COMPARISON OF MACHINE LEARNING ALGORITHMS FOR MASS APPRAISAL OF REAL ESTATE DATA, 2024
In recent years, machine learning algorithms have been used in the mass appraisal of real estate.... more In recent years, machine learning algorithms have been used in the mass appraisal of real estate. In this study, 5 machine learning algorithms are used for residential type real estate. Machine learning algorithms used for mass appraisal in this study are Artificial Neural Networks (ANN), Random Forest (RO), Multiple Regression Analysis (MRA), K-Nearest Neighborhood (k-nn), Support Vector Regression (SVR). To test the study, real estate data collected from the central districts of Ankara, were used. The main purpose of this study is to find out which machine learning algorithm gives the best results for the mass appraisal of real estates and to reveal the most important variables that affect the prices of real estate. According to the results obtained for the city of Ankara, it was observed that the best algorithm for mass appraisal is RF in residential-type real estates, followed by the ANN, k-nn, and linear regression algorithms, respectively. According to the results obtained from the residential real estate, it was concluded that heating and distances to places of importance had the greatest effect on the value. mass appraisal, machine learning algorithms, random forest, artificial neural network, real estate valuation map JEL Classification: R39
Urbanization Spatial and settlement development Dichotomy urban settlement Land cover change Land... more Urbanization Spatial and settlement development Dichotomy urban settlement Land cover change Land governance Research Article
Income from taxes constitutes the main revenue source of local and central governments in emergin... more Income from taxes constitutes the main revenue source of local and central governments in emerging economies. Urban services envisaged in development plans for rapidly growing urban populations are provided by local governments; however, available financial resources remain inadequate for effective, efficient and high quality service delivery. Despite the presence of low-density neighborhoods, especially in urban centers, the transformation of land to land lots and increasing residential density in the urban periphery place additional economic burden on local governments. In this context, a model proposal has been made in this study towards controlling by taxation, which is an application tool of land lot policy, of urban sprawl process, increasing local government revenues and reducing revenue loss to a minimum in the Ankara Metropolitan area, of which scale is gradually growing and governance efficiency is declining. The results of the evaluation carried out in a 59.7 hectare study area in the Çukurambar Region in the Çankaya District of Ankara Province have shown that it is possible to develop a real estate valuation system based on market value in order to transfer the real estate value increment resulting from development plan decisions to the public by taxation policies, and thus reduce tax and fee revenue losses resulting from tax base. It has been confirmed that the suggested tax policies will have a reducing effect on infertile land acquisition with value increment expectations and land possession trends; that local governments' tax revenues will increase by a factor of four and infrastructural investments financing might be provided by value increment financing. Using practices such as controlling the process of transformation of land to land lot, property tax, taxation of value increment, and reducing economic rent seekers' tendencies of acquiring unearned increment will be possible.
In this study, the Istanbul Province was monitored using Landsat 5 TM, MSS, Landsat 7 ETM+, and L... more In this study, the Istanbul Province was monitored using Landsat 5 TM, MSS, Landsat 7 ETM+, and Landsat 8 OLI imagery from the years 1986, 2000, 2009, 2011, 2013, and 2015 in order to assess land cover changes in the province. The aim of the study was to classify manmade structures, land, green, and water areas, and to observe the changes in the province using satellite images. After classification, the images were compared in selected years to observe land cover. Moreover, these changes were correlated with the property tax values of Istanbul by years. The findings of the study showed that manmade structure areas increased while vegetation areas decreased due to rapid population growth, urbanization, and industrial and commercial development in Istanbul. These changes also explain the transformation of land from rural and natural areas to residential use, and serve as a tool with which to assess land value increments. Land value capturing is critical for the analysis of the linkage...
