With the widespread use of aerial digital cameras such as the DMC, ADS40, RMK-D, and UltraCamD, high dynamic range imaging is generally expected for generating minuteness orthophotos in digital aerial photogrammetry. However, high dynamic range images (12-bit, 4,096 gray levels) are generally compressed into an 8-bit depth digital image (256 gray levels) owing to huge amount of data and interface with peripherals such as monitors and printers. This means that a great deal of image data is eliminated from the original image, and this introduces a new shadow problem. In particular, the influence of shadows in urban areas causes serious problems for generating minuteness orthophotos and performing house detection. In order to resolve the shadow problem in digital aerial photogrammetry, shadow areas should be recognized and corrected automatically without the loss of luminance information. With this motive, a practical shadow correction method for 8-bit depth digital image which is derived from 12-bit real data of DMC is investigated in this paper.

View full abstract

This paper describes on an accuracy investigation of point clouds which generated by MMS (Mobile Mapping System). In recent years 3D measurement technologies have advanced significantly. At the same time, MMS have become common measurement instruments for 3D application fields such as road facility management and related fields. We have to consider that the necessity of establishment positional accuracy evaluation methodology stands on the measurement mechanism of MMS point clouds. The aim of this study is to establish positional accuracy evaluation methodology for MMS point cloud measurement data and examine our LS3D (Least Squares 3D Surface matching) approach in practical experiments. Our proposed evaluation methodology consists of (1) Data Quality Assessment, (2) Precision Assessment, (3) Absolute Accuracy Assessment, (4) Relative Assessment, (5) Cross-sensor Accuracy Assessment. In order to examine our evaluation methodology, we had been carried out an actual MMS run and MMS point clouds generation at Minato-Mirai, Yokohama. Through our practical experiments and an actual evaluation, we confirmed that our proposed positional accuracy evaluation methodology is good enough procedures. As farther works, we have to accumulate much more evaluations than this moment.

View full abstract

According to appearance of low cost and high resolution digital cameras with various functions, a convenient 3D measurement using the digital cameras are enormously expected in various fields. In these circumstances, the authors have been concentrating on developing a convenient 3D measurement method using the digital cameras. Software “3DiVision” has been designed to perform convenient 3D measurement under the key words ; 3Dimension, Digital image and Visualization, and “3DiVision” shown the capability to perform camera calibration without GCPs and subsequent 3D measurement.
However, there is still issue for realize a convenient digital close range photogrammetry. The main problem is robust calibration procedure for triplet images which are taken at various distances from camera to object and ill-balance exposure stations. In order to resolve the issue, robust camera calibration method which is able to accomplish a convenient digital close range photogrammetry using triplet images of ununiform photo scales and ill-balance exposure stations are proposed in this paper.

View full abstract

This paper proposes the classification framework based on the Bayesian theory with the single polarization multi-temporal synthetic aperture radar (SAR) and an optical data, and incorporates the proposed training sample selection (SS) methods. Within this framework, the combination with gray level co-occurrence matrix (GLCM)-based textural measures is investigated. The two procedures of the classification and proposed SS are united, where SS generates the accurate and dispersed training samples. Extracted features from multi-temporal SAR data—namely, the average backscattering coefficient, the backscatter temporal variability, and the long-term coherence and the reflectance values from optical data, are integrated with the GLCM textural data. Classification results were generated by taking Osaka City, Japan, as the study area. The selected major classes were water bodies, woodlands, fields, and built-up areas. The most suitable data used for classification was the multi-temporal SAR and an optical data combination with the mean textural, because of the supplement of different data and the smoothing effect of the texture. The higher quality training samples obtained by using the combined Support Vector Machine (SVM) and Neural Network (NN)-based SS method for training the Bayesian classifier generated the highest classification accuracies in all of tested cases.

View full abstract

Forest classification map is considered as very important data in a forest management and recommended to be produced covering the whole country for public use. Therefore, it is suggested that more easy and efficient new technique is developed for producing detailed forest classification map using the remote sensing image with high resolution. Here, the object-based classification is mentioned to represent this new technique. In the object-based classification a homogeneous pixel domain in the remote sensing image is first created, and classification process is done using the statistics of a homogeneous pixel domain. Generally color information and image texture used in industrial image processing are contained in the statistics. Image texture is the numerical value that evaluated the difference in the pattern of a homogeneous pixel domain, so that it is difficult to understand the relation between image texture and actual state of the forest in situ. In order to utilize the object-based classification in a forest management it is required to make the algorithm for the object-based classification easy to understand.
Since the spatial distribution of trees reflects its growth, thinning and spacing treatment, regeneration etc, it is possible the spatial distribution can be used as classification rule for object based classification. This research examined the validity of the classification, which used brightness peak in the remote sensing image related to the spatial distribution of trees. The overall accuracy of the proposal method was 0.68. Although this result was not sufficient, the proposal method was effective in classifying early growth stage of cedar and other forest state.

View full abstract

Currently, DEM (Digital Elevation Model) of high density and wide range has been able to be created by spread of airborne laser scanner data. Because the DEM has a large amount of data, real-time communication becomes impossible. In order to reduce the amount of data of the DEM, we propose a method to compress the DEM with joint singular value decomposition. The proposed method is a lossy compression, and assume that a terrain has similar relief. We prove the effectiveness of the proposed method through the experiments using the DEM of 5m mesh.

View full abstract