Xianchang Wang

Kernel number and row structure traits, including kernels per ear, rows per ear, and kernels per row, are essential indicators for evaluating maize ear development and yield-related performance. However, conventional manual measurement is labor-intensive, time-consuming, and susceptible to subjective variation, limiting the efficiency of large-scale phenotyping in maize breeding programs. Single and multiview twodimensional imagebased algorithms are sensitive to the dense and inclined arrangement of maize kernels. This leads to discontinuities and correspondence errors during row reconstruction, degrades the estimation accuracy of kernelrelated traits, and constrains the throughput of highthroughput phenotyping in maize breeding. To overcome these limitations, this study developed an integrated rotational RGB imaging and computational phenotyping framework for automated and non-redundant quantification of maize ear structural traits. A panoramic localisation method was proposed to determine the image interval corresponding to one complete ear circumference and automatically position cropping boundaries between adjacent kernel rows, enabling accurate reconstruction of the ear surface without repeated regions. Furthermore, a directional feature density estimation network (DFDNet) was developed by integrating directional feature enhancement and a dual-branch regression architecture to predict kernel density distributions from point annotations. Based on the predicted density maps, a slope-aware row reconstruction algorithm was established to extract kernel row trajectories and quantify row-level phenotypic traits. A dataset containing 244 maize ear panoramas and 133,109 annotated kernel points was constructed for model training and evaluation. The proposed panorama localisation method achieved an ROI IoU of 0.97 ± 0.02. DFDNet achieved high-precision kernel counting with an MAE of 6.396, RMSE of 8.541, MAPE of 1.17%, and R² of 0.9908, outperforming five comparative models. The proposed framework also achieved MAEs of 0.417 and 0.665 for rows per ear and kernels per row estimation, respectively. An automated software platform was further developed to integrate panoramic reconstruction, trait visualisation, and quantitative phenotype extraction from RGB video acquisition. This study provides an efficient and scalable solution for full-surface maize ear phenotyping and demonstrates its potential for high-throughput yield-related trait measurement in precision breeding applications.

Semisupervised object detection (SSOD) has garnered significant interest for its capability to enhance the detection performance by leveraging large amounts of unlabeled data. However, current SSOD methods primarily focus on detecting horizontal objects, with little research devoted to the detection of arbitrary-oriented objects in remote sensing images. Drawing inspiration from this limitation, this article proposes a semisupervised oriented object detection framework (S $^{2}$ O-Det) to reduce annotation costs while improving detection performance in a semisupervised manner. Initially, the proposed task-consistent learning aims to alleviate the inconsistencies between classification and localization, which provides consistent confidence for the pseudolabels. Subsequently, the introduced coarse-to-fine sample mining employs dense prediction for pseudolabel assignment, adopting a divide-and-conquer approach to independently identify consistent and reliable labels for both classification and localization tasks. Finally, a probabilistic distillation loss ensures the harmonization of the probability distributions across the teacher and student feature domains, thereby reciprocally enhancing the learning competencies. Experimental results on the DOTA-v1.0 and DOTA-v1.5 datasets demonstrate that S $^{2}$ O-Det achieves promising performance across different labeling ratios.

In the field of Earth observation and computer vision, oriented object detection for remotesensing images is a crucial task that aims to locate objects more accurately in complex scenes containing a large number of densely arranged, large aspect ratio, and arbitrarily oriented objects. Although recently proposed methods have achieved remarkable performance, there are still several challenges to address: 1) interference from complex backgrounds, 2) imbalanced and mismatched label assignments caused by tiny objects and objects with large aspect ratios, and 3) misalignment between the tasks of classification and localization. In this article, we propose a frequency-assisted dynamic learning network (FADL-Net) to overcome the crucial challenges. Concretely, we introduce a spatial-spectral feature pyramid network to adaptively capture global long-range dependency feature representations containing various frequency domains. Meanwhile, to produce more reliable training samples for objects with extreme shapes, we design a geometric aware dynamic label assignment to dynamically mitigate the imbalance and mismatch in label assignment in a coarse-to-fine manner, thereby achieving more stable optimization during the training process. Moreover, we propose a joint-learning rotated quality loss that addresses the inconsistency between classification and localization by dynamically adapting the weights of different samples in the training stage. Extensive experiments on several public remote sensing datasets demonstrate that our method performs favorably against state-of-the-art detection approaches.

Hyperspectral imaging is a key technology for non-destructive detection of seed vigor presently due to its capability to capture variations of optical properties in seeds. As the seed vigor data depends on the actual germination rate, it inevitably results in an imbalance between positive and negative samples. Additionally, hyperspectral image (HSI) suffers from feature redundancy and collinearity due to its inclusion of hundreds of wavelengths. It also creates a challenge to extract effective wavelength information in feature selection, however, which limits the ability of deep learning to extract features from HSI and accurately predict seed vigor. Accordingly, in this paper, we proposed a Focal-WAResNet network to predict seed vigor end-to-end, which improves the network performance and feature representation capability, and improves the accuracy of seed vigor prediction. Firstly, the focal loss function is utilized to adjust the loss weights of different sample categories to solve the problem of sample imbalance. Secondly, a WAResNet network is proposed to select characteristic wavelengths and predict seed vigor end-to-end, focusing on wavelengths with higher network weights, which enhance the ability of seed vigor prediction. To validate the effectiveness of this method, this study collected HSI of maize seeds for experimental verification, providing a reference for plant breeding. The experimental results demonstrate a significant improvement in classification performance compared to other state-of-the-art methods, with an accuracy up to 98.48% and an F1 score of 95.9%.

In the prediction of time series, Empirical Mode Decomposition (EMD) generates subsequences and separates short-term tendencies from long-term ones. However, a single prediction model, including attention mechanism, has varying effects on each subsequence. To accurately capture the regularities of subsequences using an attention mechanism, we propose an integrated model for time series prediction based on signal decomposition and two attention mechanisms. This model combines the results of three networks—LSTM, LSTM-self-attention, and LSTM-temporal attention—all trained using subsequences obtained from EMD. Additionally, since previous research on EMD has been limited to single series analysis, this paper includes multiple series by employing two data pre-processing methods: ‘overall normalization’ and ‘respective normalization’. Experimental results on various datasets demonstrate that compared to models without attention mechanisms, temporal attention improves the prediction accuracy of short- and medium-term decomposed series by 15~28% and 45~72%, respectively; furthermore, it reduces the overall prediction error by 10~17%. The integrated model with temporal attention achieves a reduction in error of approximately 0.3%, primarily when compared to models utilizing only general forms of attention mechanisms. Moreover, after normalizing multiple series separately, the predictive performance is equivalent to that achieved for individual series.

Panax quinquefolium, as a common precious medicinal plant, has complex chemical components and unique pharmacological activities, which can play a healthcare role in the human body. With the deepening of research, the application of P. quinquefolium has become increasingly extensive. This paper summarized the research progress of the saponins isolated and identified from diffe-rent parts of P. quinquefolium, the structural classification and pharmacological activities of the saponins, and the quality control of Panacis Quinquefolii Radix. Further, this paper put forward the urgent problems to be solved in the development of P. quinquefolium. It is hoped to lay a foundation for the further study and provide reference for the research direction of P. quinquefolium.