ISPEC 9. ULUSLARARASI MÜHENDİSLİK VE FEN BİLİMLERİ KONGRESİ 13-15 Kasım 2020 Ankara, Türkiye, 2020
In recent years, studies performed by traditional methods in many branches of science have been r... more In recent years, studies performed by traditional methods in many branches of science have been replaced by the method called as machine learning. Effects such as speed and low performance observed with traditional methods are increased by using machine learning algorithms. In the field of real estate mass appraisal, machine learning algorithms are often used, however, it is almost certainly not used in practice in Turkey. On the contrary, mass appraisal by machine learning algorithms will provide both more accurate, impartial, and fast results. For the first time in the literature, it was observed that Borst performed a mass appraisal study with Artificial Neural Networks (ANN) algorithm, which is a machine learning algorithm, for the first time in 1991. After this study, the number of studies using the ANN algorithm has increased. Researchers have observed in their scientific publications that the ANN algorithm gives good results in real estate mass appraisal studies. On the other hand, some researchers compared the ANN algorithm with some classical statistical methods (hedonic, regression) in their studies and observed that ANN did not contribute much. As can be understood from the literature, although there are many researchers who observe that ANN gives good results in mass appraisal, there are some other researchers who claims that the ANN algorithm does not give good results. The reason for this dilemma is due to the uncertainty and number of the variable, type of real estate the number of the real estates. In this study, a mass appraisal is performed by the ANN method, and the results were compared with the real values. In this study 180 number of flats and their 13 variables were collected in Yenimahalle, Ankara. The training data selected randomly from 75% total flats. The reaming data were used for test purposes. According to the results, a difference of at minimum 3,199TL, maximum 150,801 TL and averagely 50.000 TL were observed between the estimated value by the ANN algorithm and the real value. According to these results, the ANN algorithm is a successful method in mass appraisal. The results showed that ANN can be used for mass appraisal studies.
Traditional methods in many areas have been replaced by modern methods known as machine learning ... more Traditional methods in many areas have been replaced by modern methods known as machine learning with the rapidly developing technology and innovations in science. One of these areas is real estate valuation (appraisal) area. Real estate appraisal can be conducted on a single real estate as well as appraisal of more than one real estate together, which is called as mass appraisal, is possible. In this study, a mass appraisal is performed by a Random Forest Regression method, and the results were evaluated. For this purpose, data of 189 flats expected real value and their 13 variables were collected in Yenimahalle, Ankara. 75% of these data were used as training data and 25% as test data. According to the results, a difference of at minimum 600 TL, maximum 60.000 TL and averagely 25.000 TL were observed between the predicted value by the Random Forest regression and the expected real value. According to these results, random forest regression is a successful method in mass appraisal, and it is observed that valuation with different machine learning methods such as random forest regression has a positive effect on time and labor force comparing with valuation of real estate by traditional methods individually. Kitlesel Değerlemede Makine Öğrenme: Rasgele Orman Regresyonu Özet-Hızla gelişen teknoloji ve bilimde bulunan yenilikler ile birçok alanda geleneksel yöntemlerin yerini makine öğrenme diye anılan modern yöntemleri almıştır. Bu alanlardan biri ise gayrimenkul değerleme alanıdır. Gayrimenkuller tek başına değerlemesi yapılabileceği gibi kitlesel olarak ta birçok gayrimenkulün bir arada değerlemesinin yapılması mümkündür. Bu çalışmada, popüler bir makine öğrenme tekniği olan Random Forest (Rasgele Orman) Regresyonu yöntemi seçilerek gayrimenkullerin kitlesel değerlemesi yapılmış ve sonuçların gerçek değere yakınlığı incelenmiştir. Bu amaçla, Ankara İli Yenimahalle İlçesinde 189 adet apartman dairesine ait değer ve bu gayrimenkullere ait 13 adet değişken verisi toplanmıştır. Bu verinin, %75'i eğitim verisi ve % 25'i ise test verisi olarak kullanılmıştır. Elde edilen sonuçlara göre, tahmin edilen değer ile olması beklenilen değer arasında en az 600 TL, en fazla 60.000 TL ve ortalama 25.000 TL fark gözlemlenmiştir. Bu sonuçlara göre rasgele orman regresyonunun kitlesel değerlemede başarılı olduğu, geleneksel yöntemlerle gayrimenkul değerlemek yerine rasgele orman regresyonu gibi farklı makine öğrenme yöntemleriyle değerleme yapılmasının zaman ve insan gücü tasarrufu açısından pozitif etkilerinin olacağı ortaya konmuştur.
