Accurate diagnosis of crop water demand is a core challenge in alleviating agricultural water scarcity. Traditional methods rely on soil moisture sensors or empirical models based on meteorological data, and have significant limitations. Therefore, developing real-time, non-destructive, and precise diagnostic technologies that reflect the crop's own water status will be crucial. In recent years, the high-throughput phenotyping technology has advanced rapidly and provided revolutionary tools to address this challenge. This paper explores the use of this technology to capture water-sensitive phenotypic traits of crops under water stress and construct water demand diagnosis models for real-time irrigation decision-making. By systematically reviewing research progress, technical methods, modeling strategies, and existing challenges, this study aims to provide theoretical support for precision irrigation and smart agriculture.