Journal of Taiwan Agricultural Research

Author:Chiao-Ying Chou, Shi-Yang Wu, and Chi-Ling Chen*

Abstract:

    The photography from unmanned aerial vehicles (UAV) provides multispectral images (i.e., red, green, blue, and near-infrared bands) and 3-dimensional points cloud with high-spatial-resolution and covering wide-region. In order to provide the maps for geographic information system (GIS) in assisting the agricultural post-disaster investigation, the goal of this study is to discover the most effective agriculture damage interpretation by using image discrimination technology with the characteristics of vehicle speed, convenience, and accuracy. In this study, image interpretation technologies, including image classification and digital surface model (DSM) classification, will be evaluated on the accuracy of rice lodging detection. Firstly, Pix4D Mapper, the professional photogrammetry and drone mapping software, is used to produce DSM and orthophotographs from the photography of UAV. Secondly, the image supervised classification, normalized difference vegetation index (NDVI) classification, and DSM classification are applied for detection of rice lodging in Wufeng District, Taichung City. The study area was damaged by the extremely torrential rain in a few days during the beginning of June, 2017. As a result, the damage region and the situation of rice lodging can be delineated by UAV orthophotographs. This study, especially on wide-region post-disaster investigation, recommends adoption of the image supervised classification on rice lodging detection, because the rate of accuracy between the disaster rate of estimation by image interpretation and the disaster rate of ground-based surveillance system reached 92.54% (under the baseline of 20% of disaster rate in accordance with the Implementation Rules of Agricultural Natural Disaster Relief). In addition, the accuracy of disaster interpretation can be improved by cooperating with the ground-based surveillance system. In conclusion, the agricultural post-disaster detection and rescue operations will be improved by the cooperation between new technologies and traditional labor-force. The capabilities and potentials from both sides need to be brought in to compensate one’s shortcomings by using the individual strengths. With that, the efficiency and high-quality of damage detection can be achieved at least cost.

Key words:Agricultural damage, Image supervised classification, Digital surface model (DSM), Normalized difference vegetation index (NDVI), Disaster aid

Download: PDF Links