AUTOMATIC GROUND EXTRACTION FOR URBAN AREAS FROM AIRBORNE LIDAR DATA, 2020
Terrain models play a key role in many applications, such as hydrological modeling, volume calcul... more Terrain models play a key role in many applications, such as hydrological modeling, volume calculation, wire and pipeline route planning as well as many engineering applications. While terrain models can be generated from traditional data sources, an advanced and recently popular geospatial technology, Light Detection and Ranging (LiDAR) data, is also a source for generating high-density terrain models in the last decades. The main advantage of LiDAR technology over traditional data sources is that it generates 3D point clouds directly so that the representation of the surfaces is obtained fast. On the other hand, before terrain modeling, ground points need to be extracted by point labeling in the 3D point cloud. In this study, a new algorithm is proposed for automatic ground point extraction from airborne LiDAR data for urban areas. The proposed algorithm is mainly based on height information of the points in the dataset and labels ground points comparing height differences in local windows. The algorithm does not require any user input threshold and a neighborhood definition. The proposed ground extraction algorithm was tested with three different urban area LiDAR data. The quality control basically performed qualitatively by visual inspection and quantitatively by calculation of overall accuracy, which is conduct by comparing the proposed algorithm results with data provider's ground classification and Cloth Simulation Filtering (CSF) algorithm's results. The overall accuracy of the proposed algorithm is found between 95%-98%. The experimental results showed that the algorithm promises reliable results to extract ground points from airborne LiDAR data for urban areas.
The extraction of building roof planes from lidar data has become a popular research topic with r... more The extraction of building roof planes from lidar data has become a popular research topic with random sample consensus (RANSAC) being one of the most commonly adopted algorithms. RANSAC extracts full planes, which is problematic when there are other points outside the plane boundary but within the plane space. This study proposes an improved RANSAC (I-RANSAC) algorithm by removing points that do not belong to the roof plane. I-RANSAC selects a random point from the extracted roof plane and then searches for its neighbours within a given threshold to identify and remove outliers. The new algorithm was tested with 14 buildings from two datasets, where quality control measures showed significant improvement over standard RANSAC.
In this study, the Istanbul Province was monitored using Landsat 5 TM, MSS, Landsat 7 ETM+, and L... more In this study, the Istanbul Province was monitored using Landsat 5 TM, MSS, Landsat 7 ETM+, and Landsat 8 OLI imagery from the years 1986, 2000, 2009, 2011, 2013, and 2015 in order to assess land cover changes in the province. The aim of the study was to classify manmade structures, land, green, and water areas, and to observe the changes in the province using satellite images. After classification, the images were compared in selected years to observe land cover. Moreover, these changes were correlated with the property tax values of Istanbul by years. The findings of the study showed that manmade structure areas increased while vegetation areas decreased due to rapid population growth, urbanization, and industrial and commercial development in Istanbul. These changes also explain the transformation of land from rural and natural areas to residential use, and serve as a tool with which to assess land value increments. Land value capturing is critical for the analysis of the linkages between the changes in land cover, and for assessing land transformation and urban growth. Due to inadequate market data, real estate tax values were used to analyze the linkages between detection changes, land cover, and taxation. In fact, the declared tax values of land owners are generally lower than the actual market values and therefore it is not possible to transfer the value increasing of land in urban areas by using property taxation from the owner to local and central governments. The research results also show that the integration of remote sensing results with real estate market data give us to determine the tax base values of real estate more realistically.
Inland waters have vital importance in terrestrial ecosystems as they contribute to the total div... more Inland waters have vital importance in terrestrial ecosystems as they contribute to the total diversity in
surrounding areas as well as enhancing horizontal and vertical ecological connectivity of various habitats.
Therefore, temporal monitoring of changes on water bodies is crucial. Morphological changes of
inland waters can be conveniently determined using Remote Sensing (RS) techniques. For instance,
optical satellite images are widely used for change detection studies; however, it might be difficult to get
a proper optical image of an area of interest all the time as the area might be covered by clouds or haze.
