Open Cotton Boll Detection Methodology Using Unmanned Aerial System (UAS) Imagery

Unmanned Aerial System (UAS) images have great potential for agriculture and can be used for various applications. Especially in case of agricultural researches, UAS images facilitate precise data collection in crop fields for detailed analysis. Additionally, crops can be monitored with short time intervals using UASs, unlike most other remote sensing platforms. In this study, an open cotton boll detection methodology using UAS imagery is proposed. We used a Phanthom 4 platform and its standard integrated sensor. Raw oblique imagery was used to detect cotton bolls not shown in orthomosaic image. Random seed distribution and region growing methods were adopted to detect homogeneous regions. Their spatial characteristics including size and roundness were then analyzed to acquire meaningful spectral information of initial cotton boll candidates. Finally, multispectral threshold values of candidate bolls are derived from the Otsu method and combination of thresholding is applied to the image. The proposed method shows good results in the visual comparison with the original image. Most of all, the proposed method can contribute to automated cotton boll detection and aid crop yield estimation.