Professor National Taiwan University Taipei, Taipei, Taiwan (Republic of China)
Body of Abstract: Choosing the appropriate seedlings and plant types is important for plant growth and horticultural design. Intelligent recognition technology can assist selecting plants that meet the specific needs. Nowadays, numerous applications and tools utilize artificial intelligence and machine learning techniques to identify plants. These tools are widely used in identifying plant species, understanding their growth requirements, and learning how to care for plants in diverse environments. By using artificial intelligence technology, the risk of selecting unsuitable plants can be minimized and plan and design of the gardening projects can be enhanced. In this paper, the intelligent recognition for seedling and plant morphology selection is studied, combining visual perception through image analysis with artificial intelligence. It aims to further enhance the detection and selection of plant morphologies. The seedlings of Oncidium, which are cultivated in a greenhouse, are investigated. This provides a diverse range of orchids at different growth stages for comprehensive data collection and further exploration. Regarding data processing, the initial step involves background removal from all images followed by labeling. This process improves accuracy compared to images without background removal, as it reduces background noise interference. During the cultivation of orchids, they typically undergo two to three repottings. The purpose of repotting is to enhance nutrient absorption, prevent the roots from escaping the pot, and accelerate plant growth. However, repotting also requires manual labor for selecting and repotting the seedlings. Traditionally, farmers determine the repotting frequency based on the passage of time. However, with the majority of orchids now being cultivated in greenhouses, their growth rate is significantly faster. Consequently, the frequency of repotting in the short term increases. Therefore, by using image recognition to directly identify the seedlings that require repotting, the time spent on manual selection can be reduced.