Xin Song

Ship image classification identifies the type of ships in an input image, which plays a significant role in the marine field. To enhance the ship classification performance, various research focuses on studying multi-modal ship classification, which aims at combining the advantages of visible images and infrared images to capture complementary information. However, the current methods simply concatenate features of different modalities to learn complementary information, which neglects the multi-level correlation between different modalities. Moreover, the existing methods require a large amount of labeled ship images to train the model. How to capture the multi-level cross-modal correlation between unlabeled and labeled data is still a challenge. In this paper, a novel semi-supervised multi-modal ship classification approach is proposed to solve these issues, which consists of two components, i.e., multi-level cross-modal interactive network and semi-supervised contrastive learning strategy. To learn comprehensive complementary information for classification, the multi-level cross-modal interactive network is designed to build local-level and global-level cross-modal feature correlation. Then, the semi-supervised contrastive learning strategy is employed to drive the optimization of the network with the intra-class consistency constraint based on supervision signals of unlabeled samples and prior label information. Extensive experiments on the public datasets demonstrate that our approach achieves state-of-the-art semi-supervised classification effectiveness.

Cryogenic extraction of water from plant and soil samples constitutes a critical element in isotope eco-hydrological studies. This study was undertaken to examine the stability and reliability of a new generation of water extraction technology based on cryogenic vacuum distillation (CVD)-automatic cryogenic vacuum distillation water extraction system, and to optimize the extraction time and improve the extraction efficiency for this system, as well as to explore the applicability of this system. For these purposes, water extraction from tap water and leaves and stems of four species of woody plants was carried out by using manual and automatic CVD systems and then the hydrogen and oxygen isotope values (δ²H and δ¹⁸O) of water extracted from the two systems were compared. Besides, rehydrated leaves, stems and soils were extracted by manual CVD system and automatic CVD system under different extraction time lengths, then the δ²H、δ¹⁸O of water extracted with the two systems were compared too. Moreover, we conducted a separate experiment to use the automatic CVD system to confirm the depth of water absorption and the utilization ratio of source water of Osmanthus fragrans (Thunb.) Loureiro. The results are as follows. There is no significant difference in the hydrogen and oxygen isotope values (δ²H, δ¹⁸O) of water extracted with the two systems and that data obtained from the two methods are strongly linearly correlated, with the correlation line being statistically indistinguishable from the 1:1 line. The δ²H and δ¹⁸O values of rehydrated sample water extracted from leaves, stems and soils for 1 h, 2 h, and 2 h by automatic CVD system are not statistically different from those extracted by manual CVD system. The depth of water uptake and the proportion of water use of Osmanthus fragrans (Thunb.) Loureiro determined based on δ²H is inconsistent with that determined based on δ¹⁸O. Therefore, this study systematically verifies for the first time that there is no difference between automatic CVD system and manual CVD system, which provides a guarantee of reliability in the related application of isotope eco-hydrology; further, this study enables us to establish recommendations on optimizing the extraction time and improving the extraction efficiency for the automatic CVD system. Meanwhile, our finding of that the δ²H of water extracted from the rehydrated samples were more depleted than the reference water adds evidence to the recently revealed phenomenon of "deuterium deviation" as caused by CVD in the ecohydrological field; as such, we suggest more caution in utilizing CVD-generated δ²H data to trace plant water sources and determine the water use strategies in plant and ecosystem studies.