Assessing the impact of soil and field conditions on cotton crop emergence using UAV-based imagery

Authors: TIAN, FENGKAI - University Of Missouri; Ransom, Curtis; ZHOU, JIANFENG - University Of Missouri; WILSON, BRADLEY - University Of Missouri; Sudduth, Kenneth - Ken

Submitted to: Computers and Electronics in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/9/2024
Publication Date: 2/17/2024
Citation: Tian, F., Ransom, C.J., Zhou, J., Wilson, B., Sudduth, K.A. 2024. Assessing the impact of soil and field conditions on cotton crop emergence using UAV-based imagery. Computers and Electronics in Agriculture. 218. Article 108738. https://doi.org/10.1016/j.compag.2024.108738
DOI: https://doi.org/10.1016/j.compag.2024.108738

Interpretive Summary: Optimizing cotton seeding rates is one method to maximize yield production. However, optimum seeding rates will vary based on environmental factors like soil texture and landscape position. Determining the optimal seeding rate has been difficult on large fields because it requires manually counting plants which is time intensive. Using aerial images taken with an unmanned aerial vehicle (UAV) is a more efficient method but there is limited research relating images to stand counts. Therefore, the objective of this research was to develop methods for extracting stand count data from images, determine if soil texture or elevation explained any differences in stand counts, and determine the optimal cotton seeding rate for different parts of an example field. Findings showed that cotton seeding rates were accurately estimated (>91% accuracy) using various machine learning techniques. Furthermore, it was found that cotton had improved germination and seed spacing in areas of higher sand content and higher elevation. The ideal seeding rate was found to be between 108,000 and 123,000 seeds ha-1.

Technical Abstract: Crop seeding rate is one of the crucial factors that affect crop production. However, acquiring adequate data of crops in multiple growing environments is time-consuming and challenging in large fields. The objective of this research was to develop and evaluate an efficient method using an unmanned aerial vehicle (UAV) imaging system with the deep learning technique to assess cotton emergence spacing uniformity at different seeding rates. A study was conducted using two cotton varieties and five seeding rates (56k, 74k, 91k, 108k, and 123k seeds ha-1), with each treatment containing four rows with three replicates in a random block design. A UAV-based imaging system was used to collect RGB images of the research field at flight heights of 10 m and 15 m above ground level at two and six weeks after planting, respectively. UAV images were stitched to build orthomosaic images segmented into small blocks. The object detection algorithm YOLOv7 was applied to identify cotton plants from each segmented image. In addition, Hough transform and polynomial regression methods were used to identify each cotton row and remove weeds. To evaluate the impact of soil conditions on cotton emergence, stand count was calculated in 5-m row segments to correlate with soil electrical conductivity (ECa) and elevation. Results show that the research could detect cotton plants with the mean average precision of 96.9% at the 50% intersection over the union threshold (mAP@50) for the two-week dataset and 92.7% mAP@50 for the six-week dataset. The results also indicate that plant uniformity was closely correlated with field elevation and ECa, with an average R2 of 0.62 using the random forest algorithm. The coefficient of variation was used to evaluate the spacing uniformity of each seeding rate and demonstrated that the 108k and 123k seeds ha-1 planting rates tended to exhibit better spacing uniformity than others under various environmental conditions. This study provides valuable insights by developing a pipeline for early-stage cotton stand count using high-resolution remote sensing techniques to evaluate the uniformity of different seeding rates for cotton, ultimately improving the efficiency of crop management.