Publication: 3b Lidar İle Üretilen Model Doğruluğunun Araştırılması
Abstract
Bu çalışmada, 3B LiDAR sensörü, mobil bir insansız kara aracına entegre edilerek bir iç mekânın taranması gerçekleştirilmiştir. Elde edilen nokta bulutu verileri, LiDAR tabanlı konumlandırma ve haritalama görevlerinde yaygın olarak kullanılan HDL-Graph-SLAM algoritması ile işlenmiş ve 3B çevre modeli oluşturulmuştur. Bu çalışmanın temel amacı, elde edilen 3B modelin doğruluğunu test etmektir. Bu kapsamda iki farklı doğruluk analizi gerçekleştirilmiştir. İlk analiz türü olan nesne temelli ölçüm analizinde, çalışma alanında yer alan 26 farklı nesnenin boyutları, yüksek hassasiyetli Total Station cihazı ile fiziksel olarak ölçülmüş ve bu veriler, LiDAR ile elde edilen ölçümlerle karşılaştırılmıştır. 'En' ve 'boy' yönlerindeki ortalama farkların sırasıyla ±0.012 m ve ±0.011 m olarak hesaplanması, sistemin 1-2 cm seviyesinde yüksek doğruluk sunduğunu göstermiştir. İkinci analiz olan genel yüzey doğruluğu analizinde ise, LiDAR tabanlı SLAM yöntemiyle oluşturulan 3B model, lazer tarayıcıdan elde edilen referans 3B model ile karşılaştırılmıştır. Bu kapsamda, 7 sol duvar, 7 sağ duvar ve 10 zemin yüzeyi olmak üzere toplam 24 yüzey incelenmiş; LiDAR-SLAM verileri ile referans veriler arasındaki farklar RMSE, ortalama mesafe ve standart sapma metrikleri kullanılarak değerlendirilmiştir. Analiz sonuçlarına göre, tüm yüzeyler için ortalama RMSE değeri ±2.12 cm, ortalama mesafe ±4.16 cm ve standart sapma değeri ±1.06 cm olarak hesaplanmıştır. Elde edilen sonuçlar, LiDAR-SLAM tekniğinin iç mekân haritalama ve çevresel modelleme gibi uygulamalarda yüksek doğruluk sunduğunu ve pratik bir çözüm alternatifi oluşturduğunu göstermektedir. Bununla birlikte, daha ileri düzey hassasiyet gerektiren özel mühendislik uygulamaları için, sistemin ilave tarama yöntemleri veya destekleyici teknolojilerle birlikte kullanılması önerilebilir. Bu çalışma, LiDAR-SLAM tekniğinin iç mekân modelleme potansiyelini ortaya koyarken, uygulama şartlarına bağlı olarak karşılaşılabilecek sınırlamaların da göz önünde bulundurulması gerektiğini vurgulamaktadır.
In this study, a 3D LiDAR sensor was integrated into a mobile unmanned ground vehicle to scan a specific indoor environment. The acquired point cloud data were processed using the HDL-Graph-SLAM algorithm, which is widely used in LiDAR-based localization and mapping tasks, and a 3D environmental model was created. The primary aim of this study is to assess the accuracy of the generated 3D model. For this purpose, two different accuracy analyses were conducted. In the first type of analysis, object-based measurement analysis, the dimensions of 26 different objects located within the study area were measured with a high-precision Total Station device, and these measurements were compared with those obtained from LiDAR-SLAM data. The calculated mean differences of ±0.012 m in width and ±0.011 m in length demonstrated that the system provides 1-2 cm accuracy. In the second analysis, general surface accuracy analysis, the 3D model created using the LiDAR-based SLAM method was compared with a reference 3D model obtained from a laser scanner. In this context, a total of 24 surfaces, including 7 left walls, 7 right walls, and 10 floor surfaces, were examined, and the differences between the LiDAR data and the reference data were evaluated using RMSE, mean distance, and standard deviation metrics. According to the analysis results, the average RMSE value for all surfaces was calculated as ±2.12 cm, the mean distance as ±4.16 cm, and the standard deviation as ±1.06 cm. The obtained results indicate that LiDAR-based SLAM provides high accuracy and offers a practical solution for applications such as indoor mapping and environmental modeling. However, for specialized engineering applications requiring higher levels of precision, the system is recommended to be used in conjunction with additional scanning methods or supporting technologies. This study highlights the potential of LiDAR technology for indoor modeling while emphasizing the need to consider the limitations that may arise depending on the application conditions.
In this study, a 3D LiDAR sensor was integrated into a mobile unmanned ground vehicle to scan a specific indoor environment. The acquired point cloud data were processed using the HDL-Graph-SLAM algorithm, which is widely used in LiDAR-based localization and mapping tasks, and a 3D environmental model was created. The primary aim of this study is to assess the accuracy of the generated 3D model. For this purpose, two different accuracy analyses were conducted. In the first type of analysis, object-based measurement analysis, the dimensions of 26 different objects located within the study area were measured with a high-precision Total Station device, and these measurements were compared with those obtained from LiDAR-SLAM data. The calculated mean differences of ±0.012 m in width and ±0.011 m in length demonstrated that the system provides 1-2 cm accuracy. In the second analysis, general surface accuracy analysis, the 3D model created using the LiDAR-based SLAM method was compared with a reference 3D model obtained from a laser scanner. In this context, a total of 24 surfaces, including 7 left walls, 7 right walls, and 10 floor surfaces, were examined, and the differences between the LiDAR data and the reference data were evaluated using RMSE, mean distance, and standard deviation metrics. According to the analysis results, the average RMSE value for all surfaces was calculated as ±2.12 cm, the mean distance as ±4.16 cm, and the standard deviation as ±1.06 cm. The obtained results indicate that LiDAR-based SLAM provides high accuracy and offers a practical solution for applications such as indoor mapping and environmental modeling. However, for specialized engineering applications requiring higher levels of precision, the system is recommended to be used in conjunction with additional scanning methods or supporting technologies. This study highlights the potential of LiDAR technology for indoor modeling while emphasizing the need to consider the limitations that may arise depending on the application conditions.
Description
Citation
WoS Q
Scopus Q
Source
Volume
Issue
Start Page
End Page
89
