Monitor cotton budding using SVM and UAV images

Authors: XIA, LANG - NATIONAL ENGINEERING RESEARCH CENTER FOR INFORMATION TECHNOLOGY IN AGRICULTURE; ZHANG, RUIRUI - NATIONAL ENGINEERING RESEARCH CENTER FOR INFORMATION TECHNOLOGY IN AGRICULTURE; CHEN, LIPING - NATIONAL ENGINEERING RESEARCH CENTER FOR INFORMATION TECHNOLOGY IN AGRICULTURE; Huang, Yanbo; XU, GANG - NATIONAL ENGINEERING RESEARCH CENTER FOR INFORMATION TECHNOLOGY IN AGRICULTURE; WEN, YAO - NATIONAL ENGINEERING RESEARCH CENTER FOR INFORMATION TECHNOLOGY IN AGRICULTURE; YI, TONGCHUAN - NATIONAL ENGINEERING RESEARCH CENTER FOR INFORMATION TECHNOLOGY IN AGRICULTURE

Submitted to: Applied Sciences
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/30/2019
Publication Date: 10/14/2019
Citation: Xia, L., Zhang, R., Chen, L., Huang, Y., Xu, G., Wen, Y., Yi, T. 2019. Monitor cotton budding using SVM and UAV images. Applied Sciences. 9(4312): 1-13. https://doi.org/10.3390/app9204312.
DOI: https://doi.org/10.3390/app9204312

Interpretive Summary: Rapid and timely monitoring cotton germination rate is important for cotton growers. Scientists in National Engineering Research Center for Information Technology in Agriculture in Beijing, China mounted a true-color camera on an unmanned aerial vehicle and collected images of young cotton plants in a field in Xinjiang, China to estimate the germination of cotton plants. In a collaborative research with a scientist of USDA-ARS Crop Production Systems Research Unit at Stoneville, Mississippi the collected images were processed and classified to identify the cotton plants from the images. In image processing with an improved method a 6.3% increase of plant detection accuracy was obtained in comparison with using the conventional image processing method so that the eventual plant detection accuracy reached over 91%. With the plastic film covered in the field to avoid weed infestation the developed UAV imaging system observed a relatively low cotton germination rate of 56.26%, which illustrates the effectiveness of the system for practical application.

Technical Abstract: Monitoring the cotton budding rate is important for growers, so that they can replant cotton timely at the locations in which cotton density is sparse. In this study, a true-color camera was mounted on an unmanned aerial vehicle and used to collect images of young cotton plants to estimate the germination of cotton plants. The collected images were preprocessed by stitching them together to obtain the single orthomosaic image, and the maximum likelihood classification was conducted to identify the cotton plants in the image. In order to avoid the influence of overlapping cotton plants, which often leads to underestimation of the cotton budding rate, a method based on the maximal and minimal lengths of a single cotton plant is proposed and developed in this study to identify and count the overlapped cotton plants. A 6.3% increase of detection accuracy was obtained when using this newly developed method compared to not using it. The validation based on visual interpretation indicated that the new method presented an accuracy of 91.13%. Due to the coverage of plastic film in the field, the study is unaffected by weeds. However, the covered plastic film hinders the unearthing of the seeds and makes less cotton plants observed by the camera, as a result, the monitored budding rate in the study is relatively low, with a rate of 56.26%.