Determination of rice leaf midrib deflection in field environment by using semantic segmentation and shortest distance algorithm

• Published in: Computers and electronics in agriculture 2023-12, Vol.215, p.108326, Article 108326
• Main Authors: Luo, Jiaqi, Dai, Baisheng, Chang, Penghao, Gao, Rui, Su, Zhongbin
• Format: Article
• Language: English
• Subjects: age determination; agricultural colleges; agricultural extension; agriculture; algorithms; China; electronics; leaves; photography; rice
• ISSN: 0168-1699
• DOI: 10.1016/j.compag.2023.108326

In order to automatically and accurately determine rice leaf age in a complex field environment, a method for determining rice leaf vein deflection is proposed. The method is based on semantic segmentation and a shortest distance algorithm. First, a semantic segmentation model named RLNet is constructed to segment leaf structure curved regions (LSCR). These regions are then clustered and merged using a coordinate-based algorithm, and leaf structure curved lines (LSCL) are fitted by regions. Finally, the shortest distance algorithm is used to determine the deflection of the middle leaf structure curved line, from which the rice leaf age is obtained. Unlike commonly used methods for rice leaf age determination, the proposed method does not require complex manual assistance and harsh experimental conditions, it only requires photography of rice leaves in a field environment. To evaluate the practicality and effectiveness of the method, an experiment was conducted at the Agricultural Extension Center of the Northeast Agricultural University, Harbin, Heilongjiang, China. The method segmented LSCR with an accuracy of 84.71% and determined the rice leaf midrib deflection with an accuracy of 95%. Experimental results showed that the proposed method could determine rice leaf age automatically and accurately in a field environment and thereby reduce dependence on professionals.