Location: Water Management and Systems ResearchTitle: Enhancing model performance in detecting lodging areas in wheat fields using UAV RGB imagery: Considering spatial and temporal variations
|ZHANG, DONGYAN - Northwest A&f University
|ZHANG, GAN - Anhui Agricultural University
|TAO, CHENG - Anhui Agricultural University
Submitted to: Computers and Electronics in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/29/2023
Publication Date: 10/14/2023
Citation: Zhang, D., Zhang, G., Tao, C., Zhang, H. 2023. Enhancing model performance in detecting lodging areas in wheat fields using UAV RGB imagery: Considering spatial and temporal variations. Computers and Electronics in Agriculture. 214. Article e108297. https://doi.org/10.1016/j.compag.2023.108297.
Interpretive Summary: Detecting and measuring lodging areas in wheat fields quickly and accurately is important for preventing disasters and settling insurance claims. However, it's challenging because lodging detection can vary in different places and times, causing inconsistent results in the images. In this study, we developed a method to evaluate lodging area detection models over multiple years and growth stages. We used three indicators—accuracy, potential, and stability—to assess the models and combined the results using a weighted method. We evaluated different models, like CBAM-unet, SE-unet, Swin-transformer, and Unet, using images from wheat fields over three years and two key growth stages. The results showed that Swin-transformer had the highest accuracy and potential but needed some improvements for better predictions. CBAM-unet was the most stable among the models. Our research aimed to solve the challenges of evaluating lodging area detection and improve its accuracy and reliability in wheat fields.
Technical Abstract: Fast and accurate assessment of wheat lodging area holds significant importance for disaster prevention and agricultural insurance claim settlement. However, lodging detection is often influenced by the spatial and temporal heterogeneity of fields, leading to inconsistent lodging features in images captured at different times and locations. This inconsistency poses challenges for evaluating model performance and obtaining consistent results. In this study, we proposed a comprehensive lodging area detection model performance evaluation method for multi-year and multi-phenological periods. Three evaluation indicators—accuracy, potential, and stability—were introduced to assess model performance from different perspectives. Additionally, a weighted method is employed to combine the results of multiple experiments. To evaluate the performance of lodging area segmentation in wheat fields, we selected three feature adaptive models (CBAM-unet, SE-unet, and Swin-transformer) along with the Unet model. The dataset consisted of images acquired over three years and during two key phenological periods. The experimental results are as follows: (1) Swin-transformer achieved the highest weighted average accuracy (Acc_cp) of 83.53% among the four models evaluated. (2) Swin-transformer exhibited the highest upper limit of segmentation accuracy, reaching 97.98%. However, its average accuracy in the prediction experiment was 85% of the upper limit, suggesting potential for optimization. (3) CBAM-unet demonstrated the lowest overall weighted segmentation accuracy variance (35.92) compared to the other three models, indicating higher stability. Through this research, we aimed to address the challenges associated with evaluating lodging area detection models in the presence of spatial and temporal heterogeneity. By proposing a comprehensive evaluation method and selecting the most suitable feature adaptive model, we sought to enhance the accuracy and reliability of lodging area detection in wheat fields.