Photo-realistic texture mapping of 3D object is one of the major subjects in the study of landscape or city modeling. This paper presents an automated method to map realistic texture onto the facades of 3D building models from multiple overlapped images. This method works on condition that the geometric representation of buildings and the image orientation are known. Images used for texture mapping are mainly obtained from close-range photogrammetry. Rigorous mapping function is developed and implemented with the consideration of selecting the most appropriate images to obtain the best mapping texture. Two examples of building structures are demonstrated. The results will be expressed in Virtual Reality Modeling Language （VRML） so that they can be shown with VRML-compatible visualization software.

Airborne light detection and ranging (LiDAR) is an active remote sensing technique, which sends pulses of laser light toward the ground and detects the return times of back-scattered energy in order to determine ranges of the surface. The suitability of LiDAR in forestry is demonstrated by application such as assessment of timber resources and biomass, quantification of 3D canopy structures, as well as derivation of single trees properties like individual tree positions and tree height. In this study, we used the single tree height data was investigated in the Alishan area and LAI data was investigated in the Chitou and Alishan area, and use airborne LiDAR data to measured mean tree height, analyze the relationship between LAI and Laser Penetration Index (LPI) suitable raster cell size and estimation model. The results of the study indicated that the laser single tree height overestimates the ground truth tree height, especially in compression tree. The laser mean height is computed as the arithmetic mean of the largest laser values within grid size of 15m, and the results was R2 = 0.993. The results was R2 = 0.979 when the LiDAR raster grid size was 15× 15 m 2. Therefore, LiDAR data is useful to estimate the LAI. The estimate map of LAI was obtained according to the results of regression analysis, and it can be found that the LiDAR point cloud of three-dimensional structure made the LAI values between forested region and non-forested region had significant differences. It shows that the LiDAR data have better detection ability for the forest canopy. It provides a new remote sensing technique for forestry investigation in Taiwan.

Geometric correction of remote sensing imagery is indispensable for the purpose of data integration and applications. Conventional ortho- recti-fication only considers the correction of tilt displacements and terrain relief displacements. Since high-resolution imagery captures more detailed information, the abrupt changes of object elevation result in the relief displacements and occlusions in images. In urban areas, high-rise buildings and multi-layer road systems of orthoimages cause serious image deformation. True-orthorectification procedures take account of ground objects for the correction of relief displacements and compensation of the hidden areas. Accordingly, the true orthoimages provide the exact position of land objects. This research focuses on the man-made constructions, including buildings and roads. Thus, the generation of true orthoimages is done by the integration of building, road, and terrain models. The test data comprises three dimensional building models, road models and digital terrain models. Multi-view aerial images are employed for the compensation of occlusions. The proposed scheme includes: (1) visibility analysis, (2) hidden detection, (3) hidden compensation, and (4) shadow enhancement. UltraCam-D imagery is used to validate the proposed scheme. The experimental results indicate that the proposed methods may produce quality true-ortho images with the accuracy between 1 and 2 pixels.

In this study, the wavelet approach is used for airborne LiDAR data filtering. An initial digital surface model (DSM) is generated by wavelet approach. The method utilizes then geodesic dilation, height difference and second order difference of heights of neighboring grid points, and then re-selects LiDAR points by generating a reference surface to select more reliable terrain points. Finally, those selected terrain points are used to generate the digital elevation model (DEM) in that area. In the process, digital aerial color images and their derived feature figures such as NDVI and GI can be also applied. In addition, a method is proposed to exclude those non-terrain points in a bridge region. A point scanning method is presented to sort the seeds located on the bridge surface. These seeds can be adopted for region growing process. Then all non-terrain points within the bridge area can be found. Both test area I with simpler surface topography and test area II with more complicated surface topography as well as break lines show that the total error of LiDAR point classification is 0.5% and 6%, respectively. They verify the method is applicable for LiDAR point filtering and DEM generation.

This paper is concerning detection and characterization of wave breaking using X-band pulsed Doppler microwave radar nearby the Shuang-Si estuary at Fu-Long village in Taiwan. Along the same measuring line, as the character of subjective scatterers under the radar illuminated area changes, and the features of radar backscatter will be different. The analysis of radar backscatter involved radar cross section, Doppler frequency shift, and Doppler bandwidth from the sea surface is accomplished. Eventually, the maximum radar cross section and Doppler frequency shift on the same measuring line, connect the position on every measuring line respectively. The radar system was shown the capability to delineate the location of breaking wave clearly and that is sufficient for nearshore events of air-sea interactions to achieve a measurement operation. It’s safe and secure for investigators and instruments that they accomplish all detective operation successfully.

For remotely sensed images, region growing segmentation is a widely-used technique for object extraction and identification. However, the main drawback of region growing segmentation is the threshold setting for stopping the growth of a region. Improper threshold setting not only causes over-segmentation or under-segmentation, but also cannot create needed object regions for various targets. In order to overcome these drawbacks, a multi-scale region growing segmentation method, based on the maximization of the change of edge density, is proposed. Experiments, including different kinds of high resolution remotely sensed images are used to test the performance of the proposed scheme. The experimental results show that the proposed scheme can both remove the limitation of threshold setting and generate relatively reasonable segmentation output for different types of objects.