2017 International Artificial Intelligence and Data Processing Symposium (IDAP), 2017
Kinect sensörler yakin geçmişte oyun konsollari ile birlikte kendini göstersede nesne takip, pate... more Kinect sensörler yakin geçmişte oyun konsollari ile birlikte kendini göstersede nesne takip, patern tanima, nesne ebat kontrolü, engel tanima ve iç mekan haritalama gibi bir çok mühendislik alanida kullanilmaktadir. Kinect sensörü içerdiği RBG kamera ve IR kamera ile ayni anda iki farkli kameradan veri alarak nesneye ait farkli özelliklerin kayit edilmesini sağlamaktadir. Bu bildiride RGB-D sensörlerin iç mekan haritalamada sağladiği doğruluk hem görsel hem de istatistiksel olarak sunulmuştur. RGB-D sensörler ile elde edilen nokta bulutu gaussian, median, ortalama ve Diferansiyel Arama Algoritmasi(DSA) tabanli yüzey uydurma filtresi filtrelenmiştir. Filtrelenen verilen yersel lazer tarayici ile elde edilen nokta bulutu ile çakişitirilmiş, çakiştirma sonucundaki standart sapmalar belirlenmiş ve çakiştirmada meydana gelen hata yüzeyleri ise görsel olarak karşilaştirilmiştir. Sonuçlara göre RGB-D sensörlerle elde edilen ortam haritasi her ne kadar zaman maliyeti yüksek olsada çok hassa...
2017 8th International Conference on Information Technology (ICIT), 2017
Image edge information is very important in application areas such as machine learning, image pro... more Image edge information is very important in application areas such as machine learning, image processing, stereo vision, object tracking and pattern recognition. Intensity discontinuities or sudden intensity changes in a region are indicative of the edge region in that region. Although there are many approaches to detecting edge, generally intensity discontinuities or sudden intensity changes in a region are described as edge. In this study, we proposed a Backtracking Search (BSA) clustering based edge detection approach for noisy images. Proposed approach has two stages. In first stage, the edge map is calculated using the max-min filter defined in a window. In second stage, edge map is calculated via BSA based clustering with using a cost function.
Abstract Solar photovoltaic (PV) technologies receive investment, support, and incentives despite... more Abstract Solar photovoltaic (PV) technologies receive investment, support, and incentives despite their apparent high costs. In terms of renewable energy policies and spatial planning, site selection for solar PV farms is a critical issue. A novel comprehensive framework for assessing the site suitability of solar PV farms is proposed, which considers the preservation of natural, ecological conservation, and cultural areas. The framework combines a Geographical Information System (GIS) with layers of satellite-derived data for energy resources as well as locally collected data, and the Multi Criteria Decision Making (MCDM) method based on the Analytic Hierarchy Process (AHP). In the GIS environment, maps belonging to fourteen sub-criteria of the three main criteria (Geography, Climate and Location) for Kahramanmaras, Turkey were obtained, as well as suitability map derived from their weighted sum. The AHP method generates the appropriate weight for each input criteria using a pairwise comparison matrix. GHI, aspect, distance from power line network, land use/cover, annual average temperature, and others were found as the most necessary sub-criteria, respectively. According to the findings, 9.62% of study area very low, 20.15% is low, 22.51% is moderate, 23.98% is high and 23.74% is very high, and 26% is unsuitable for solar PV farms. Based on the results, the northern districts of the study region were determined to be the most suitable for the construction of solar PV farms. When existing five PV farms' location selection decisions were examined, it was observed that the investment outcomes were consistent with the results of the study. The proposed framework is projected to minimize the cost, time, and resources used on the construction of solar PV farms.
