• Proceedings of the Korean Society of Surveying, Geodesy, Photogrammetry, and Cartography Conference
• /
• 2007.04a
• /
• pp.273-276
• /
• 2007

In this study, we generated DSM(Digital Surface Model)s and orthophotos with both LIDAR data and scanned aerial photos and compared them with those generated from only the scanned photos. We checked the relief displacements of buildings appearing in the generated orthophotos, where the displacement should not be exist in a true-orthophoto. The RMSE of the relief displacement in the orthophoto generated using a LIDAR DSM is 3 m while the RMSE in the orthophotos from a DSM based on the image matching is 6.1 m. It was revealed that the orthophoto from a LIDAR DSM are closer to a true-orthophoto. But the results in the accuracy test and similarity evaluation of the generated orthophotos were contrary to former results because the roof texture of buildings were expanded to occlusion areas around the buildings. With the central area of the photo, we can generate sufficiently accurate true-orthophotos using a LIDAR DSM.

• Proceedings of the KSRS Conference
• /
• 2003.11a
• /
• pp.885-887
• /
• 2003

The purpose of this investigation is to generate true orthophotos from high resolution satellite images. The major works of this research include 4 parts: (1) determination of orientation parameters, (2) generating traditional orthophotos using terrain model, (3) relief correction for buildings, and (4) process for hidden areas. To determine the position of satellites, we correct the onboard orientation parameters to fine tune the orbit. In the generation of traditional orthophotos, we employ orientation parameters and digital terrain model(DTM) to rectify tilt displacements and relief displacements for terrain. We, then, compute relief displacements for buildings with digital building model (DBM). To avoid double mapping, we detect hidden areas. Due to the satellite’s small field of view, an efficient method for the detection of hidden areas and building rectification will be proposed in this paper. Test areas cover the city of Kaohsiung in southern Taiwan. Test images are from the QuickBird satellite.

• Korean Journal of Geomatics
• /
• v.2 no.2
• /
• pp.131-137
• /
• 2002

The digital photogrammetric products have been increasingly used as an accurate foundation for representing information associated with infrastructure management. The technological advances in merging raster and vector data within the framework of GIS have allowed for the inclusion of DTMs and digital orthophotos with vector data and its associated attributes. This study addresses not only generating DEMs and digital orthophotos but producing three dimensional building models from aerial photos of an urban area by employing the digital photogrammetric technology. DEMs and digital orthophotos were automatically generated through the process of orientations, image matching and so on, and then the practical problems, which must be solved especially in applying to urban areas, were considered. The accuracy of produced digital orthophotos was derived by using check points. Also three dimensional visualization imagery, which is useful in the landform analysis, and 3D building models were produced. Digital photogrammetric products would be used widely not only as GIS framework data layers by using the GIS link function which links attribute and image information in the database for applying to infrastructure management and but as geospatial data for especially 3D GIS in urban areas.

• Korean Journal of Remote Sensing
• /
• v.33 no.4
• /
• pp.423-436
• /
• 2017

This paper presents a deep learning-based road segmentation framework from very high-resolution orthophotos. The proposed method uses Deep Convolutional Autoencoders for end-to-end mapping of orthophotos to road segmentations. In addition, a set of post-processing steps were applied to make the model outputs GIS-ready data that could be useful for various applications. The optimization of the model's parameters is explained which was conducted via grid search method. The model was trained and implemented in Keras, a high-level deep learning framework run on top of Tensorflow. The results show that the proposed model with the best-obtained hyperparameters could segment road objects from orthophotos at an average accuracy of 88.5%. The results of optimization revealed that the best optimization algorithm and activation function for the studied task are Stochastic Gradient Descent (SGD) and Exponential Linear Unit (ELU), respectively. In addition, the best numbers of convolutional filters were found to be 8 for the first and second layers and 128 for the third and fourth layers of the proposed network architecture. Moreover, the analysis on the time complexity of the model showed that the model could be trained in 4 hours and 50 minutes on 1024 high-resolution images of size $106{\times}106pixels$, and segment road objects from similar size and resolution images in around 14 minutes. The results show that the deep learning models such as Convolutional Autoencoders could be a best alternative to traditional machine learning models for road segmentation from aerial photographs.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.36 no.3
• /
• pp.117-126
• /
• 2018

Due to intelligent transport systems, location-based applications, and augmented reality, demand for image maps and 3D (Three-Dimensional) maps is increasing. As a result, data acquisition using UAV (Unmanned Aerial Vehicles) has flourished in recent years. However, even though orthophoto map production and research using UAVs are flourishing, few studies on 3D modeling have been conducted. In this study, orthophoto and 3D modeling research was performed using various angle images acquired by a UAV. For orthophotos, accuracy was evaluated using a GPS (Global Positioning System) survey that employed VRS (Virtual Reference Station) acquired checkpoints. 3D modeling was evaluated by calculating the RMSE (Root Mean Square Error) of the difference between the outline height values of buildings obtained from the GPS survey to the corresponding 3D modeling height values. The orthophotos satisfied the acceptable accuracy of NGII (National Geographic Information Institute) for a 1/500 scale map from all angles. In the case of 3D modeling, models based on images taken at 45 degrees revealed better accuracy of building outlines than models based on images taken at 30, 60, or 75 degrees. To summarize, it was shown that for orthophotos, the accuracy for 1/500 maps was satisfied at all angles; for 3D modeling, images taken at 45 degrees produced the most accurate models.