Moreover, aerial images collected by an optical sensor mounted on aircraft can also be employed for
monitoring inland water change. On the other hand, optical images are two dimensional which makes
very difficult to detect changes in three dimensions such as for inland water bodies. Hence, alternative
technologies such as Light Detection and Ranging (LiDAR) which has direct and fast 3D data acquisition
can be used instead of images for change detection. Since LiDAR sensors are mounted on an aircraft, the
data collection time can be scheduled according to weather conditions for avoiding from rain and haze.
Therefore, in this study, LiDAR technology was chosen as the source of the data and two algorithms are
proposed for extracting boundary of inland water bodies. It is known that inland water bodies are
generally planar, as a first step of the proposed methodology, point cloud of the water surface was
extracted using RANdom SAmple Consensus (RANSAC) algorithm from LiDAR data. For the second step,
two algorithms were proposed for delineating of inland water surface boundary. The first algorithm is
the Angles of Points (AOP) which is mainly based on line and angle properties of point cloud. The second
algorithm is the LiDAR to Image (LTI) and it basically involves conversion of point clouds to binary images
for extraction of boundary of water bodies. The two algorithms for boundary extraction of water bodies
were tested with three different inland water bodies. Finally, the results produced by the algorithms
were compared to each other and with manually extracted boundaries for different time periods. The
experimental results showed that the proposed algorithms are capable of extracting the boundaries of
water bodies; however, the LTI algorithm performed better than the AOP when applied on water surfaces
with complex geometry.
COMPARISON OF MACHINE LEARNING ALGORITHMS FOR MASS APPRAISAL OF REAL ESTATE DATA, 2024
In recent years, machine learning algorithms have been used in the mass appraisal of real estate.... more In recent years, machine learning algorithms have been used in the mass appraisal of real estate. In this study, 5 machine learning algorithms are used for residential type real estate. Machine learning algorithms used for mass appraisal in this study are Artificial Neural Networks (ANN), Random Forest (RO), Multiple Regression Analysis (MRA), K-Nearest Neighborhood (k-nn), Support Vector Regression (SVR). To test the study, real estate data collected from the central districts of Ankara, were used. The main purpose of this study is to find out which machine learning algorithm gives the best results for the mass appraisal of real estates and to reveal the most important variables that affect the prices of real estate. According to the results obtained for the city of Ankara, it was observed that the best algorithm for mass appraisal is RF in residential-type real estates, followed by the ANN, k-nn, and linear regression algorithms, respectively. According to the results obtained from the residential real estate, it was concluded that heating and distances to places of importance had the greatest effect on the value. mass appraisal, machine learning algorithms, random forest, artificial neural network, real estate valuation map JEL Classification: R39
Urbanization Spatial and settlement development Dichotomy urban settlement Land cover change Land... more Urbanization Spatial and settlement development Dichotomy urban settlement Land cover change Land governance Research Article
Income from taxes constitutes the main revenue source of local and central governments in emergin... more Income from taxes constitutes the main revenue source of local and central governments in emerging economies. Urban services envisaged in development plans for rapidly growing urban populations are provided by local governments; however, available financial resources remain inadequate for effective, efficient and high quality service delivery. Despite the presence of low-density neighborhoods, especially in urban centers, the transformation of land to land lots and increasing residential density in the urban periphery place additional economic burden on local governments. In this context, a model proposal has been made in this study towards controlling by taxation, which is an application tool of land lot policy, of urban sprawl process, increasing local government revenues and reducing revenue loss to a minimum in the Ankara Metropolitan area, of which scale is gradually growing and governance efficiency is declining. The results of the evaluation carried out in a 59.7 hectare study area in the Çukurambar Region in the Çankaya District of Ankara Province have shown that it is possible to develop a real estate valuation system based on market value in order to transfer the real estate value increment resulting from development plan decisions to the public by taxation policies, and thus reduce tax and fee revenue losses resulting from tax base. It has been confirmed that the suggested tax policies will have a reducing effect on infertile land acquisition with value increment expectations and land possession trends; that local governments' tax revenues will increase by a factor of four and infrastructural investments financing might be provided by value increment financing. Using practices such as controlling the process of transformation of land to land lot, property tax, taxation of value increment, and reducing economic rent seekers' tendencies of acquiring unearned increment will be possible.
