Irrigation Water Requirement of Complicated Agricultural Land by Using of Airborne Digital Sensors Image

• Main Authors: HSU, CHIA-SHENG, 徐家盛
• Other Authors: CHEN, LI
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/kv66gu

博士 === 中華大學 === 土木工程學系 === 105 === Due to the impact of global climate change, rainfall in Taiwan tends to be extreme and rainfall days decrease. However, the frequency of heavy rainfall increases. In the agricultural irrigation area adjacent to urban areas, crop types and water demand need to be more accurately grasped to balance agricultural, industry and domestic water usage. Particularly, the land use of irrigated area nearby the metropolitan is complex. There may be different crop growth on a field so as the water demand is different. It is necessary to accurately distinguish among the crops, so that the water demand can be calculated accurately. Therefore the water supply can accurately meet the demand of crops and save excess water resources. In recent years, the image classification method has been changed from the traditional pixel-based classification to the combined object oriented classification. Compared with traditional pixel-based method which is tend to have spot-like pepper and salt effect, the object-based classification can reduce the salt and pepper effect significantly, also greatly simplify the amount of data required for analysis. In order to obtain better spectral recognition result, most of image information is described by color, texture and shape to enhance image recognition results. In this study, image information extraction and crop interpretation were carried out by using ADS40 as the experimental data, and comparing the traditional supervised classification, the artificial classification of cadastral-parcel units and the image object classification model. The accuracy of the classification of crops was compared by different methods, and irrigation water was estimated using the crop acreage. The results showed that the three image classification methods could achieve the overall accuracy of higher than 80%, and the object-based classification was the highest, reaching 88.68%. In the study area, the daily water demand of the crops was 2,586 m3 per day, and the difference between the methods was about 4 ~ 5%.