Camera calibration is an avoidable process for computational vision applications, such as 3D reve... more Camera calibration is an avoidable process for computational vision applications, such as 3D reverse engineering, industrial robot calibration, optic-pattern recognition, simultaneous localization and mapping, autonomous visual-driving and photogrammetric vision. The camera calibration problem is too complex, nonlinear and multimodal. Traditional camera calibration methods using gradient-based optimization often trap to one of the many local solutions available. Accurate computation ability of traditional camera calibration methods is limited since they use gradient-based optimization methods. Since evolutionary computing algorithms can avoid local solutions of numerical problems, they have the potential to accurately compute the required camera calibration parameters for high-precision computational vision applications. In this paper, the camera calibration parameters are computed by using 11 evolutionary computing algorithms, i.e., WDE, ABC, PSO, COBIDE, DE, CS, GWO, TLBO, MVMO, FOA and LSHADE. In order to make unbiased evaluation of the camera calibration results provided by the related evolutionary computing algorithms, two gradientbased traditional camera calibration methods, i.e., Zhang and Bouguet, have been used in the conducted experiments in this paper. The camera calibration results of the related methods were used to model a 3D physical test scene by using Structure from Motion photogrammetry method. The reference data set of the related 3D physical scene has been captured by using a 3D terrestrial laser scanner. Statistical comparison of the camera calibration results exposed that WDE supplies statistically better results than other comparison algorithms.
VII. UZAKTAN ALGILAMA VE CBS SEMPOZYUMU UZAL-CBS2018, 2018
Yapay sinir ağları makine öğrenimi alanındaki birçok problemde sıklıkla kullanılmaktadır. Yapay s... more Yapay sinir ağları makine öğrenimi alanındaki birçok problemde sıklıkla kullanılmaktadır. Yapay sinir ağları ilk ortaya çıktıklarında donanım yetersizliği sebebiyle çok fazla ilgi görememiş olsa da günümüzde GPU teknolojisindeki önemli gelişmeler sayesinde birçok farklı disiplin tarafından kullanılmaktadır. Örüntü tanıma, medikal uygulamalar, ses tanıma, otonom sürüş ve görüntü sınıflandırma gibi geniş bir yelpazede makine öğrenimi uygulamaları yapılmaktadır. Bütün bu çalışma alanlarının hepsinin ortak problemi verinin ne şekilde temsil edildiğine bağlı olarak değişen performanslarıdır. Bundan dolayı veriyi en iyi şekilde temsil edecek özniteliklerin çıkartılarak daha iyi sınıflandırma performansı sağlanabilmektedir. Derin öğrenme yöntemleri herhangi bir veri ön işlemesi yapmadan veriyi en iyi temsil eden öznitelikleri çıkartır. Yığınlanmış Autoencoder yapısı giriş verisinden yeni öznitelik bilgisi üreten Oto-Kodlayıcılardan (OK) oluşur. OKlar giriş verisini gizli katmanlardan geçirerek özyinelemeli olarak farklı bir şekilde temsil edilmesini sağlarlar. Problemlerin tipine bağlı olarak bir veya birden fazla OKlar birbiri arkasına bağlanabilir. OK ile işlenen verinin sınıflandırılması için kullanılan ve mantıksal regresyonun genişletilmiş bir formu olan Softmax, Yığınlanmış Autoencoder yapısının son adımı olarak kullanılabilmektedir. Bu çalışmada, çalışma alanı olarak AVIRIS sensörü ile elde edilen 16 farklı sınıf etiket değerine sahip, 200 bantlı Indian Pines ve 7 farklı sınıf etiket değerine sahip Salinas-A veri seti kullanılmıştır. Çalışma kapsamında iki katmanlı Yığınlanmış Oto-Kodlayıcı yapısı kullanılarak rastgele belirlenen test alanındaki sınıflandırma doğruluğunu en fazla yapan parametre (Saklı katman sayıları, epok sayıları, transfer fonksiyonları) değerleri araştırılmıştır. Elde edilen sonuçlar kendi içinde değerlendirildikten sonra Destek Vektör Makineleri, Karar ağaçları, Diskriminant Analizi, k-En Yakın Komşuluk, Torbalama-Karar Ağaçları ve Rastgele Alt Uzay k-En Yakın Komşuluk yöntemleri ile elde edilen test doğruluğu ile karşılaştırılmıştır. Derin öğrenme yapısı donanıma bağlı olarak oldukça yavaş çalışmasından dolayı parametre belirleme işlemi oldukça zordur. Buna karşın GPU teknolojisindeki hızlı gelişmeler bu durumu tersine çevirmektedir. Yapılan deneysel çalışmalar sonucunda Yığınlanmış Oto-Kodlayıcı tekniğinin diğer yöntemlerden daha başarılı sonuçlar elde ettiği görülmüştür.