In this study, the Istanbul Province was monitored using Landsat 5 TM, MSS, Landsat 7 ETM+, and L... more In this study, the Istanbul Province was monitored using Landsat 5 TM, MSS, Landsat 7 ETM+, and Landsat 8 OLI imagery from the years 1986, 2000, 2009, 2011, 2013, and 2015 in order to assess land cover changes in the province. The aim of the study was to classify manmade structures, land, green, and water areas, and to observe the changes in the province using satellite images. After classification, the images were compared in selected years to observe land cover. Moreover, these changes were correlated with the property tax values of Istanbul by years. The findings of the study showed that manmade structure areas increased while vegetation areas decreased due to rapid population growth, urbanization, and industrial and commercial development in Istanbul. These changes also explain the transformation of land from rural and natural areas to residential use, and serve as a tool with which to assess land value increments. Land value capturing is critical for the analysis of the linkage...
ISPEC 9. ULUSLARARASI MÜHENDİSLİK VE FEN BİLİMLERİ KONGRESİ 13-15 Kasım 2020 Ankara, Türkiye, 2020
In recent years, studies performed by traditional methods in many branches of science have been r... more In recent years, studies performed by traditional methods in many branches of science have been replaced by the method called as machine learning. Effects such as speed and low performance observed with traditional methods are increased by using machine learning algorithms. In the field of real estate mass appraisal, machine learning algorithms are often used, however, it is almost certainly not used in practice in Turkey. On the contrary, mass appraisal by machine learning algorithms will provide both more accurate, impartial, and fast results. For the first time in the literature, it was observed that Borst performed a mass appraisal study with Artificial Neural Networks (ANN) algorithm, which is a machine learning algorithm, for the first time in 1991. After this study, the number of studies using the ANN algorithm has increased. Researchers have observed in their scientific publications that the ANN algorithm gives good results in real estate mass appraisal studies. On the other hand, some researchers compared the ANN algorithm with some classical statistical methods (hedonic, regression) in their studies and observed that ANN did not contribute much. As can be understood from the literature, although there are many researchers who observe that ANN gives good results in mass appraisal, there are some other researchers who claims that the ANN algorithm does not give good results. The reason for this dilemma is due to the uncertainty and number of the variable, type of real estate the number of the real estates. In this study, a mass appraisal is performed by the ANN method, and the results were compared with the real values. In this study 180 number of flats and their 13 variables were collected in Yenimahalle, Ankara. The training data selected randomly from 75% total flats. The reaming data were used for test purposes. According to the results, a difference of at minimum 3,199TL, maximum 150,801 TL and averagely 50.000 TL were observed between the estimated value by the ANN algorithm and the real value. According to these results, the ANN algorithm is a successful method in mass appraisal. The results showed that ANN can be used for mass appraisal studies.
Traditional methods in many areas have been replaced by modern methods known as machine learning ... more Traditional methods in many areas have been replaced by modern methods known as machine learning with the rapidly developing technology and innovations in science. One of these areas is real estate valuation (appraisal) area. Real estate appraisal can be conducted on a single real estate as well as appraisal of more than one real estate together, which is called as mass appraisal, is possible. In this study, a mass appraisal is performed by a Random Forest Regression method, and the results were evaluated. For this purpose, data of 189 flats expected real value and their 13 variables were collected in Yenimahalle, Ankara. 75% of these data were used as training data and 25% as test data. According to the results, a difference of at minimum 600 TL, maximum 60.000 TL and averagely 25.000 TL were observed between the predicted value by the Random Forest regression and the expected real value. According to these results, random forest regression is a successful method in mass appraisal, and it is observed that valuation with different machine learning methods such as random forest regression has a positive effect on time and labor force comparing with valuation of real estate by traditional methods individually. Kitlesel Değerlemede Makine Öğrenme: Rasgele Orman Regresyonu Özet-Hızla gelişen teknoloji ve bilimde bulunan yenilikler ile birçok alanda geleneksel yöntemlerin yerini makine öğrenme diye anılan modern yöntemleri almıştır. Bu alanlardan biri ise gayrimenkul değerleme alanıdır. Gayrimenkuller tek başına değerlemesi yapılabileceği gibi kitlesel olarak ta birçok gayrimenkulün bir arada değerlemesinin yapılması mümkündür. Bu çalışmada, popüler bir makine öğrenme tekniği olan Random Forest (Rasgele Orman) Regresyonu yöntemi seçilerek gayrimenkullerin kitlesel değerlemesi yapılmış ve sonuçların gerçek değere yakınlığı incelenmiştir. Bu amaçla, Ankara İli Yenimahalle İlçesinde 189 adet apartman dairesine ait değer ve bu gayrimenkullere ait 13 adet değişken verisi toplanmıştır. Bu verinin, %75'i eğitim verisi ve % 25'i ise test verisi olarak kullanılmıştır. Elde edilen sonuçlara göre, tahmin edilen değer ile olması beklenilen değer arasında en az 600 TL, en fazla 60.000 TL ve ortalama 25.000 TL fark gözlemlenmiştir. Bu sonuçlara göre rasgele orman regresyonunun kitlesel değerlemede başarılı olduğu, geleneksel yöntemlerle gayrimenkul değerlemek yerine rasgele orman regresyonu gibi farklı makine öğrenme yöntemleriyle değerleme yapılmasının zaman ve insan gücü tasarrufu açısından pozitif etkilerinin olacağı ortaya konmuştur.
