Correct segmentation scale

Oct 13, 2025 by vinooalluri.com 76 views

Introduction

The classification of object-based land cover is an increasingly applied method, especially in the context of land cover analysis in urban areas. In Medan Belawan District, the use of high-resolution imagery, especially the image of SpOT, is one of the focus of interesting research. Although pixel-based classifications have been widely carried out, information about the classification of object-based land cover (OBIA) in this area is still limited. This study aims to analyze various types of land cover using the OBIA method and the Decision Tree algorithm.

Background

Land cover classification is a crucial aspect of remote sensing and geographic information systems (GIS). It involves identifying and categorizing different types of land cover, such as vegetation, water bodies, and built-up areas, using satellite or aerial imagery. The accuracy of land cover classification is essential for various applications, including environmental monitoring, urban planning, and natural resource management.

In recent years, object-based image analysis (OBIA) has emerged as a popular method for land cover classification. OBIA involves segmenting an image into objects or regions of interest, which are then classified based on their characteristics. This approach has several advantages over traditional pixel-based classification methods, including improved accuracy and reduced computational complexity.

Methodology

This study lasted for five months, from December 2019 to April 2020, with a focus on 14 types of land cover in accordance with the real conditions in the field. The study area was Medan Belawan District, which is located in the province of North Sumatra, Indonesia. The study used high-resolution SpOT images, which were acquired in 2019.

The OBIA method was used to classify the land cover, and the Decision Tree algorithm was employed to classify the objects. The Decision Tree algorithm is a popular machine learning algorithm that is widely used for classification tasks. It works by recursively partitioning the data into smaller subsets based on the values of the input variables.

The study used two different segmentation scales, namely 50 and 150, to evaluate the effect of segmentation scale on the accuracy of land cover classification. The segmentation scale refers to the size of the objects or regions of interest that are created during the segmentation process.

Results

The results of this study show that the segmentation scale used greatly affects the accuracy of the classification results. The use of the 50 segmentation scale provides overall accuracy (Overall accuracy/OA) and Kappa accuracy reaches 100%, and the correlation value is 0.57. Conversely, on the segmentation scale of 150, the overall accuracy and accuracy of Kappa dropped to 97%, with a correlation value of only 0.46.

The validation results of the land cover map also show a significant difference between the two segmentation scales. The 50 segmentation scale produces a very good level of validation, which is 91.46%, while the 150 scale only reaches 81.66%. This shows that smaller segmentation scale can provide more accurate details in the classification of land cover.

Discussion

The results of this study confirm the importance of choosing the right methods and tools in the classification of land cover, especially in areas that continue to change, such as urban areas. The use of high-resolution imagery, such as SpOT images, can provide more accurate results than traditional pixel-based classification methods.

The Decision Tree algorithm is a powerful tool for land cover classification, and its use in this study has provided accurate results. However, the accuracy of land cover classification can be affected by various factors, including the quality of the imagery, the segmentation scale, and the classification algorithm used.

Conclusion

In conclusion, this study has demonstrated the effectiveness of the OBIA method and the Decision Tree algorithm in land cover classification. The use of high-resolution imagery and the correct segmentation scale can provide accurate results, which can be used for various applications, including environmental monitoring, urban planning, and natural resource management.

Further research is needed to explore various other classification techniques and increase the accuracy of measurement of land cover in other regions in Indonesia. Additionally, the use of other machine learning algorithms, such as random forest and support vector machines, can be explored to improve the accuracy of land cover classification.

Recommendations

Based on the results of this study, the following recommendations are made:

Use of high-resolution imagery: The use of high-resolution imagery, such as SpOT images, can provide more accurate results than traditional pixel-based classification methods.
Correct segmentation scale: The correct segmentation scale is essential for accurate land cover classification. A smaller segmentation scale can provide more accurate details in the classification of land cover.
Decision Tree algorithm: The Decision Tree algorithm is a powerful tool for land cover classification, and its use in this study has provided accurate results.
Further research: Further research is needed to explore various other classification techniques and increase the accuracy of measurement of land cover in other regions in Indonesia.

