveness of fuzzy clustering algorithms for segmenting multispectral remote sensing images and validate the performance of the proposed segmentation methods for detecting spatial features and accurately extracting information from images. In accordance with the research results, using clustering algorithms and fuzzy features is an optimal method for integrating information from HR-PRS satellite images of a geographical area for segmentation purposes.
Although no single method can be considered optimal in every respect or for every indicator, depending on the images, the geographical area, and the intended application, different algorithms may perform optimally. Combining the investigated algorithms in future research using Gaussian fuzzy numbers and GG-FCM clustering with fuzzy parameters could lead to an efficient method suitable for various geographical areas. Gaussian fuzzy numbers are the most suitable type for image segmentation using HR-PRS images, and the use of fuzzy numbers in general can lead to better results.
Given the favorable characteristics of fuzzy clustering algorithms, including robustness to noise and outliers, the use of Gaussian fuzzy features and fuzzy clustering methods based on Fuzzy C-Means (FCM) constitutes one of the most effective approaches for segmentation. Furthermore, the application of various types of fuzzy numbers and different metrics, as well as the fuzzification of membership degrees and distance parameters, can lead to optimal and desirable performance for more robust fuzzy clustering.
ACKNOWLEDGEMENT
We would like to thank Khorramshahr University of Marine Science and Technology for supporting this work under research grant contract No. 160.
REFERENCES
Adachi, M., Ito, A., Yonemura, S. and Takeuchi, W., 2017. Estimation of global soil respiration by accounting for land-use changes derived from remote sensing data. Journal of Environmental Management, 200, pp.97-104. https://doi.org/10.1016/j.jenvman.2017.05.076
Arai, R., Kodaira, S., Takahashi, T., Miura, S. and Kaneda, Y., 2018. Seismic evidence for arc segmentation, active magmatic intrusions and syn-rift fault system in the northern Ryukyu volcanic arc. Earth, Planets and Space, 70(1). https://doi.org/10.1186/s40645-017-0157-2
Bayram, B., Demir, N., Akpinar, B., Oy, S., Erdem, F., Vögtle, T. and Seker, D.Z., 2018. Effect of Different Segmentation Methods Using Optical Satellite Imagery to Estimate Fuzzy Clustering Parameters for SENTINEL-1A SAR Images. International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLII-1, pp.39-43. https://doi.org/10.5194/isprs-archives-XLII-1-39-2018
Ben Salah, M., Mitiche, A. and Ben Ayed, I., 2010. Effective level set image segmentation with a kernel induced data term. IEEE Transactions on Image Processing, 19, pp.220-232. https://doi.org/10.1109/tip.2009.2032940
Benincasa, M., Falcini, F., Adduce, C., Sannino, G. and Santoleri, R., 2019. Synergy of Satellite Remote Sensing and Numerical Ocean Modelling for Coastal Geomorphology Diagnosis. Remote Sensing, 11(22), p.2636. https://doi.org/10.3390/rs11222636
Chang, N.B., Bai, K. and Chen, C.-F., 2015. Smart information reconstruction via time-space-spectrum continuum for cloud removal in satellite images. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 8(5), pp.1898-1912. https://doi.org/10.1109/JSTARS.2015.2400636
Hua, A.K., 2017. Land Use Land Cover Changes in Detection of Water Quality: A Study Based on Remote Sensing and Multivariate Statistics. Journal of Environmental and Public Health, 2017, pp.1-12. https://doi.org/10.1155/2017/7515130
Jurado, J.M., Cárdenas, J.L., Ogayar, C.J., Ortega, L. and Feito, F.R., 2020. Semantic Segmentation of Natural Materials on a Point Cloud Using Spatial and Multispectral Features. Sensors, 20(8), p.2244. https://doi.org/10.3390/s20082244
Swetnam, T.L., Gillan, J.K., Sankey, T.T., McClaran, M.P., Nichols, M.H., Heilman, P. and McVay, J., 2018. Considerations for Achieving Cross-Platform Point Cloud Data Fusion across Different Dryland Ecosystem Structural States. Frontiers in Plant Science, 8. https://doi.org/10.3389/fpls.2017.02144
Wang, T., Yan, G., Mu, X., Jiao, Z., Chen, L. and Chu, Q., 2018. Toward operational shortwave radiation modeling and retrieval over rugged terrain. Remote Sensing of Environment, 205, pp.419-433. https://doi.org/10.1016/j.rse.2017.11.006