In this paper, weighted differential evolution algorithm (WDE) has been proposed for solving real... more In this paper, weighted differential evolution algorithm (WDE) has been proposed for solving real-valued numerical optimization problems. When all parameters of WDE are determined randomly, in practice, WDE has no control parameter but the pattern size. WDE can solve unimodal, multimodal, separable, scalable, and hybrid problems. WDE has a very fast and quite simple structure, in addition, it can be parallelized due to its non-recursive nature. WDE has a strong exploration and exploitation capability. In this paper, WDE's success in solving CEC' 2013 problems was compared to 4 different EAs (i.e., CS, ABC, JADE, and BSA) statistically. One 3D geometric optimization problem (i.e., GPS network adjustment problem) and 4 constrained engineering design problems were used to examine the WDE's ability to solve real-world problems. Results obtained from the performed tests showed that, in general, problem-solving success of WDE is statistically better than the comparison algorithms that have been used in this paper.
Environmental Science and Pollution Research, 2021
The fact that traditional energy sources have limited reserves and have a negative impact on the ... more The fact that traditional energy sources have limited reserves and have a negative impact on the environment increases the demand for renewable energy sources. Environmental, economic, and sustainability concerns have led researchers, investors, and policy makers to seek the potential of renewable energy sources. Suitable site selection for new-generation renewable resources is vital in large-scale projects. In this paper, geographic information systems and multicriteria decision-making (MCDM) methods were integrated to exploit and construct the best location for solar PV power plants in Kayseri, Turkey. Three main, twelve subcriteria, and their indicators related to the study area were determined. The rank-based (i.e., rank sum, rank reciprocal weights, and rank order centroid weights) and analytical hierarchical process (AHP) MCDM methods were used to determine the weights of the criteria. Thirty-three existing solar PV power plants were used to verify the success of MCDM methods. Four MCDM methods yielded effective results according to the proposed criteria, and most of the existing solar PV power plants match the convenient regions on the suitability map provided by geographic information systems-based rank reciprocal method. In addition, according to the experimental study, Sarız, Tomarza, and Incesu districts of Kayseri were the most suitable sites.
Environmental Science and Pollution Research, 2021
The fact that traditional energy sources have limited reserves and have a negative impact on the ... more The fact that traditional energy sources have limited reserves and have a negative impact on the environment increases the demand for renewable energy sources. Environmental, economic, and sustainability concerns have led researchers, investors, and policy makers to seek the potential of renewable energy sources. Suitable site selection for new-generation renewable resources is vital in large-scale projects. In this paper, geographic information systems and multicriteria decision-making (MCDM) methods were integrated to exploit and construct the best location for solar PV power plants in Kayseri, Turkey. Three main, twelve subcriteria, and their indicators related to the study area were determined. The rank-based (i.e., rank sum, rank reciprocal weights, and rank order centroid weights) and analytical hierarchical process (AHP) MCDM methods were used to determine the weights of the criteria. Thirty-three existing solar PV power plants were used to verify the success of MCDM methods. Four MCDM methods yielded effective results according to the proposed criteria, and most of the existing solar PV power plants match the convenient regions on the suitability map provided by geographic information systems-based rank reciprocal method. In addition, according to the experimental study, Sarız, Tomarza, and Incesu districts of Kayseri were the most suitable sites.