AUTOMATIC GROUND EXTRACTION FOR URBAN AREAS FROM AIRBORNE LIDAR DATA, 2020
Terrain models play a key role in many applications, such as hydrological modeling, volume calcul... more Terrain models play a key role in many applications, such as hydrological modeling, volume calculation, wire and pipeline route planning as well as many engineering applications. While terrain models can be generated from traditional data sources, an advanced and recently popular geospatial technology, Light Detection and Ranging (LiDAR) data, is also a source for generating high-density terrain models in the last decades. The main advantage of LiDAR technology over traditional data sources is that it generates 3D point clouds directly so that the representation of the surfaces is obtained fast. On the other hand, before terrain modeling, ground points need to be extracted by point labeling in the 3D point cloud. In this study, a new algorithm is proposed for automatic ground point extraction from airborne LiDAR data for urban areas. The proposed algorithm is mainly based on height information of the points in the dataset and labels ground points comparing height differences in local windows. The algorithm does not require any user input threshold and a neighborhood definition. The proposed ground extraction algorithm was tested with three different urban area LiDAR data. The quality control basically performed qualitatively by visual inspection and quantitatively by calculation of overall accuracy, which is conduct by comparing the proposed algorithm results with data provider's ground classification and Cloth Simulation Filtering (CSF) algorithm's results. The overall accuracy of the proposed algorithm is found between 95%-98%. The experimental results showed that the algorithm promises reliable results to extract ground points from airborne LiDAR data for urban areas.
The extraction of building roof planes from lidar data has become a popular research topic with r... more The extraction of building roof planes from lidar data has become a popular research topic with random sample consensus (RANSAC) being one of the most commonly adopted algorithms. RANSAC extracts full planes, which is problematic when there are other points outside the plane boundary but within the plane space. This study proposes an improved RANSAC (I-RANSAC) algorithm by removing points that do not belong to the roof plane. I-RANSAC selects a random point from the extracted roof plane and then searches for its neighbours within a given threshold to identify and remove outliers. The new algorithm was tested with 14 buildings from two datasets, where quality control measures showed significant improvement over standard RANSAC.
In this study, the Istanbul Province was monitored using Landsat 5 TM, MSS, Landsat 7 ETM+, and L... more In this study, the Istanbul Province was monitored using Landsat 5 TM, MSS, Landsat 7 ETM+, and Landsat 8 OLI imagery from the years 1986, 2000, 2009, 2011, 2013, and 2015 in order to assess land cover changes in the province. The aim of the study was to classify manmade structures, land, green, and water areas, and to observe the changes in the province using satellite images. After classification, the images were compared in selected years to observe land cover. Moreover, these changes were correlated with the property tax values of Istanbul by years. The findings of the study showed that manmade structure areas increased while vegetation areas decreased due to rapid population growth, urbanization, and industrial and commercial development in Istanbul. These changes also explain the transformation of land from rural and natural areas to residential use, and serve as a tool with which to assess land value increments. Land value capturing is critical for the analysis of the linkages between the changes in land cover, and for assessing land transformation and urban growth. Due to inadequate market data, real estate tax values were used to analyze the linkages between detection changes, land cover, and taxation. In fact, the declared tax values of land owners are generally lower than the actual market values and therefore it is not possible to transfer the value increasing of land in urban areas by using property taxation from the owner to local and central governments. The research results also show that the integration of remote sensing results with real estate market data give us to determine the tax base values of real estate more realistically.
