Few-shot segmentation (FSS) is a cutting-edge technology that can meet requirements using a small workload. With the development of China Aerospace Engineering, FSS plays a fundamental role in astronaut working scene (AWS) intelligent parsing. Although mainstream FSS methods have made considerable breakthroughs in natural data, they are not suitable for AWSs. AWSs are characterized by a similar foreground (FG) and background (BG), indistinguishable categories, and the strong influence of light, all of which place higher demands on FSS methods. We design a pixel-wise and class-wise network (PCNet) to match support and query features using pixel-wise and class-wise semantic cues. Specifically, PCNet extracts pixel-wise semantic information at each layer of the backbone using novel cross-attention. Dense prototypes are further utilized to extract class-wise semantic cues as a supplement. In addition, the deep prototype is distilled in reverse to the shallow layer to improve its quality. Furthermore, we customize a dataset for AWSs and conduct abundant experiments. The results indicate that PCNet outperforms the published best method by 4.34% and 5.15% in accuracy under one-shot and five-shot settings, respectively. Moreover, PCNet compares favorably with the traditional semantic segmentation model under the 13-shot setting.
The application of intelligent video monitoring for natural resource protection and management has become increasingly common in recent years. To enhance safety monitoring during the grazing prohibition and rest period of grassland, this paper proposes a multi-object segmentation algorithm based on mid-infrared images for all-weather surveillance. The approach integrates the Image Timing Features–Gaussian Mixture Model (ITF-GMM) and Convolutional-UNet (Con-UNet) to improve the accuracy of target detection. First, a robust background modelling, i.e., the ITF-GMM, is proposed. Unlike the basic Gaussian Mixture Model (GMM), the proposed model dynamically adjusts the learning rate according to the content difference between adjacent frames and optimizes the number of Gaussian distributions through time series histogram analysis of pixels. Second, a segmentation framework based on Con-UNet is developed to improve the feature extraction ability of UNet. In this model, the maximum pooling layer is replaced with a convolutional layer, addressing the challenge of limited training data and improving the network’s ability to preserve spatial features. Finally, an integrated computation strategy is designed to combine the outputs of ITF-GMM and Con-UNet at the pixel level, and morphological operations are performed to refine the segmentation results and suppress noises, ensuring clearer object boundaries. The experimental results show the effectiveness of proposed approach, achieving a precision of 96.92%, an accuracy of 99.87%, an intersection over union (IOU) of 94.81%, and a recall of 97.75%. Furthermore, the proposed algorithm meets real-time processing requirements, confirming its capability to enhance small-target detection in complex outdoor environments and supporting the automation of grassland monitoring and enforcement.