Integrating spectral and textural information for identifying the tasseling date of summer maize using UAV based RGB images

• Published in: International journal of applied earth observation and geoinformation 2021-10, Vol.102, p.102435, Article 102435
• Main Authors: Guo, Yahui, Fu, Yongshuo H., Chen, Shouzhi, Robin Bryant, Christopher, Li, Xinxi, Senthilnath, J., Sun, Hongyong, Wang, Shuxin, Wu, Zhaofei, de Beurs, Kirsten
• Format: Article
• Language: English
• Subjects: Phenology extraction; Spectral information; Textural information; UAV
• ISSN: 1569-8432; 1872-826X
• DOI: 10.1016/j.jag.2021.102435

•Textural information was integrated to monitor the growth of summer maize.•The IAFWM was introduced by integrating the spectral and textural information.•The new index achieved the highest precision in extracting tasseling dates. The extraction of phenological events in forest and agriculture commonly relies on Vegetation Indices (VI) composed by visible and near infrared bands from satellite images. However, the textural information playing an important role in image fusion, image classification and change detection is commonly ignored. In this study, high-throughput images collected from an Unmanned Aerial Vehicle (UAV) platform during the growth stages of summer maize were used to identify the Tasseling Date (TD) based on both spectral and textural information. The spectral and textural information were extracted using various VI and the Gray Level Co-occurrence Matrix (GLCM), respectively. The results showed that the Normalized Green Blue Difference Index (NGBDI), and the Green Blue Difference Index (GBDI) of VI and the Contrast Information (Contrast) of GLCM performed better than other variables. A new index was generated by integrating spectral and textural information using the Improved Adaptive Feature Weighting Method (IAFWM), and then the TDs were identified for each plot. The Root Mean Square Error (RMSE) of new index was 5.77 days and it was the lowest among all variables. The potential ability of more advanced machine learning and deep learning in integrating the spectral and textural information should be investigated.