2017 International Artificial Intelligence and Data Processing Symposium (IDAP), 2017
Kinect sensörler yakin geçmişte oyun konsollari ile birlikte kendini göstersede nesne takip, pate... more Kinect sensörler yakin geçmişte oyun konsollari ile birlikte kendini göstersede nesne takip, patern tanima, nesne ebat kontrolü, engel tanima ve iç mekan haritalama gibi bir çok mühendislik alanida kullanilmaktadir. Kinect sensörü içerdiği RBG kamera ve IR kamera ile ayni anda iki farkli kameradan veri alarak nesneye ait farkli özelliklerin kayit edilmesini sağlamaktadir. Bu bildiride RGB-D sensörlerin iç mekan haritalamada sağladiği doğruluk hem görsel hem de istatistiksel olarak sunulmuştur. RGB-D sensörler ile elde edilen nokta bulutu gaussian, median, ortalama ve Diferansiyel Arama Algoritmasi(DSA) tabanli yüzey uydurma filtresi filtrelenmiştir. Filtrelenen verilen yersel lazer tarayici ile elde edilen nokta bulutu ile çakişitirilmiş, çakiştirma sonucundaki standart sapmalar belirlenmiş ve çakiştirmada meydana gelen hata yüzeyleri ise görsel olarak karşilaştirilmiştir. Sonuçlara göre RGB-D sensörlerle elde edilen ortam haritasi her ne kadar zaman maliyeti yüksek olsada çok hassa...
2017 8th International Conference on Information Technology (ICIT), 2017
Image edge information is very important in application areas such as machine learning, image pro... more Image edge information is very important in application areas such as machine learning, image processing, stereo vision, object tracking and pattern recognition. Intensity discontinuities or sudden intensity changes in a region are indicative of the edge region in that region. Although there are many approaches to detecting edge, generally intensity discontinuities or sudden intensity changes in a region are described as edge. In this study, we proposed a Backtracking Search (BSA) clustering based edge detection approach for noisy images. Proposed approach has two stages. In first stage, the edge map is calculated using the max-min filter defined in a window. In second stage, edge map is calculated via BSA based clustering with using a cost function.
Abstract Solar photovoltaic (PV) technologies receive investment, support, and incentives despite... more Abstract Solar photovoltaic (PV) technologies receive investment, support, and incentives despite their apparent high costs. In terms of renewable energy policies and spatial planning, site selection for solar PV farms is a critical issue. A novel comprehensive framework for assessing the site suitability of solar PV farms is proposed, which considers the preservation of natural, ecological conservation, and cultural areas. The framework combines a Geographical Information System (GIS) with layers of satellite-derived data for energy resources as well as locally collected data, and the Multi Criteria Decision Making (MCDM) method based on the Analytic Hierarchy Process (AHP). In the GIS environment, maps belonging to fourteen sub-criteria of the three main criteria (Geography, Climate and Location) for Kahramanmaras, Turkey were obtained, as well as suitability map derived from their weighted sum. The AHP method generates the appropriate weight for each input criteria using a pairwise comparison matrix. GHI, aspect, distance from power line network, land use/cover, annual average temperature, and others were found as the most necessary sub-criteria, respectively. According to the findings, 9.62% of study area very low, 20.15% is low, 22.51% is moderate, 23.98% is high and 23.74% is very high, and 26% is unsuitable for solar PV farms. Based on the results, the northern districts of the study region were determined to be the most suitable for the construction of solar PV farms. When existing five PV farms' location selection decisions were examined, it was observed that the investment outcomes were consistent with the results of the study. The proposed framework is projected to minimize the cost, time, and resources used on the construction of solar PV farms.
