Rice lodging severely affects harvest yield. Traditional evaluation methods and manual on-site measurement are found to be time-consuming, labor-intensive, and cost-intensive. In this study, a new method for rice lodging assessment based on a deep learning UNet (U-shaped Network) architecture was proposed. The UAV (unmanned aerial vehicle) equipped with a high-resolution digital camera and a three-band multispectral camera synchronously was used to collect lodged and non-lodged rice images at an altitude of 100 m. After splicing and cropping the original images, the datasets with the lodged and non-lodged rice image samples were established by augmenting for building a UNet model. The research results showed that the dice coefficients in RGB (Red, Green and Blue) image and multispectral image test set were 0.9442 and 0.9284, respectively. The rice lodging recognition effect using the RGB images without feature extraction is better than that of multispectral images. The findings of this study are useful for rice lodging investigations by different optical sensors, which can provide an important method for large-area, high-efficiency, and low-cost rice lodging monitoring research.

IntroductionFine-tuning (FT) is a generally adopted transfer learning method for deep learning-based magnetic resonance imaging (MRI) reconstruction. In this approach, the reconstruction model is initialized with pre-trained weights derived from a source domain with ample data and subsequently updated with limited data from the target domain. However, the direct full-weight update strategy can pose the risk of "catastrophic forgetting" and overfitting, hindering its effectiveness. The goal of this study is to develop a zero-weight update transfer strategy to preserve pre-trained generic knowledge and reduce overfitting.MethodsBased on the commonality between the source and target domains, we assume a linear transformation relationship of the optimal model weights from the source domain to the target domain. Accordingly, we propose a novel transfer strategy, linear fine-tuning (LFT), which introduces scaling and shifting (SS) factors into the pre-trained model. In contrast to FT, LFT only updates SS factors in the transfer phase, while the pre-trained weights remain fixed.ResultsTo evaluate the proposed LFT, we designed three different transfer scenarios and conducted a comparative analysis of FT, LFT, and other methods at various sampling rates and data volumes. In the transfer scenario between different contrasts, LFT outperforms typical transfer strategies at various sampling rates and considerably reduces artifacts on reconstructed images. In transfer scenarios between different slice directions or anatomical structures, LFT surpasses the FT method, particularly when the target domain contains a decreasing number of training images, with a maximum improvement of up to 2.06 dB (5.89%) in peak signal-to-noise ratio.DiscussionThe LFT strategy shows great potential to address the issues of "catastrophic forgetting" and overfitting in transfer scenarios for MRI reconstruction, while reducing the reliance on the amount of data in the target domain. Linear fine-tuning is expected to shorten the development cycle of reconstruction models for adapting complicated clinical scenarios, thereby enhancing the clinical applicability of deep MRI reconstruction.