Due to the rapid artificial intelligence technology progress and innovation in various fields, this research aims to use science mapping tools to comprehensively and objectively analyze recent advances, hot-spots, and challenges in artificial intelligence-based microplastic-imaging field from the Web of Science (2019–2022). By text mining and visualization in the scientific literature we emphasized some opportunities to bring forward further explication and analysis by (i) exploring efficient and low-cost automatic quantification methods in the appearance properties of microplastics, such as shape, size, volume, and topology, (ii) investigating microplastics water-soluble synthetic polymers and interaction with other soil and water ecology environments via artificial intelligence technologies, (iii) advancing efficient artificial intelligence algorithms and models, even including intelligent robot technology, (iv) seeking to create and share robust data sets, such as spectral libraries and toxicity database and co-operation mechanism, (v) optimizing the existing deep learning models based on the readily available data set to balance the related algorithm performance and interpretability, (vi) facilitating Unmanned Aerial Vehicle technology coupled with artificial intelligence technologies and data sets in the mass quantities of microplastics. Our major findings were that the research of artificial intelligence methods to revolutionize environmental science was progressing toward multiple cross-cutting areas, dramatically increasing aspects of the ecology of plastisphere, microplastics toxicity, rapid identification, and volume assessment of microplastics. The above findings can not only determine the characteristics and track of scientific development, but also help to find suitable research opportunities to carry out more in-depth research with many problems remaining.

Knowledge distillation (KD) is a well-established technique for compressing neural networks and has gained increasing attention in object detection tasks. However, typical object detection distillation methods use fixed-level semantic features for distillation, which might not be best for all training stages and samples. In this paper, a multilayer semantic feature adaptive distillation (MSFAD) method is proposed that uses a routing network composed of a teacher and a student detector, along with an agent network for decision making. Specifically, the inputs to the proxy network consist of the features output by the neck structures of the teacher and student detectors, and the output is a decision on which features to choose for distillation. The MSFAD method improves the distillation training process by enabling the student detector to automatically select valuable semantic-level features from the teacher detector. Experimental results demonstrated that the proposed method increased the mAP₅₀ of YOLOv5s by 3.4% and the mAP_50–90 by 3.3%. Additionally, YOLOv5n with only 1.9 M parameters achieved detection performance comparable to that of YOLOv5s.

Semi-supervised learning is a learning pattern that can utilize labeled data and unlabeled data to train deep neural networks. In semi-supervised learning methods, self-training-based methods do not depend on a data augmentation strategy and have better generalization ability. However, their performance is limited by the accuracy of predicted pseudo-labels. In this paper, we propose to reduce the noise in the pseudo-labels from two aspects: the accuracy of predictions and the confidence of the predictions. For the first aspect, we propose a similarity graph structure learning (SGSL) model that considers the correlation between unlabeled and labeled samples, which facilitates the learning of more discriminative features and, thus, obtains more accurate predictions. For the second aspect, we propose an uncertainty-based graph convolutional network (UGCN), which can aggregate similar features based on the learned graph structure in the training phase, making the features more discriminative. It can also output the uncertainty of predictions in the pseudo-label generation phase, generating pseudo-labels only for unlabeled samples with low uncertainty; thus, reducing the noise in the pseudo-labels. Further, a positive and negative self-training framework is proposed, which combines the proposed SGSL model and UGCN into the self-training framework for end-to-end training. In addition, in order to introduce more supervised signals in the self-training process, negative pseudo-labels are generated for unlabeled samples with low prediction confidence, and then the positive and negative pseudo-labeled samples are trained together with a small number of labeled samples to improve the performance of semi-supervised learning. The code is available upon request.

Short-term hourly reliable prediction of significant wave height is an important research topic in coastal engineering. Many researchers have carried out in-depth studies in many ocean regions. Generally, most of this work is implemented through numerical models. However, as for numerical models, with the increase of hourly prediction duration, the accumulation of wave randomness leads to the poor prediction effect. In this paper, four buoy stations in the Taiwan Strait are taken as the research objects. We propose a significant wave height prediction algorithm, which combines numerical weather prediction model WRF and deep-learning model, called WRF-CLSF(Convolution-LSTM-FC). WRF can forecast 24-h wind speed in real-time, based on a variety of interpretable physical mechanisms. CLSF aims to extract the information of historical wind-wave interaction scale and the trend of wind-wave coherence, with the help of convolution operation and the time series model(LSTM), respectively. In the experiment, the proposed model was compared with the state-of-the-art prediction model. The results show that WRF-CLSF has an outstanding prediction effect at four buoy observation stations along with the Taiwan Strait.

Accurate prediction of Sea Surface Temperature (SST) plays a critical role in climate research and marine ecosystem management. Traditional models predict trends by analyzing and fitting data, but they struggle with capturing long-range dependencies and complex spatiotemporal patterns. The Transformer’s attention mechanism effectively addresses long-range dependencies, but its high computational complexity poses challenges. To overcome these limitations, this study proposes a novel spatiotemporal sequence prediction model: the Spatiotemporal Vision Mamba Recurrent Neural Network (SVRNN). The model innovatively integrates a bidirectional state-space processing mechanism and decoupled memory modules. The bidirectional mechanism maintains a global receptive field with linear computational complexity, while the decoupled memory modules explicitly separate spatiotemporal dependencies, enhancing the model’s ability to capture complex spatiotemporal patterns. During the experiment on hourly SST prediction in the Taiwan Strait, where the SST of the next 12 hours was predicted using data from the previous 12 hours, the SVRNN model demonstrated superior performance, achieving a Root Mean Square Error (RMSE) of 0.159°C, a Mean Absolute Error (MAE) of 0.105°C, and a Mean Absolute Percentage Error (MAPE) of 0.496%. Furthermore, our seasonal error analysis reveals that the model exhibits robust performance in different seasons, providing more reliable technical support for SST prediction in the Taiwan Strait.