Automated Extraction of Shoreline in Tuban Regency, East Java from Google Earth Imagery by Integrating Canny Edge Detector

The shoreline is an area that becomes the boundary between land and sea and experiences morphological changes over time. This region has a dynamic condition where various components (air, rocks, water) are interconnected. Multitemporal shoreline analysis is one of the critical parameters for monitoring coastal areas. This information can be used for morphodynamic modeling, coastal area management, and erosion and accretion studies. This study aims to analyze shoreline changes in the North Coast of Tuban Regency, East Java using the Canny Algorithm and Google Earth Imagery from 2000 to 2020. The Canny algorithm was chosen because it has been tested to produce sharp and good edges compared to other edge detection algorithms. From this research, it can be concluded that in the north coast of Tuban Regency, based on the sample years taken, the area's shoreline experienced the erosion of 0.297 1.566 meters/ five years. The edges generated using the Canny algorithm are practical in interpreting shorelines and making analysis faster. In the future, there is a need for more elaboration regarding the use of Google Earth imagery in shoreline analysis, especially in geometric corrections (Georeference). This elaboration is essential because it will affect the analysis results, especially the shoreline position.


Introduction
Tuban Regency is one of East Java districts from all 38 districts and cities in the province (Tuban Regency Government, 2018). Tuban Regency is located on the northern coast of Java Island with a shoreline of about 65 km with an area of 1,904.70 km² (Tuban Regency Government, 2018). In some of its areas, Tuban Regency is located in a coastal environment that makes these communities rely on the sea's produce by becoming fishermen. The coast becomes one of the areas of human activity that can be used for various fishing activities and settlements (Driptufany, 2020). The utilization will reduce land in the coastal area, and the land is carrying capacity, which causes erosion and sedimentation (Driptufany, 2020). The shoreline becomes the barrier between the land and the ocean, which changes in morphology from time to time, which can be influenced by the sea level rise (Utami et al., 2017). This region has a very dynamic condition where air, rocks, and water are interconnected (Kasim, 2012). The coastal dynamics process is closely related to the coastal areas' management (Kasim, 2012). Alesheikh et al. (2007); Kasim (2012) argues that multitemporal shoreline analysis is one of the critical parameters for monitoring coastal areas. This information can be used for morphodynamic modeling, coastal area management, and erosion and accretion studies (Chand & Acharya, 2010;Kasim, 2012). Fuad et al. (2019); Suniada (2015) states that remote sensing techniques can analyze shoreline change. Along with the development of technology, one of the methods for shoreline extraction is to use edge detection techniques. Edge detection is the image processing stage to produce each object's edges in the image (Munir, 2019). The image's edge can be seen from the neighboring points' grey points (x and y). The benefits of edge detection can also reduce the amount of data processed and can be used for change detection on the shoreline (Munir, 2019 Ramadhani et al. (2021) in the Coastal District of Saying, Demak Regency using remote sensing methods and the Digital Shoreline Analysis System (DSAS). The results showed a change in shoreline abrasion and accretion by 82% and 18%, with a tendency to abrasion.
From the explanation above, a problem of how edge detection performs detecting shorelines in Google Earth imagery arises. Identifying the object's edge is vital because it is a preliminary study to observe changes in the shoreline more quickly. Therefore, this study aims to analyze changes in the shoreline in the North Coast of Tuban Regency, East Java using the Canny Algorithm. We have conducted a canny edge detector in Gili Raja Island, Sumenep (Prayogo & Hidayah, 2021). This algorithm was chosen because it has been tested to produce sharp and good edges compared to other edge detection algorithms (Maini & Aggarwal, 2009).

Research Location
This research is located at 6° 53'27.51 "S and 112° 3'38.10" E in the North of Tuban Regency, East Java. After all, this location is suspected of experiencing abrasion yearly because it is located directly opposite the open sea. The shoreline observed in this study is approximately 650 meters long. Figure 1 shows the research location displayed on the Basemap World imagery. Canny Algorithm Canny edge detection is a technique for extracting structural information that aims to reduce processed data. Based on Canny (1986); Deriche (1987) states that this process consists of at least five stages, namely: Applying a Gaussian filter so that the image becomes smoother and minimizes noise with the following equation (Gaussian filter (2k + 1) × (2k + 1))) (equation 1): Then determine the image intensity gradient with the following equation (equation 2) The edge direction angles represent vertical, horizontal, and two diagonals (0°, 45°, 90°, and 135°). Then it can apply steps such as (1) applying nonmaximum compressions to eliminate spurious responses to edge detection, (2) specifying a double threshold for determining potential image edges, and (3) Track edges with hysteresis: suppressing all other weak edges and not connected to the firm edge (Canny, 1986;Deriche, 1987).

Image Preprocessing
Before the image is filtered, the first thing to do is create a Ground Control Point (GCP) in Google Earth. GCP aims to adjust the coordinates on the map with coordinates in the field (Danoedoro, 2012). There are four GCPs used for georeferencing on five Google Earth maps. Tables 1 and 2 are information on the recording date of images and GCP used in this study. Image edge processing using the Canny algorithm is carried out at least in several steps (Chapter 2). This step is carried out to obtain structural information for each observed object, namely the shoreline in Google Earth imagery, from 2000 to 2020. This detection has a strict definition compared to other edge detections, so that the results of Canny detection are better than other edge detection. Figure 2 shows the Google Earth multitemporal imagery data from 2000 to 2020 used in this study. July 11, 2000July 28, 2003 August 16, 2010 July 8, 2016 September 6, 2020 The Canny algorithm has general criteria for detecting object edges. The operator's detected edge must be accurately localized in the center with marked once in each object. The detection must capture as many edges as possible to produce a suitable edge with minimal errors. According to requirements, the technique used in Canny detection to obtain edge information uses the calculus of variations function. The first derivative of Gaussian can explain this function. Figure 3 shows the results of Canny edge detection for shoreline analysis in the Tuban Regency, East Java. 2000 2003 2010 2016 The Gaussian filter on the image minimizes noise so that the object's edges are easily detected. The image's noise significantly affects the shoreline's extraction, so smoothing is needed at this stage. Besides, this filter uses a kernel window that is not static and can be changed according to each object's needs being filtered. The next step is to thin the edges of the image. This process is carried out to determine the change's location in the highest / sharpest intensity value to produce a more authentic and accurate image edge. The final process of shoreline detection with Canny is edge tracking with hysteresis. This stage aims to trace the edges of the weak pixels caused by the unconnected image's noise response. From the detection process, the following is the appearance of the shoreline in 2000, 2003, 2010, 2016, and 2020 in the North Coast of Tuban Regency, East Java (Figures 4).  Table 3 shows the shoreline shift information for each sample of the study locations.

Conclusion
From this research, it can be concluded that in Pesisir Utara, Tuban Regency, based on the sample years taken, the area's shoreline experienced the erosion of 0.297 -1.566 meters / five years. The edges generated using the Canny algorithm are very helpful in interpreting shorelines and making analysis faster. In the future, there is a need for more elaboration regarding the use of Google Earth imagery in