| Research projects |
| Principal Investigator |
Project name |
Funding Agency |
Project No. |
Period |
Abstract |
keywords |
| YISHUO HUANG |
Estimation of Flooded Areas by Integrating Satellite Observation Information—Taking the Zhuoshuixi River Basin as an Example |
National Science and Technology Council |
|
2022/08/01~2023/07/31 |
In this study, a variety of satellite images were integrated and fused to obtain possible flooding areas. Based on the floodplain area, terrain information and water level information, the flood volume is obtained. Yin Jun's in-depth learning, training and learning modules from previous cases, predicts the possible flooding areas when the rain exceeds the warning value. This research can reduce the impact range of disasters, especially floods, and can be used for disaster prevention. |
|
| YISHUO HUANG |
Image Segmentation, Intensity Inhomogeneity, Thin Clouds. |
National Science and Technology Council |
MOST 109-2121-M-324-001- |
2019/8/1-2020/7/31 |
Removing clouds is an important issue in satellite image interpretation. Intensity inhomogeneity means that the pixel values are contaminated by noise, shadow, etc., and it can be estimated in a multiplicative way. In this paper, the satellite image containing thin clouds are treated as intensity inhomogeneity; it means thin clouds contaminate the pixel values of the given satellite images. Hence, where the thin cloud is? How to remove thin clouds? Those raised questions can be answered by employing the image segmentation techniques. Traditionally, image segmentation can usually cluster the given image into few classes, but it is difficult to correctly segment the given image while intensity inhomogeneity exists. An approach modeling the thin clouds in a multiplicative way is proposed, and the thin cloud effects can be approximated by finding the convergence till the given image is fully segmented. The image segmentation cooperating the multiplicative model is numerically implemented by introducing level set function and iteration scheme. In doing so, the image can be simplified to the limited regions such that the differences between the segmented and given images can reach a minimum. The idea of removing thin clouds based on segmentation is to replace the given image with the segmented image such that the intensity inhomogeneity can be approximated with the differences between the given image and the segmented image. Several test areas covered with thin clouds are evaluated, and the processed results demonstrate the proposed approach can remove thin clouds efficiently. |
Image Segmentation, Intensity Inhomogeneity, Thin Clouds. |
| YISHUO HUANG |
Integrating 3D building models and infrared thermal images for structural defect detection research |
National Science and Technology Council |
MOST 106-2119-M-324-001 |
2017/8/1-2018/7/31 |
Infrared thermography is used to record surface temperature data. The recorded surface temperature data does not offer much information about the defects present on the building facades unless the data contained in the given thermographs can be processed and analyzed. The potential defect locations can usually be located in areas where the surface temperature is higher than the surrounding area. The recorded surface temperature data can be corrupted by the surrounding environmental conditions; shadows and sun glare can affect the surface temperature data in the thermograph such that the recorded surface temperature is higher or lower than expected. The effects have a profound influence on identifying the defects because of the corrupted thermography. This paper proposes an approach based on the multiplicative model to approximate the shadow or glare effects as well as develop the optimal approximation of the segmented results by employing the principle of total variation. In the segmented results, the data distribution is homogeneous in each segmented region and is estimated by the average value of all the pixels in the segmented region. With the segmented regions, the temperature |
Thermograph, Defect Location, Image Segmentation |