Inland waters have vital importance in terrestrial ecosystems as they contribute to the total div... more Inland waters have vital importance in terrestrial ecosystems as they contribute to the total diversity in
surrounding areas as well as enhancing horizontal and vertical ecological connectivity of various habitats.
Therefore, temporal monitoring of changes on water bodies is crucial. Morphological changes of
inland waters can be conveniently determined using Remote Sensing (RS) techniques. For instance,
optical satellite images are widely used for change detection studies; however, it might be difficult to get
a proper optical image of an area of interest all the time as the area might be covered by clouds or haze.
Moreover, aerial images collected by an optical sensor mounted on aircraft can also be employed for
monitoring inland water change. On the other hand, optical images are two dimensional which makes
very difficult to detect changes in three dimensions such as for inland water bodies. Hence, alternative
technologies such as Light Detection and Ranging (LiDAR) which has direct and fast 3D data acquisition
can be used instead of images for change detection. Since LiDAR sensors are mounted on an aircraft, the
data collection time can be scheduled according to weather conditions for avoiding from rain and haze.
Therefore, in this study, LiDAR technology was chosen as the source of the data and two algorithms are
proposed for extracting boundary of inland water bodies. It is known that inland water bodies are
generally planar, as a first step of the proposed methodology, point cloud of the water surface was
extracted using RANdom SAmple Consensus (RANSAC) algorithm from LiDAR data. For the second step,
two algorithms were proposed for delineating of inland water surface boundary. The first algorithm is
the Angles of Points (AOP) which is mainly based on line and angle properties of point cloud. The second
algorithm is the LiDAR to Image (LTI) and it basically involves conversion of point clouds to binary images
for extraction of boundary of water bodies. The two algorithms for boundary extraction of water bodies
were tested with three different inland water bodies. Finally, the results produced by the algorithms
were compared to each other and with manually extracted boundaries for different time periods. The
experimental results showed that the proposed algorithms are capable of extracting the boundaries of
water bodies; however, the LTI algorithm performed better than the AOP when applied on water surfaces
with complex geometry.
Building roof plane extraction from Light Detection And Ranging (LiDAR) data has
become very popu... more Building roof plane extraction from Light Detection And Ranging (LiDAR) data has become very popular study. One of most popular algorithms for planar feature extraction is Random Sample and Concencus (RANSAC). The RANSAC algorithm defines the planes in a continuous infinite planimetric space, and the points on the continuing plane planimetry, which are not within the bounds of the plane, are extracted as if they are in plane boundary. The aim of this study is to develop an algorithm for defining and eliminating the outliers from building roof planes, which are extracted using RANSAC algorithm to enhance/improve RANSAC plane extraction results. Hence, an algorithm was develop (called as Improved-RANSAC, I-RANSAC) to enhance RANSAC algorithm for planar feature extraction. To extract planar feature from Lidar data, ground and non-ground points need to be classified. Using only non-ground points from the whole LiDAR data, the proposed plane extraction algorithm (I-RANSAC) was tested with 8 single building LiDAR data and 3 LiDAR data sets that contain more than one buildings. Precision, Recall and Fmeasures are calculated and observed as around 0.95.
Uploads
surrounding areas as well as enhancing horizontal and vertical ecological connectivity of various habitats.
Therefore, temporal monitoring of changes on water bodies is crucial. Morphological changes of
inland waters can be conveniently determined using Remote Sensing (RS) techniques. For instance,
optical satellite images are widely used for change detection studies; however, it might be difficult to get
a proper optical image of an area of interest all the time as the area might be covered by clouds or haze.