• Korean Journal of Remote Sensing
• /
• v.35 no.6_1
• /
• pp.945-957
• /
• 2019

In this study, we propose a coarse to fine image registration method for eliminating geometric error between images over agricultural areas acquired using Unmanned Aerial Vehicle (UAV). First, images of agricultural area were acquired using UAV, and then orthophotos were generated. In order to reduce the probability of extracting outliers that cause errors during image registration, the region of interest is selected by using the metadata of the generated orthophotos to minimize the search area. The coarse image registration was performed based on the extracted tie-points using the Speeded-Up Robust Features (SURF) method to eliminate geometric error between orthophotos. Subsequently, the fine image registration was performed using tie-points extracted through the Mutual Information (MI) method, which can extract the tie-points effectively even if there is no outstanding spatial properties or structure in the image. To verify the effectiveness and superiority of the proposed method, a comparison analysis using 8 orthophotos was performed with the results of image registration using the SURF method and the MI method individually. As a result, we confirmed that the proposed method can effectively eliminated the geometric errors between the orthophotos.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.25 no.5
• /
• pp.415-425
• /
• 2007

In this study, we generated an orthophoto with both LIDAR data and aerial images and compared it with that generated from only the images. For the accuracy assessment of these orthophotos, we performed not only qualitative analysis based on visual inspection but also quantitative analysis by measuring horizontal inconsistency, boundary coordinates and similarity measures on buildings. Based on the visual inspection and horizontal inconsistency, the orthophoto based on LIDAR DSM appeared to be more closer to a true-orthophoto. However, the analysis on measurements of boundary coordinates and similarity measures indicates that the orthophoto based on LIDAR DSM is more vulnerable to double mapping on occluded areas. Accordingly, if we apply an effective solution on double mapping or use only the central areas of the aerial images where occluded areas are rarely founded, we can generate automatically true-orthophotos based on a LIDAR DSM.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.36 no.4
• /
• pp.223-233
• /
• 2018

In this study was classified the cadastral control points and parcel boundary points into 40m, 100m by flight altitude of UAV (Unmanned Aerial Vehicle) which compared the coordinates extracted from the orthophoto with the parcel boundary point coordinates by GNSS (Global Navigation Satellite System) ground survey. As a results of this study, first, in the spatial resolution analysis that the average error of the orthoimage by flight altitude were 0.024m at 40m, and 0.034m at 100m which were higher 40m than 100m for spatial resolution of orthophotos and position accuracy. Second, in order to analyze the accuracy of image recognition by airmark of flight altitude that was divided into three cases of nothing, green, and red of RMSE (Root Mean Square Error) were X=0.039m, Y=0.019m and Z=0.055m, the highest accuracy. Third, the result of the comparison between orthophotos and field survey results that showed the total RMSE error of the cadastral control points were X=0.029m, Y=0.028m, H=0.051m, and the parcel boundary points were X=0.041m, Y=0.030m. In conclusion, based on the results of this study, it is expected that if the average error of flight altitude is limited to less than 0.05m in the legal regulations related to orthophotos for cadastral surveying, it will be an economical and efficient method for cadastral survey as well as spatial information acquisition.

• Korean Journal of Remote Sensing
• /
• v.34 no.1
• /
• pp.45-74
• /
• 2018

Landslide is a natural hazard that threats lives and properties in many areas around the world. Landslides are difficult to recognize, particularly in rainforest regions. Thus, an accurate, detailed, and updated inventory map is required for landslide susceptibility, hazard, and risk analyses. The inconsistency in the results obtained using different features selection techniques in the literature has highlighted the importance of evaluating these techniques. Thus, in this study, six techniques of features selection were evaluated. Very-high-resolution LiDAR point clouds and orthophotos were acquired simultaneously in a rainforest area of Cameron Highlands, Malaysia by airborne laser scanning (LiDAR). A fuzzy-based segmentation parameter (FbSP optimizer) was used to optimize the segmentation parameters. Training samples were evaluated using a stratified random sampling method and set to 70% training samples. Two machine-learning algorithms, namely, Support Vector Machine (SVM) and Random Forest (RF), were used to evaluate the performance of each features selection algorithm. The overall accuracies of the SVM and RF models revealed that three of the six algorithms exhibited higher ranks in landslide detection. Results indicated that the classification accuracies of the RF classifier were higher than the SVM classifier using either all features or only the optimal features. The proposed techniques performed well in detecting the landslides in a rainforest area of Malaysia, and these techniques can be easily extended to similar regions.

• Korean Journal of Remote Sensing
• /
• v.38 no.5_1
• /
• pp.597-608
• /
• 2022

Remotely sensed data have been used in various fields, such as disasters, agriculture, urban planning, and the military. Recently, the demand for the multitemporal dataset with the high-spatial-resolution has increased. This manuscript proposed an automatic image matching algorithm using a deep learning technique to utilize a multitemporal remotely sensed dataset. The proposed deep learning model was based on High Resolution Net (HRNet), widely used in image segmentation. In this manuscript, denseblock was added to calculate the correlation map between images effectively and to increase learning efficiency. The training of the proposed model was performed using the multitemporal orthophotos of the National Geographic Information Institute (NGII). In order to evaluate the performance of image matching using a deep learning model, a comparative evaluation was performed. As a result of the experiment, the average horizontal error of the proposed algorithm based on 80% of the image matching rate was 3 pixels. At the same time, that of the Zero Normalized Cross-Correlation (ZNCC) was 25 pixels. In particular, it was confirmed that the proposed method is effective even in mountainous and farmland areas where the image changes according to vegetation growth. Therefore, it is expected that the proposed deep learning algorithm can perform relative image registration and image matching of a multitemporal remote sensed dataset.