Infrared and visible image fusion is to combine the information of thermal radiation and detailed texture from the two images into one informative fused image. Recently, deep learning methods have been widely applied in this task; however, those methods usually fuse multiple extracted features with the same fusion strategy, which ignores the differences in the representation of these features, resulting in the loss of information in the fusion process. To address this issue, we propose a novel method named multi-modal feature self-adaptive transformer (MFST) to preserve more significant information about the source images. Firstly, multi-modal features are extracted from the input images by a convolutional neural network (CNN). Then, these features are fused by the focal transformer blocks that can be trained through an adaptive fusion strategy according to the characteristics of different features. Finally, the fused features and saliency information of the infrared image are considered to obtain the fused image. The proposed fusion framework is evaluated on TNO, LLVIP, and FLIR datasets with various scenes. Experimental results demonstrate that our method outperforms several state-of-the-art methods in terms of subjective and objective evaluation.

Tibetan medicinal materials play a significant role in Tibetan culture. However, some types of Tibetan medicinal materials share similar shapes and colors, but possess different medicinal properties and functions. The incorrect use of such medicinal materials may lead to poisoning, delayed treatment, and potentially severe consequences for patients. Historically, the identification of ellipsoid-like herbaceous Tibetan medicinal materials has relied on manual identification methods, including observation, touching, tasting, and nasal smell, which heavily rely on the technicians’ accumulated experience and are prone to errors. In this paper, we propose an image-recognition method for ellipsoid-like herbaceous Tibetan medicinal materials that combines texture feature extraction and a deep-learning network. We created an image dataset consisting of 3200 images of 18 types of ellipsoid-like Tibetan medicinal materials. Due to the complex background and high similarity in the shape and color of the ellipsoid-like herbaceous Tibetan medicinal materials in the images, we conducted a multi-feature fusion experiment on the shape, color, and texture features of these materials. To leverage the importance of texture features, we utilized an improved LBP (local binary pattern) algorithm to encode the texture features extracted by the Gabor algorithm. We inputted the final features into the DenseNet network to recognize the images of the ellipsoid-like herbaceous Tibetan medicinal materials. Our approach focuses on extracting important texture information while ignoring irrelevant information such as background clutter to eliminate interference and improve recognition performance. The experimental results show that our proposed method achieved a recognition accuracy of 93.67% on the original dataset and 95.11% on the augmented dataset. In conclusion, our proposed method could aid in the identification and authentication of ellipsoid-like herbaceous Tibetan medicinal materials, reducing errors and ensuring the safe use of Tibetan medicinal materials in healthcare.

Tibetan murals are a cultural heritage form rich in historical and artistic information, but many surviving images are affected by cracks, flaking, pigment loss, abrasion, and other deterioration. Accurately segmenting damaged regions is challenging because of complex decorative textures, irregular degradation morphology, domain discrepancy, and limited pixel-level annotations. This study proposes an unsupervised domain-adaptive damage segmentation framework that transfers damage knowledge from labeled Dunhuang murals to unlabeled Tibetan mural images. The framework integrates DINOv3-based multi-layer visual representations, an FPN-inspired lightweight decoder, and teacher-student pseudo-label adaptation with dynamic filtering and target-domain reweighting. Experiments use 961 labeled Dunhuang mural image-mask pairs from MuralDH and 906 Tibetan mural images. On the held-out Tibetan test set, the method achieves 61.54% Dice and 44.45% IoU, supporting low-annotation damage localization for Tibetan mural preservation.

Error confusion sets refer to collections of words or characters that, within the same group, are similar in spelling and grammar to the corresponding words or characters in the correct text, thereby easily causing confusion. Such confusion sets are commonly used in automatic text correction systems, where the system replaces potential errors with other candidates from the confusion set to enhance the accuracy of the correction. This paper introduces a method for constructing Tibetan syllable confusion sets aimed at text correction, and validates the confusion sets in experiments on Tibetan text correction. The Tibetan syllable confusion sets presented in this paper are composed of three distinct subsets based on the characteristics of Tibetan text: a confusion subset based on Tibetan syllable component information and edit distance, a confusion subset based on input key position distance, and a confusion subset based on Tibetan grammatical knowledge. These subsets are merged sequentially to form a comprehensive Tibetan syllable confusion set for text correction, totaling 59.2M in size. Experiments on public data with this confusion set have shown that the accuracy of the Bert+ confusion set correction method is 10.63% higher than that of a standalone Bert model, demonstrating the effectiveness of the Tibetan syl-lable confusion set in Tibetan text correction tasks.

Research on Tibetan dialect identification is crucial for exploring the linguistic features of various dialects, advancing intelligent speech technology, and enriching linguistic theory. The construction of a high-quality speech data set forms the foundation of such research. This study integrates insights from Tibetan linguistics and speech signal processing to propose a method for building a data set specifically designed for Tibetan dialect identification and quantification research. First, based on the Tibetan phonemic system, example words and sentences were designed to include Tibetan vowels, consonants, and the “eight-case” grammatical structure. Second, to address the problem of low-quality speech data, an automatic detection method based on signal-to-noise ratio (SNR), pitch, speech rate, and other indicators was introduced to identify and handle abnormal samples. Third, to overcome the issue of limited pronunciation styles in some dialect regions, the GPT-Sovits 4.0 voice cloning model was employed to enrich pronunciation variations. As a result, a speech data set comprising 56,760 samples and approximately 39.8 hours of speech data was developed. Comparative experiments with publicly available data sets of similar scale demonstrate that the constructed data set improves Tibetan dialect identification accuracy by 6.12% (using the k-nearest neighbor algorithm) and 12.27% (using long short-term memory networks) compared to existing public data sets. These findings indicate that the newly constructed data set more effectively captures the phonetic distinctions among Tibetan dialects and enhances performance in dialect identification tasks.