Limitations

This study has several limitations, including:

Study area: The study area was limited to Medan Belawan District, which may not be representative of other regions in Indonesia.
Imagery quality: The quality of the imagery used in this study may not be representative of other imagery used in land cover classification.
Classification algorithm: The use of the Decision Tree algorithm may not be the best approach for land cover classification in all cases.

Future Research Directions

Future research directions include:

Exploring other classification techniques: Exploring other classification techniques, such as random forest and support vector machines, can be used to improve the accuracy of land cover classification.
Increasing the accuracy of measurement: Increasing the accuracy of measurement of land cover in other regions in Indonesia can be achieved by using high-resolution imagery and the correct segmentation scale.
Applying land cover classification to other applications: Applying land cover classification to other applications, such as environmental monitoring and urban planning, can be used to improve the accuracy of land cover classification.

Introduction

In our previous article, we discussed the classification of object-based land cover using the Decision Tree algorithm in Medan Belawan District. In this article, we will answer some frequently asked questions (FAQs) related to this topic.

Q: What is object-based land cover classification?

A: Object-based land cover classification is a method of classifying land cover into different categories based on the characteristics of the objects or regions of interest. This approach is different from traditional pixel-based classification methods, which classify each pixel individually.

Q: What is the Decision Tree algorithm?

A: The Decision Tree algorithm is a popular machine learning algorithm used for classification tasks. It works by recursively partitioning the data into smaller subsets based on the values of the input variables.

Q: What is the significance of segmentation scale in object-based land cover classification?

A: The segmentation scale refers to the size of the objects or regions of interest that are created during the segmentation process. The correct segmentation scale is essential for accurate land cover classification. A smaller segmentation scale can provide more accurate details in the classification of land cover.

Q: What are the advantages of using high-resolution imagery in object-based land cover classification?

A: The use of high-resolution imagery, such as SpOT images, can provide more accurate results than traditional pixel-based classification methods. High-resolution imagery can capture more detailed information about the land cover, which can lead to more accurate classification results.

Q: What are the limitations of this study?

A: This study has several limitations, including:

Study area: The study area was limited to Medan Belawan District, which may not be representative of other regions in Indonesia.
Imagery quality: The quality of the imagery used in this study may not be representative of other imagery used in land cover classification.
Classification algorithm: The use of the Decision Tree algorithm may not be the best approach for land cover classification in all cases.

Q: What are the future research directions?

A: Future research directions include:

Exploring other classification techniques: Exploring other classification techniques, such as random forest and support vector machines, can be used to improve the accuracy of land cover classification.
Increasing the accuracy of measurement: Increasing the accuracy of measurement of land cover in other regions in Indonesia can be achieved by using high-resolution imagery and the correct segmentation scale.
Applying land cover classification to other applications: Applying land cover classification to other applications, such as environmental monitoring and urban planning, can be used to improve the accuracy of land cover classification.

Q: What are the practical implications of this study?

A: The practical implications of this study are:

Improved land cover classification: The use of object-based land cover classification and the Decision Tree algorithm can provide more accurate results than traditional pixel-based classification methods.
Better decision-making: The accurate classification of land cover can lead to better decision-making in various applications, such as environmental monitoring and urban planning.
Increased efficiency: The use of high-resolution imagery and the correct segmentation scale can increase the efficiency of land cover classification.

Q: What are the future applications of this study?

A: The future applications of this study are:

Environmental monitoring: The accurate classification of land cover can be used to monitor environmental changes and track the impact of human activities on the environment.
Urban planning: The accurate classification of land cover can be used to plan and manage urban areas more effectively.
Natural resource management: The accurate classification of land cover can be used to manage natural resources more effectively and sustainably.

Conclusion

In conclusion, this Q&A article has provided answers to some frequently asked questions related to the classification of object-based land cover using the Decision Tree algorithm in Medan Belawan District. The use of object-based land cover classification and the Decision Tree algorithm can provide more accurate results than traditional pixel-based classification methods. The accurate classification of land cover can lead to better decision-making in various applications, such as environmental monitoring and urban planning.