Camera calibration is an avoidable process for computational vision applications, such as 3D reve... more Camera calibration is an avoidable process for computational vision applications, such as 3D reverse engineering, industrial robot calibration, optic-pattern recognition, simultaneous localization and mapping, autonomous visual-driving and photogrammetric vision. The camera calibration problem is too complex, nonlinear and multimodal. Traditional camera calibration methods using gradient-based optimization often trap to one of the many local solutions available. Accurate computation ability of traditional camera calibration methods is limited since they use gradient-based optimization methods. Since evolutionary computing algorithms can avoid local solutions of numerical problems, they have the potential to accurately compute the required camera calibration parameters for high-precision computational vision applications. In this paper, the camera calibration parameters are computed by using 11 evolutionary computing algorithms, i.e., WDE, ABC, PSO, COBIDE, DE, CS, GWO, TLBO, MVMO, FOA and LSHADE. In order to make unbiased evaluation of the camera calibration results provided by the related evolutionary computing algorithms, two gradientbased traditional camera calibration methods, i.e., Zhang and Bouguet, have been used in the conducted experiments in this paper. The camera calibration results of the related methods were used to model a 3D physical test scene by using Structure from Motion photogrammetry method. The reference data set of the related 3D physical scene has been captured by using a 3D terrestrial laser scanner. Statistical comparison of the camera calibration results exposed that WDE supplies statistically better results than other comparison algorithms.
VII. UZAKTAN ALGILAMA VE CBS SEMPOZYUMU UZAL-CBS2018, 2018
Yapay sinir ağları makine öğrenimi alanındaki birçok problemde sıklıkla kullanılmaktadır. Yapay s... more Yapay sinir ağları makine öğrenimi alanındaki birçok problemde sıklıkla kullanılmaktadır. Yapay sinir ağları ilk ortaya çıktıklarında donanım yetersizliği sebebiyle çok fazla ilgi görememiş olsa da günümüzde GPU teknolojisindeki önemli gelişmeler sayesinde birçok farklı disiplin tarafından kullanılmaktadır. Örüntü tanıma, medikal uygulamalar, ses tanıma, otonom sürüş ve görüntü sınıflandırma gibi geniş bir yelpazede makine öğrenimi uygulamaları yapılmaktadır. Bütün bu çalışma alanlarının hepsinin ortak problemi verinin ne şekilde temsil edildiğine bağlı olarak değişen performanslarıdır. Bundan dolayı veriyi en iyi şekilde temsil edecek özniteliklerin çıkartılarak daha iyi sınıflandırma performansı sağlanabilmektedir. Derin öğrenme yöntemleri herhangi bir veri ön işlemesi yapmadan veriyi en iyi temsil eden öznitelikleri çıkartır. Yığınlanmış Autoencoder yapısı giriş verisinden yeni öznitelik bilgisi üreten Oto-Kodlayıcılardan (OK) oluşur. OKlar giriş verisini gizli katmanlardan geçirerek özyinelemeli olarak farklı bir şekilde temsil edilmesini sağlarlar. Problemlerin tipine bağlı olarak bir veya birden fazla OKlar birbiri arkasına bağlanabilir. OK ile işlenen verinin sınıflandırılması için kullanılan ve mantıksal regresyonun genişletilmiş bir formu olan Softmax, Yığınlanmış Autoencoder yapısının son adımı olarak kullanılabilmektedir. Bu çalışmada, çalışma alanı olarak AVIRIS sensörü ile elde edilen 16 farklı sınıf etiket değerine sahip, 200 bantlı Indian Pines ve 7 farklı sınıf etiket değerine sahip Salinas-A veri seti kullanılmıştır. Çalışma kapsamında iki katmanlı Yığınlanmış Oto-Kodlayıcı yapısı kullanılarak rastgele belirlenen test alanındaki sınıflandırma doğruluğunu en fazla yapan parametre (Saklı katman sayıları, epok sayıları, transfer fonksiyonları) değerleri araştırılmıştır. Elde edilen sonuçlar kendi içinde değerlendirildikten sonra Destek Vektör Makineleri, Karar ağaçları, Diskriminant Analizi, k-En Yakın Komşuluk, Torbalama-Karar Ağaçları ve Rastgele Alt Uzay k-En Yakın Komşuluk yöntemleri ile elde edilen test doğruluğu ile karşılaştırılmıştır. Derin öğrenme yapısı donanıma bağlı olarak oldukça yavaş çalışmasından dolayı parametre belirleme işlemi oldukça zordur. Buna karşın GPU teknolojisindeki hızlı gelişmeler bu durumu tersine çevirmektedir. Yapılan deneysel çalışmalar sonucunda Yığınlanmış Oto-Kodlayıcı tekniğinin diğer yöntemlerden daha başarılı sonuçlar elde ettiği görülmüştür.