Moreover, aerial images collected by an optical sensor mounted on aircraft can also be employed for
monitoring inland water change. On the other hand, optical images are two dimensional which makes
very difficult to detect changes in three dimensions such as for inland water bodies. Hence, alternative
technologies such as Light Detection and Ranging (LiDAR) which has direct and fast 3D data acquisition
can be used instead of images for change detection. Since LiDAR sensors are mounted on an aircraft, the
data collection time can be scheduled according to weather conditions for avoiding from rain and haze.
Therefore, in this study, LiDAR technology was chosen as the source of the data and two algorithms are
proposed for extracting boundary of inland water bodies. It is known that inland water bodies are
generally planar, as a first step of the proposed methodology, point cloud of the water surface was
extracted using RANdom SAmple Consensus (RANSAC) algorithm from LiDAR data. For the second step,
two algorithms were proposed for delineating of inland water surface boundary. The first algorithm is
the Angles of Points (AOP) which is mainly based on line and angle properties of point cloud. The second
algorithm is the LiDAR to Image (LTI) and it basically involves conversion of point clouds to binary images
for extraction of boundary of water bodies. The two algorithms for boundary extraction of water bodies
were tested with three different inland water bodies. Finally, the results produced by the algorithms
were compared to each other and with manually extracted boundaries for different time periods. The
experimental results showed that the proposed algorithms are capable of extracting the boundaries of
water bodies; however, the LTI algorithm performed better than the AOP when applied on water surfaces
with complex geometry.
surrounding areas as well as enhancing horizontal and vertical ecological connectivity of various habitats.
Therefore, temporal monitoring of changes on water bodies is crucial. Morphological changes of
inland waters can be conveniently determined using Remote Sensing (RS) techniques. For instance,
optical satellite images are widely used for change detection studies; however, it might be difficult to get
a proper optical image of an area of interest all the time as the area might be covered by clouds or haze.
Moreover, aerial images collected by an optical sensor mounted on aircraft can also be employed for
monitoring inland water change. On the other hand, optical images are two dimensional which makes
very difficult to detect changes in three dimensions such as for inland water bodies. Hence, alternative
technologies such as Light Detection and Ranging (LiDAR) which has direct and fast 3D data acquisition
can be used instead of images for change detection. Since LiDAR sensors are mounted on an aircraft, the
data collection time can be scheduled according to weather conditions for avoiding from rain and haze.
Therefore, in this study, LiDAR technology was chosen as the source of the data and two algorithms are
proposed for extracting boundary of inland water bodies. It is known that inland water bodies are
generally planar, as a first step of the proposed methodology, point cloud of the water surface was
extracted using RANdom SAmple Consensus (RANSAC) algorithm from LiDAR data. For the second step,
two algorithms were proposed for delineating of inland water surface boundary. The first algorithm is
the Angles of Points (AOP) which is mainly based on line and angle properties of point cloud. The second
algorithm is the LiDAR to Image (LTI) and it basically involves conversion of point clouds to binary images
for extraction of boundary of water bodies. The two algorithms for boundary extraction of water bodies
were tested with three different inland water bodies. Finally, the results produced by the algorithms
were compared to each other and with manually extracted boundaries for different time periods. The
experimental results showed that the proposed algorithms are capable of extracting the boundaries of
water bodies; however, the LTI algorithm performed better than the AOP when applied on water surfaces
with complex geometry.
become very popular study. One of most popular algorithms for planar feature extraction is
Random Sample and Concencus (RANSAC). The RANSAC algorithm defines the planes in
a continuous infinite planimetric space, and the points on the continuing plane planimetry,
which are not within the bounds of the plane, are extracted as if they are in plane boundary.
The aim of this study is to develop an algorithm for defining and eliminating the outliers
from building roof planes, which are extracted using RANSAC algorithm to
enhance/improve RANSAC plane extraction results. Hence, an algorithm was develop
(called as Improved-RANSAC, I-RANSAC) to enhance RANSAC algorithm for planar
feature extraction. To extract planar feature from Lidar data, ground and non-ground points
need to be classified. Using only non-ground points from the whole LiDAR data, the
proposed plane extraction algorithm (I-RANSAC) was tested with 8 single building LiDAR
data and 3 LiDAR data sets that contain more than one buildings. Precision, Recall and Fmeasures
are calculated and observed as around 0.95.