Researchers explored a number of approaches for wheat lodging detection, starting with taking advantage of satellite images
[23-25]. Though the performance of satellite images on crop lodging has been validated
[8,26], its wide adoption as a universal approach is challenging due to its low spatial (finest as tens of centimeters) and coarse temporal (multiple days) resolution. Since the satellite images cannot meet the practical application requirements, researchers started to explore other potential technologies. Most recently, researchers started to explore the potential of using unmanned aerial system (UAS) images for crop lodging detection given the recent development on UASs, sensors, and associated software for data processing, coupled with new and effective machine learning (ML) and deep learning (DL) algorithms
[27]. Compared to the satellite images, the UASs images have several advantages. First, the resolution of the collected UAS aerial images is much finer than those by satellite, with some of them reaching millimeter level. In addition, UASs allow for a shorter revisit time of the same area (temporal resolution) compared to satellite, given weather conditions are suitable for flights, allowing even for more than one flight a day over the same area of interest. During the past few years, there has been a sharply increase on research exploring the use of UASs for crop lodging—though it is at a nascent stage, study results have demonstrated the potential on different crops, such as sugar beet, canola, and wheat
[28,29]. So far, a majority of the studies treat crop lodging as a binary issue—lodging or non-lodging. From a plant breeder perspective, though the lodging and non-lodging information is crucial, it would be more meaningful to know the lodging as a percentage of each experimental unit (plot), which can be a key piece of information when making breeding decisions. However, few studies have been conducted for this specific purpose.