In this paper, weighted differential evolution algorithm (WDE) has been proposed for solving real... more In this paper, weighted differential evolution algorithm (WDE) has been proposed for solving real-valued numerical optimization problems. When all parameters of WDE are determined randomly, in practice, WDE has no control parameter but the pattern size. WDE can solve unimodal, multimodal, separable, scalable, and hybrid problems. WDE has a very fast and quite simple structure, in addition, it can be parallelized due to its non-recursive nature. WDE has a strong exploration and exploitation capability. In this paper, WDE's success in solving CEC' 2013 problems was compared to 4 different EAs (i.e., CS, ABC, JADE, and BSA) statistically. One 3D geometric optimization problem (i.e., GPS network adjustment problem) and 4 constrained engineering design problems were used to examine the WDE's ability to solve real-world problems. Results obtained from the performed tests showed that, in general, problem-solving success of WDE is statistically better than the comparison algorithms that have been used in this paper.
Environmental Science and Pollution Research, 2021
The fact that traditional energy sources have limited reserves and have a negative impact on the ... more The fact that traditional energy sources have limited reserves and have a negative impact on the environment increases the demand for renewable energy sources. Environmental, economic, and sustainability concerns have led researchers, investors, and policy makers to seek the potential of renewable energy sources. Suitable site selection for new-generation renewable resources is vital in large-scale projects. In this paper, geographic information systems and multicriteria decision-making (MCDM) methods were integrated to exploit and construct the best location for solar PV power plants in Kayseri, Turkey. Three main, twelve subcriteria, and their indicators related to the study area were determined. The rank-based (i.e., rank sum, rank reciprocal weights, and rank order centroid weights) and analytical hierarchical process (AHP) MCDM methods were used to determine the weights of the criteria. Thirty-three existing solar PV power plants were used to verify the success of MCDM methods. Four MCDM methods yielded effective results according to the proposed criteria, and most of the existing solar PV power plants match the convenient regions on the suitability map provided by geographic information systems-based rank reciprocal method. In addition, according to the experimental study, Sarız, Tomarza, and Incesu districts of Kayseri were the most suitable sites.
Environmental Science and Pollution Research, 2021
The fact that traditional energy sources have limited reserves and have a negative impact on the ... more The fact that traditional energy sources have limited reserves and have a negative impact on the environment increases the demand for renewable energy sources. Environmental, economic, and sustainability concerns have led researchers, investors, and policy makers to seek the potential of renewable energy sources. Suitable site selection for new-generation renewable resources is vital in large-scale projects. In this paper, geographic information systems and multicriteria decision-making (MCDM) methods were integrated to exploit and construct the best location for solar PV power plants in Kayseri, Turkey. Three main, twelve subcriteria, and their indicators related to the study area were determined. The rank-based (i.e., rank sum, rank reciprocal weights, and rank order centroid weights) and analytical hierarchical process (AHP) MCDM methods were used to determine the weights of the criteria. Thirty-three existing solar PV power plants were used to verify the success of MCDM methods. Four MCDM methods yielded effective results according to the proposed criteria, and most of the existing solar PV power plants match the convenient regions on the suitability map provided by geographic information systems-based rank reciprocal method. In addition, according to the experimental study, Sarız, Tomarza, and Incesu districts of Kayseri were the most suitable sites.
Uploads
Papers by Akif Gunen